diff options
| author | Radosław Kujawa <radoslaw.kujawa@c0ff33.net> | 2026-07-17 14:54:32 +0200 |
|---|---|---|
| committer | Radosław Kujawa <radoslaw.kujawa@c0ff33.net> | 2026-07-17 14:54:32 +0200 |
| commit | 2702c90d7a39fca2a3319c5e915fd290cb31fc70 (patch) | |
| tree | 283c4d851efef7e3a7cd03ba80c87e2bffcdbb2c /src/mesa/x86 | |
| parent | 836167945ec2c7ad5acdf0aa17ce35ec1b9428a6 (diff) | |
Initial import.
Diffstat (limited to 'src/mesa/x86')
36 files changed, 12134 insertions, 0 deletions
diff --git a/src/mesa/x86/3dnow.c b/src/mesa/x86/3dnow.c new file mode 100644 index 0000000..c0a4e7a --- /dev/null +++ b/src/mesa/x86/3dnow.c @@ -0,0 +1,92 @@ +/* $Id: 3dnow.c,v 1.23 2003/03/29 16:38:37 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 5.0.1 + * + * Copyright (C) 1999-2003 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * 3DNow! optimizations contributed by + * Holger Waechtler <holger@akaflieg.extern.tu-berlin.de> + */ + +#include "glheader.h" +#include "context.h" +#include "math/m_xform.h" +#include "tnl/t_context.h" + +#include "3dnow.h" +#include "common_x86_macros.h" + +#ifdef DEBUG +#include "math/m_debug.h" +#endif + + +#ifdef USE_3DNOW_ASM +DECLARE_XFORM_GROUP( 3dnow, 2 ) +DECLARE_XFORM_GROUP( 3dnow, 3 ) +DECLARE_XFORM_GROUP( 3dnow, 4 ) + +DECLARE_NORM_GROUP( 3dnow ) + + +extern void _ASMAPI +_mesa_v16_3dnow_general_xform( GLfloat *first_vert, + const GLfloat *m, + const GLfloat *src, + GLuint src_stride, + GLuint count ); + +extern void _ASMAPI +_mesa_3dnow_project_vertices( GLfloat *first, + GLfloat *last, + const GLfloat *m, + GLuint stride ); + +extern void _ASMAPI +_mesa_3dnow_project_clipped_vertices( GLfloat *first, + GLfloat *last, + const GLfloat *m, + GLuint stride, + const GLubyte *clipmask ); +#endif + + +void _mesa_init_3dnow_transform_asm( void ) +{ +#ifdef USE_3DNOW_ASM + ASSIGN_XFORM_GROUP( 3dnow, 2 ); + ASSIGN_XFORM_GROUP( 3dnow, 3 ); + ASSIGN_XFORM_GROUP( 3dnow, 4 ); + + /* There's a bug somewhere in the 3dnow_normal.S file that causes + * bad shading. Disable for now. + ASSIGN_NORM_GROUP( 3dnow ); + */ + +#ifdef DEBUG + _math_test_all_transform_functions( "3DNow!" ); + _math_test_all_normal_transform_functions( "3DNow!" ); +#endif +#endif +} diff --git a/src/mesa/x86/3dnow.h b/src/mesa/x86/3dnow.h new file mode 100644 index 0000000..1f2fd8e --- /dev/null +++ b/src/mesa/x86/3dnow.h @@ -0,0 +1,39 @@ +/* $Id: 3dnow.h,v 1.6 2002/04/09 14:58:03 keithw Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * 3DNow! optimizations contributed by + * Holger Waechtler <holger@akaflieg.extern.tu-berlin.de> + */ + +#ifndef __3DNOW_H__ +#define __3DNOW_H__ + +#include "math/m_xform.h" + +void _mesa_init_3dnow_transform_asm( void ); + +#endif diff --git a/src/mesa/x86/3dnow_normal.S b/src/mesa/x86/3dnow_normal.S new file mode 100644 index 0000000..4c1aca2 --- /dev/null +++ b/src/mesa/x86/3dnow_normal.S @@ -0,0 +1,840 @@ +/* $Id: 3dnow_normal.S,v 1.8 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2003 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * 3Dnow assembly code by Holger Waechtler + */ + +#ifdef USE_3DNOW_ASM + +#include "matypes.h" +#include "norm_args.h" + + SEG_TEXT + +#define M(i) REGOFF(i * 4, ECX) +#define STRIDE REGOFF(12, ESI) + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_normalize_normals) +GLNAME(_mesa_3dnow_transform_normalize_normals): + +#define FRAME_OFFSET 12 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + PUSH_L ( EBP ) + + MOV_L ( ARG_LENGTHS, EDI ) + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EBP ) /* dest->count = in->count */ + MOV_L ( EBP, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + + CMP_L ( CONST(0), EBP ) /* count > 0 ?? */ + JE ( LLBL (G3TN_end) ) + + MOV_L ( REGOFF (V4F_COUNT, ESI), EBP ) + FEMMS + + PUSH_L ( EBP ) + PUSH_L ( EAX ) + PUSH_L ( EDX ) /* save counter & pointer for */ + /* the normalize pass */ +#undef FRAME_OFFSET +#define FRAME_OFFSET 24 + + MOVQ ( M(0), MM3 ) /* m1 | m0 */ + MOVQ ( M(4), MM4 ) /* m5 | m4 */ + + MOVD ( M(2), MM5 ) /* | m2 */ + PUNPCKLDQ ( M(6), MM5 ) /* m6 | m2 */ + + MOVQ ( M(8), MM6 ) /* m9 | m8 */ + MOVQ ( M(10), MM7 ) /* | m10 */ + + CMP_L ( CONST(0), EDI ) /* lengths == 0 ? */ + JNE ( LLBL (G3TN_scale_end ) ) + + MOVD ( ARG_SCALE, MM0 ) /* | scale */ + PUNPCKLDQ ( MM0, MM0 ) /* scale | scale */ + + PFMUL ( MM0, MM3 ) /* scale * m1 | scale * m0 */ + PFMUL ( MM0, MM4 ) /* scale * m5 | scale * m4 */ + PFMUL ( MM0, MM5 ) /* scale * m6 | scale * m2 */ + PFMUL ( MM0, MM6 ) /* scale * m9 | scale * m8 */ + PFMUL ( MM0, MM7 ) /* | scale * m10 */ + +ALIGNTEXT32 +LLBL (G3TN_scale_end): +LLBL (G3TN_transform): + MOVQ ( REGIND (EDX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF (8, EDX), MM2 ) /* | x2 */ + + MOVQ ( MM0, MM1 ) /* x1 | x0 */ + PUNPCKLDQ ( MM2, MM2 ) /* x2 | x2 */ + + PFMUL ( MM3, MM0 ) /* x1*m1 | x0*m0 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PREFETCHW ( REGIND(EAX) ) + + PFMUL ( MM4, MM1 ) /* x1*m5 | x0*m4 */ + PFACC ( MM1, MM0 ) /* x0*m4+x1*m5 | x0*m0+x1*m1 */ + + PFMUL ( MM5, MM2 ) /* x2*m6 | x2*m2 */ + PFADD ( MM2, MM0 ) /* x0*m4+x1*m5+x2*m6| x0*m0+...+x2**/ + + MOVQ ( REGIND (EDX), MM1 ) /* x1 | x0 */ + MOVQ ( MM0, REGOFF(-16, EAX) ) /* write r0, r1 */ + + PFMUL ( MM6, MM1 ) /* x1*m9 | x0*m8 */ + MOVD ( REGOFF (8, EDX), MM2 ) /* | x2 */ + + PFMUL ( MM7, MM2 ) /* | x2*m10 */ + PFACC ( MM1, MM1 ) /* *not used* | x0*m8+x1*m9 */ + + PFADD ( MM2, MM1 ) /* *not used* | x0*m8+x1*m9+x2*m*/ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + MOVD ( MM1, REGOFF(-8, EAX) ) /* write r2 */ + DEC_L ( EBP ) /* decrement normal counter */ + JA ( LLBL (G3TN_transform) ) + + + POP_L ( EDX ) /* end of transform --- */ + POP_L ( EAX ) /* now normalizing ... */ + POP_L ( EBP ) + + CMP_L ( CONST(0), EDI ) /* lengths == 0 ? */ + JE ( LLBL (G3TN_norm ) ) /* calculate lengths */ + + +ALIGNTEXT32 +LLBL (G3TN_norm_w_lengths): + + PREFETCHW ( REGOFF(12,EAX) ) + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM1 ) /* | x2 */ + + MOVD ( REGIND (EDI), MM3 ) /* | length (x) */ + PFMUL ( MM3, MM1 ) /* | x2 (normalize*/ + + PUNPCKLDQ ( MM3, MM3 ) /* length (x) | length (x) */ + PFMUL ( MM3, MM0 ) /* x1 (normalized) | x0 (normalize*/ + + ADD_L ( STRIDE, EDX ) /* next normal */ + ADD_L ( CONST(4), EDI ) /* next length */ + + PREFETCH ( REGIND(EDI) ) + + MOVQ ( MM0, REGIND(EAX) ) /* write new x0, x1 */ + MOVD ( MM1, REGOFF(8, EAX) ) /* write new x2 */ + + ADD_L ( CONST(16), EAX ) /* next r */ + DEC_L ( EBP ) /* decrement normal counter */ + + JA ( LLBL (G3TN_norm_w_lengths) ) + JMP ( LLBL (G3TN_exit_3dnow) ) + +ALIGNTEXT32 +LLBL (G3TN_norm): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND (EAX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM1 ) /* | x2 */ + + MOVQ ( MM0, MM3 ) /* x1 | x0 */ + MOVQ ( MM1, MM4 ) /* | x2 */ + + PFMUL ( MM0, MM3 ) /* x1*x1 | x0*x0 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PFMUL ( MM1, MM4 ) /* | x2*x2 */ + PFADD ( MM4, MM3 ) /* | x0*x0+x2*x2 */ + + PFACC ( MM3, MM3 ) /* **not used** | x0*x0+x1*x1+x2**/ + PFRSQRT ( MM3, MM5 ) /* 1/sqrt (x0*x0+x1*x1+x2*x2) */ + + MOVQ ( MM5, MM4 ) + PUNPCKLDQ ( MM3, MM3 ) + + DEC_L ( EBP ) /* decrement normal counter */ + PFMUL ( MM5, MM5 ) + + PFRSQIT1 ( MM3, MM5 ) + PFRCPIT2 ( MM4, MM5 ) + + PFMUL ( MM5, MM0 ) /* x1 (normalized) | x0 (normalize*/ + + MOVQ ( MM0, REGOFF(-16, EAX) ) /* write new x0, x1 */ + PFMUL ( MM5, MM1 ) /* | x2 (normalize*/ + + MOVD ( MM1, REGOFF(-8, EAX) ) /* write new x2 */ + JA ( LLBL (G3TN_norm) ) + +LLBL (G3TN_exit_3dnow): + FEMMS + +LLBL (G3TN_end): + POP_L ( EBP ) + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_normalize_normals_no_rot) +GLNAME(_mesa_3dnow_transform_normalize_normals_no_rot): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 12 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + PUSH_L ( EBP ) + + MOV_L ( ARG_LENGTHS, EDI ) + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EBP ) /* dest->count = in->count */ + MOV_L ( EBP, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + + CMP_L ( CONST(0), EBP ) /* count > 0 ?? */ + JE ( LLBL (G3TNNR_end) ) + + FEMMS + + MOVD ( M(0), MM0 ) /* | m0 */ + PUNPCKLDQ ( M(5), MM0 ) /* m5 | m0 */ + + MOVD ( M(10), MM2 ) /* | m10 */ + PUNPCKLDQ ( MM2, MM2 ) /* m10 | m10 */ + + CMP_L ( CONST(0), EDI ) /* lengths == 0 ? */ + JNE ( LLBL (G3TNNR_scale_end ) ) + + MOVD ( ARG_SCALE, MM7 ) /* | scale */ + PUNPCKLDQ ( MM7, MM7 ) /* scale | scale */ + + PFMUL ( MM7, MM0 ) /* scale * m5 | scale * m0 */ + PFMUL ( MM7, MM2 ) /* scale * m10 | scale * m10 */ + +ALIGNTEXT32 +LLBL (G3TNNR_scale_end): + CMP_L ( CONST(0), EDI ) /* lengths == 0 ? */ + JE ( LLBL (G3TNNR_norm) ) /* need to calculate lengths */ + + MOVD ( REGIND(EDI), MM3 ) /* | length (x) */ + + +ALIGNTEXT32 +LLBL (G3TNNR_norm_w_lengths): /* use precalculated lengths */ + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM6 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM7 ) /* | x2 */ + + PFMUL ( MM0, MM6 ) /* x1*m5 | x0*m0 */ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + PFMUL ( MM2, MM7 ) /* | x2*m10 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PFMUL ( MM3, MM7 ) /* | x2 (normalized) */ + PUNPCKLDQ ( MM3, MM3 ) /* length (x) | length (x) */ + + ADD_L ( CONST(4), EDI ) /* next length */ + PFMUL ( MM3, MM6 ) /* x1 (normalized) | x0 (normalized) */ + + DEC_L ( EBP ) /* decrement normal counter */ + MOVQ ( MM6, REGOFF(-16, EAX) ) /* write r0, r1 */ + + MOVD ( MM7, REGOFF(-8, EAX) ) /* write r2 */ + MOVD ( REGIND(EDI), MM3 ) /* | length (x) */ + + JA ( LLBL (G3TNNR_norm_w_lengths) ) + JMP ( LLBL (G3TNNR_exit_3dnow) ) + +ALIGNTEXT32 +LLBL (G3TNNR_norm): /* need to calculate lengths */ + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM6 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM7 ) /* | x2 */ + + PFMUL ( MM0, MM6 ) /* x1*m5 | x0*m0 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PFMUL ( MM2, MM7 ) /* | x2*m10 */ + MOVQ ( MM6, MM3 ) /* x1 (transformed)| x0 (transformed) */ + + MOVQ ( MM7, MM4 ) /* | x2 (transformed) */ + PFMUL ( MM6, MM3 ) /* x1*x1 | x0*x0 */ + + + PFMUL ( MM7, MM4 ) /* | x2*x2 */ + PFACC ( MM3, MM3 ) /* **not used** | x0*x0+x1*x1 */ + + PFADD ( MM4, MM3 ) /* | x0*x0+x1*x1+x2*x2*/ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + PFRSQRT ( MM3, MM5 ) /* 1/sqrt (x0*x0+x1*x1+x2*x2) */ + MOVQ ( MM5, MM4 ) + + PUNPCKLDQ ( MM3, MM3 ) + PFMUL ( MM5, MM5 ) + + PFRSQIT1 ( MM3, MM5 ) + DEC_L ( EBP ) /* decrement normal counter */ + + PFRCPIT2 ( MM4, MM5 ) + PFMUL ( MM5, MM6 ) /* x1 (normalized) | x0 (normalized) */ + + MOVQ ( MM6, REGOFF(-16, EAX) ) /* write r0, r1 */ + PFMUL ( MM5, MM7 ) /* | x2 (normalized) */ + + MOVD ( MM7, REGOFF(-8, EAX) ) /* write r2 */ + JA ( LLBL (G3TNNR_norm) ) + + +LLBL (G3TNNR_exit_3dnow): + FEMMS + +LLBL (G3TNNR_end): + POP_L ( EBP ) + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_rescale_normals_no_rot) +GLNAME(_mesa_3dnow_transform_rescale_normals_no_rot): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 12 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + PUSH_L ( EBP ) + + MOV_L ( ARG_IN, EAX ) + MOV_L ( ARG_DEST, EDX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EBP ) /* dest->count = in->count */ + MOV_L ( EBP, REGOFF(V4F_COUNT, EDX) ) + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + MOV_L ( REGOFF(V4F_START, EDX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + + CMP_L ( CONST(0), EBP ) + JE ( LLBL (G3TRNR_end) ) + + FEMMS + + MOVD ( ARG_SCALE, MM6 ) /* | scale */ + PUNPCKLDQ ( MM6, MM6 ) /* scale | scale */ + + MOVD ( REGIND(ECX), MM0 ) /* | m0 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m5 | m0 */ + + PFMUL ( MM6, MM0 ) /* scale*m5 | scale*m0 */ + MOVD ( REGOFF(40, ECX), MM2 ) /* | m10 */ + + PFMUL ( MM6, MM2 ) /* | scale*m10 */ + +ALIGNTEXT32 +LLBL (G3TRNR_rescale): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM4 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM5 ) /* | x2 */ + + PFMUL ( MM0, MM4 ) /* x1*m5 | x0*m0 */ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + PFMUL ( MM2, MM5 ) /* | x2*m10 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + DEC_L ( EBP ) /* decrement normal counter */ + MOVQ ( MM4, REGOFF(-16, EAX) ) /* write r0, r1 */ + + MOVD ( MM5, REGOFF(-8, EAX) ) /* write r2 */ + JA ( LLBL (G3TRNR_rescale) ) /* cnt > 0 ? -> process next normal */ + + FEMMS + +LLBL (G3TRNR_end): + POP_L ( EBP ) + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_rescale_normals) +GLNAME(_mesa_3dnow_transform_rescale_normals): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 8 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EDI ) /* dest->count = in->count */ + MOV_L ( EDI, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + + CMP_L ( CONST(0), EDI ) + JE ( LLBL (G3TR_end) ) + + FEMMS + + MOVQ ( REGIND(ECX), MM3 ) /* m1 | m0 */ + + MOVQ ( REGOFF(16,ECX), MM4 ) /* m5 | m4 */ + MOVD ( ARG_SCALE, MM0 ) /* scale */ + + MOVD ( REGOFF(8,ECX), MM5 ) /* | m2 */ + PUNPCKLDQ ( MM0, MM0 ) /* scale | scale */ + + PUNPCKLDQ ( REGOFF(24, ECX), MM5 ) + PFMUL ( MM0, MM3 ) /* scale*m1 | scale*m0 */ + + MOVQ ( REGOFF(32, ECX), MM6 ) /* m9 | m8*/ + PFMUL ( MM0, MM4 ) /* scale*m5 | scale*m4 */ + + MOVD ( REGOFF(40, ECX), MM7 ) /* | m10 */ + PFMUL ( MM0, MM5 ) /* scale*m6 | scale*m2 */ + + PFMUL ( MM0, MM6 ) /* scale*m9 | scale*m8 */ + + PFMUL ( MM0, MM7 ) /* | scale*m10 */ + +ALIGNTEXT32 +LLBL (G3TR_rescale): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM2 ) /* | x2 */ + + MOVQ ( MM0, MM1 ) /* x1 | x0 */ + PUNPCKLDQ ( MM2, MM2 ) /* x2 | x2 */ + + PFMUL ( MM3, MM0 ) /* x1*m1 | x0*m0 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PFMUL ( MM4, MM1 ) /* x1*m5 | x0*m4 */ + PFACC ( MM1, MM0 ) /* x0*m4+x1*m5 | x0*m0+x1*m1 */ + + MOVQ ( REGIND(EDX), MM1 ) /* x1 | x0 */ + + PFMUL ( MM5, MM2 ) /* x2*m6 | x2*m2 */ + PFADD ( MM2, MM0 ) /* x0*m4...+x2*m6| x0*m0+x1*m1+x2*m2 */ + + MOVD ( REGOFF(8, EDX), MM2 ) /* | x2 */ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + MOVQ ( MM0, REGOFF(-16, EAX) ) /* write r0, r1 */ + PFMUL ( MM6, MM1 ) /* x1*m9 | x0*m8 */ + + PFMUL ( MM7, MM2 ) /* | x2*m10 */ + PFACC ( MM1, MM1 ) /* *not used* | x0*m8+x1*m9 */ + + PFADD ( MM2, MM1 ) /* *not used* | x0*m8+x1*m9+x2*m10 */ + MOVD ( MM1, REGOFF(-8, EAX) ) /* write r2 */ + + DEC_L ( EDI ) /* decrement normal counter */ + JA ( LLBL (G3TR_rescale) ) + + FEMMS + +LLBL (G3TR_end): + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_normals_no_rot) +GLNAME(_mesa_3dnow_transform_normals_no_rot): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 8 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EDI ) /* dest->count = in->count */ + MOV_L ( EDI, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + + CMP_L ( CONST(0), EDI ) + JE ( LLBL (G3TNR_end) ) + + FEMMS + + MOVD ( REGIND(ECX), MM0 ) /* | m0 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m5 | m0 */ + + MOVD ( REGOFF(40, ECX), MM2 ) /* | m10 */ + PUNPCKLDQ ( MM2, MM2 ) /* m10 | m10 */ + +ALIGNTEXT32 +LLBL (G3TNR_transform): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM4 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM5 ) /* | x2 */ + + PFMUL ( MM0, MM4 ) /* x1*m5 | x0*m0 */ + ADD_L ( STRIDE, EDX) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + PFMUL ( MM2, MM5 ) /* | x2*m10 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + DEC_L ( EDI ) /* decrement normal counter */ + MOVQ ( MM4, REGOFF(-16, EAX) ) /* write r0, r1 */ + + MOVD ( MM5, REGOFF(-8, EAX) ) /* write r2 */ + JA ( LLBL (G3TNR_transform) ) + + FEMMS + +LLBL (G3TNR_end): + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_transform_normals) +GLNAME(_mesa_3dnow_transform_normals): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 8 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( ARG_MAT, ECX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EDI ) /* dest->count = in->count */ + MOV_L ( EDI, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), EDX ) /* in->start */ + MOV_L ( REGOFF(MATRIX_INV, ECX), ECX ) /* mat->inv */ + + CMP_L ( CONST(0), EDI ) /* count > 0 ?? */ + JE ( LLBL (G3T_end) ) + + FEMMS + + MOVQ ( REGIND(ECX), MM3 ) /* m1 | m0 */ + MOVQ ( REGOFF(16, ECX), MM4 ) /* m5 | m4 */ + + MOVD ( REGOFF(8, ECX), MM5 ) /* | m2 */ + PUNPCKLDQ ( REGOFF(24, ECX), MM5 ) /* m6 | m2 */ + + MOVQ ( REGOFF(32, ECX), MM6 ) /* m9 | m8 */ + MOVD ( REGOFF(40, ECX), MM7 ) /* | m10 */ + +ALIGNTEXT32 +LLBL (G3T_transform): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(EDX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EDX), MM2 ) /* | x2 */ + + MOVQ ( MM0, MM1 ) /* x1 | x0 */ + PUNPCKLDQ ( MM2, MM2 ) /* x2 | x2 */ + + PFMUL ( MM3, MM0 ) /* x1*m1 | x0*m0 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + PFMUL ( MM4, MM1 ) /* x1*m5 | x0*m4 */ + PFACC ( MM1, MM0 ) /* x0*m4+x1*m5 | x0*m0+x1*m1 */ + + PFMUL ( MM5, MM2 ) /* x2*m6 | x2*m2 */ + PFADD ( MM2, MM0 ) /* x0*m4...+x2*m6| x0*m0+x1*m1+x2*m2 */ + + MOVQ ( REGIND(EDX), MM1 ) /* x1 | x0 */ + MOVQ ( MM0, REGOFF(-16, EAX) ) /* write r0, r1 */ + + PFMUL ( MM6, MM1 ) /* x1*m9 | x0*m8 */ + MOVD ( REGOFF(8, EDX), MM2 ) /* | x2 */ + + PFMUL ( MM7, MM2 ) /* | x2*m10 */ + ADD_L ( STRIDE, EDX ) /* next normal */ + + PREFETCH ( REGIND(EDX) ) + + PFACC ( MM1, MM1 ) /* *not used* | x0*m8+x1*m9 */ + PFADD ( MM2, MM1 ) /* *not used* | x0*m8+x1*m9+x2*m10 */ + + MOVD ( MM1, REGOFF(-8, EAX) ) /* write r2 */ + DEC_L ( EDI ) /* decrement normal counter */ + + JA ( LLBL (G3T_transform) ) + + FEMMS + +LLBL (G3T_end): + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_normalize_normals) +GLNAME(_mesa_3dnow_normalize_normals): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 12 + + PUSH_L ( EDI ) + PUSH_L ( ESI ) + PUSH_L ( EBP ) + + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EBP ) /* dest->count = in->count */ + MOV_L ( EBP, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), ECX ) /* in->start */ + MOV_L ( ARG_LENGTHS, EDX ) + + CMP_L ( CONST(0), EBP ) /* count > 0 ?? */ + JE ( LLBL (G3N_end) ) + + FEMMS + + CMP_L ( CONST(0), EDX ) /* lengths == 0 ? */ + JE ( LLBL (G3N_norm2) ) /* calculate lengths */ + +ALIGNTEXT32 +LLBL (G3N_norm1): /* use precalculated lengths */ + + PREFETCH ( REGIND(EAX) ) + + MOVQ ( REGIND(ECX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, ECX), MM1 ) /* | x2 */ + + MOVD ( REGIND(EDX), MM3 ) /* | length (x) */ + PFMUL ( MM3, MM1 ) /* | x2 (normalized) */ + + PUNPCKLDQ ( MM3, MM3 ) /* length (x) | length (x) */ + ADD_L ( STRIDE, ECX ) /* next normal */ + + PREFETCH ( REGIND(ECX) ) + + PFMUL ( MM3, MM0 ) /* x1 (normalized) | x0 (normalized) */ + MOVQ ( MM0, REGIND(EAX) ) /* write new x0, x1 */ + + MOVD ( MM1, REGOFF(8, EAX) ) /* write new x2 */ + ADD_L ( CONST(16), EAX ) /* next r */ + + ADD_L ( CONST(4), EDX ) /* next length */ + DEC_L ( EBP ) /* decrement normal counter */ + + JA ( LLBL (G3N_norm1) ) + + JMP ( LLBL (G3N_end1) ) + +ALIGNTEXT32 +LLBL (G3N_norm2): /* need to calculate lengths */ + + PREFETCHW ( REGIND(EAX) ) + + PREFETCH ( REGIND(ECX) ) + + MOVQ ( REGIND(ECX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, ECX), MM1 ) /* | x2 */ + + MOVQ ( MM0, MM3 ) /* x1 | x0 */ + ADD_L ( STRIDE, ECX ) /* next normal */ + + PFMUL ( MM0, MM3 ) /* x1*x1 | x0*x0 */ + MOVQ ( MM1, MM4 ) /* | x2 */ + + ADD_L ( CONST(16), EAX ) /* next r */ + PFMUL ( MM1, MM4 ) /* | x2*x2 */ + + PFADD ( MM4, MM3 ) /* | x0*x0+x2*x2 */ + PFACC ( MM3, MM3 ) /* x0*x0+...+x2*x2 | x0*x0+x1*x1+x2*x2*/ + + PFRSQRT ( MM3, MM5 ) /* 1/sqrt (x0*x0+x1*x1+x2*x2) */ + MOVQ ( MM5, MM4 ) + + PUNPCKLDQ ( MM3, MM3 ) + PFMUL ( MM5, MM5 ) + + PFRSQIT1 ( MM3, MM5 ) + DEC_L ( EBP ) /* decrement normal counter */ + + PFRCPIT2 ( MM4, MM5 ) + + PFMUL ( MM5, MM0 ) /* x1 (normalized) | x0 (normalized) */ + MOVQ ( MM0, REGOFF(-16, EAX) ) /* write new x0, x1 */ + + PFMUL ( MM5, MM1 ) /* | x2 (normalized) */ + MOVD ( MM1, REGOFF(-8, EAX) ) /* write new x2 */ + + JA ( LLBL (G3N_norm2) ) + +LLBL (G3N_end1): + FEMMS + +LLBL (G3N_end): + POP_L ( EBP ) + POP_L ( ESI ) + POP_L ( EDI ) + RET + + + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_3dnow_rescale_normals) +GLNAME(_mesa_3dnow_rescale_normals): + +#undef FRAME_OFFSET +#define FRAME_OFFSET 8 + PUSH_L ( EDI ) + PUSH_L ( ESI ) + + MOV_L ( ARG_IN, ESI ) + MOV_L ( ARG_DEST, EAX ) + MOV_L ( REGOFF(V4F_COUNT, ESI), EDX ) /* dest->count = in->count */ + MOV_L ( EDX, REGOFF(V4F_COUNT, EAX) ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) /* dest->start */ + MOV_L ( REGOFF(V4F_START, ESI), ECX ) /* in->start */ + + CMP_L ( CONST(0), EDX ) + JE ( LLBL (G3R_end) ) + + FEMMS + + MOVD ( ARG_SCALE, MM0 ) /* scale */ + PUNPCKLDQ ( MM0, MM0 ) + +ALIGNTEXT32 +LLBL (G3R_rescale): + + PREFETCHW ( REGIND(EAX) ) + + MOVQ ( REGIND(ECX), MM1 ) /* x1 | x0 */ + MOVD ( REGOFF(8, ECX), MM2 ) /* | x2 */ + + PFMUL ( MM0, MM1 ) /* x1*scale | x0*scale */ + ADD_L ( STRIDE, ECX ) /* next normal */ + + PREFETCH ( REGIND(ECX) ) + + PFMUL ( MM0, MM2 ) /* | x2*scale */ + ADD_L ( CONST(16), EAX ) /* next r */ + + MOVQ ( MM1, REGOFF(-16, EAX) ) /* write r0, r1 */ + MOVD ( MM2, REGOFF(-8, EAX) ) /* write r2 */ + + DEC_L ( EDX ) /* decrement normal counter */ + JA ( LLBL (G3R_rescale) ) + + FEMMS + +LLBL (G3R_end): + POP_L ( ESI ) + POP_L ( EDI ) + RET + +#endif diff --git a/src/mesa/x86/3dnow_xform1.S b/src/mesa/x86/3dnow_xform1.S new file mode 100644 index 0000000..f551f56 --- /dev/null +++ b/src/mesa/x86/3dnow_xform1.S @@ -0,0 +1,426 @@ +/* $Id: 3dnow_xform1.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifdef USE_3DNOW_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FRAME_OFFSET 4 + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_general ) +GLNAME( _mesa_3dnow_transform_points1_general ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPGR_3 ) ) + + MOVQ ( REGIND(ECX), MM0 ) /* m01 | m00 */ + MOVQ ( REGOFF(8, ECX), MM1 ) /* m03 | m02 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + MOVQ ( REGOFF(56, ECX), MM3 ) /* m33 | m32 */ + +ALIGNTEXT16 +LLBL( G3TPGR_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + PUNPCKLDQ ( MM4, MM4 ) /* x0 | x0 */ + + MOVQ ( MM4, MM5 ) /* x0 | x0 */ + PFMUL ( MM0, MM4 ) /* x0*m01 | x0*m00 */ + + PFMUL ( MM1, MM5 ) /* x0*m03 | x0*m02 */ + PFADD ( MM2, MM4 ) /* x0*m01+m31 | x0*m00+m30 */ + + PFADD ( MM3, MM5 ) /* x0*m03+m33 | x0*m02+m32 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + + MOVQ ( MM5, REGOFF(8, EDX) ) /* write r3, r2 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TPGR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPGR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_identity ) +GLNAME( _mesa_3dnow_transform_points1_identity ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(1), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_1), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPIR_4) ) + +ALIGNTEXT16 +LLBL( G3TPIR_3 ): + + MOVD ( REGIND(EAX), MM0 ) /* | x0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + MOVD ( MM0, REGIND(EDX) ) /* | r0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPIR_3 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPIR_4 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_3d_no_rot ) +GLNAME( _mesa_3dnow_transform_points1_3d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3NRR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TP3NRR_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + PFMUL ( MM0, MM4 ) /* | x0*m00 */ + + PFADD ( MM2, MM4 ) /* m31 | x0*m00+m30 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + + MOVD ( MM3, REGOFF(8, EDX) ) /* write r2 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP3NRR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3NRR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_perspective ) +GLNAME( _mesa_3dnow_transform_points1_perspective ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPPR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TPPR_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* 0 | x0 */ + PFMUL ( MM0, MM4 ) /* 0 | x0*m00 */ + + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + MOVQ ( MM3, REGOFF(8, EDX) ) /* write r2 (=m32), r3 (=0) */ + + ADD_L ( EDI, EAX ) /* next vertex */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPPR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPPR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_2d ) +GLNAME( _mesa_3dnow_transform_points1_2d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(2), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2R_3 ) ) + + MOVQ ( REGIND(ECX), MM0 ) /* m01 | m00 */ + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2R_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + PUNPCKLDQ ( MM4, MM4 ) /* x0 | x0 */ + + PFMUL ( MM0, MM4 ) /* x0*m01 | x0*m00 */ + PFADD ( MM2, MM4 ) /* x0*m01+m31 | x0*m00+m30 */ + + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP2R_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2R_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_2d_no_rot ) +GLNAME( _mesa_3dnow_transform_points1_2d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(2), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2NRR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2NRR_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + PFMUL ( MM0, MM4 ) /* | x0*m00 */ + PFADD ( MM2, MM4 ) /* m31 | x0*m00+m30 */ + + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP2NRR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2NRR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points1_3d ) +GLNAME( _mesa_3dnow_transform_points1_3d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(4, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3R_3 ) ) + + MOVQ ( REGIND(ECX), MM0 ) /* m01 | m00 */ + MOVD ( REGOFF(8, ECX), MM1 ) /* | m02 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TP3R_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + PUNPCKLDQ ( MM4, MM4 ) /* x0 | x0 */ + + MOVQ ( MM4, MM5 ) /* | x0 */ + PFMUL ( MM0, MM4 ) /* x0*m01 | x0*m00 */ + + PFMUL ( MM1, MM5 ) /* | x0*m02 */ + PFADD ( MM2, MM4 ) /* x0*m01+m31 | x0*m00+m30 */ + + PFADD ( MM3, MM5 ) /* | x0*m02+m32 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + + MOVD ( MM5, REGOFF(8, EDX) ) /* write r2 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP3R_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3R_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + +#endif diff --git a/src/mesa/x86/3dnow_xform2.S b/src/mesa/x86/3dnow_xform2.S new file mode 100644 index 0000000..cbe3a19 --- /dev/null +++ b/src/mesa/x86/3dnow_xform2.S @@ -0,0 +1,466 @@ +/* $Id: 3dnow_xform2.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifdef USE_3DNOW_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FRAME_OFFSET 4 + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_general ) +GLNAME( _mesa_3dnow_transform_points2_general ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPGR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(16, ECX), MM0 ) /* m10 | m00 */ + + MOVD ( REGOFF(4, ECX), MM1 ) /* | m01 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM1 ) /* m11 | m01 */ + + MOVD ( REGOFF(8, ECX), MM2 ) /* | m02 */ + PUNPCKLDQ ( REGOFF(24, ECX), MM2 ) /* m12 | m02 */ + + MOVD ( REGOFF(12, ECX), MM3 ) /* | m03 */ + PUNPCKLDQ ( REGOFF(28, ECX), MM3 ) /* m13 | m03 */ + + MOVQ ( REGOFF(48, ECX), MM4 ) /* m31 | m30 */ + MOVQ ( REGOFF(56, ECX), MM5 ) /* m33 | m32 */ + +ALIGNTEXT16 +LLBL( G3TPGR_2 ): + + MOVQ ( REGIND(EAX), MM6 ) /* x1 | x0 */ + MOVQ ( MM6, MM7 ) /* x1 | x0 */ + + PFMUL ( MM0, MM6 ) /* x1*m10 | x0*m00 */ + PFMUL ( MM1, MM7 ) /* x1*m11 | x0*m01 */ + + PFACC ( MM7, MM6 ) /* x0*m01+x1*m11 | x0*x00+x1*m10 */ + PFADD ( MM4, MM6 ) /* x0*...*m11+m31 | x0*...*m10+m30 */ + + MOVQ ( MM6, REGIND(EDX) ) /* write r1, r0 */ + MOVQ ( REGIND(EAX), MM6 ) /* x1 | x0 */ + + MOVQ ( MM6, MM7 ) /* x1 | x0 */ + PFMUL ( MM2, MM6 ) /* x1*m12 | x0*m02 */ + + PFMUL ( MM3, MM7 ) /* x1*m13 | x0*m03 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + PFACC ( MM7, MM6 ) /* x0*m03+x1*m13 | x0*x02+x1*m12 */ + PFADD ( MM5, MM6 ) /* x0*...*m13+m33 | x0*...*m12+m32 */ + + MOVQ ( MM6, REGOFF(8, EDX) ) /* write r3, r2 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPGR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPGR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_perspective ) +GLNAME( _mesa_3dnow_transform_points2_perspective ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPPR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TPPR_2 ): + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + MOVQ ( MM3, REGOFF(8, EDX) ) /* write r2 (=m32), r3 (=0) */ + + ADD_L ( EDI, EAX ) /* next vertex */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPPR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPPR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_3d ) +GLNAME( _mesa_3dnow_transform_points2_3d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3 ), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3R_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(16, ECX), MM0 ) /* m10 | m00 */ + + MOVD ( REGOFF(4, ECX), MM1 ) /* | m01 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM1 ) /* m11 | m01 */ + + MOVD ( REGOFF(8, ECX), MM2 ) /* | m02 */ + PUNPCKLDQ ( REGOFF(24, ECX), MM2 ) /* m12 | m02 */ + + MOVQ ( REGOFF(48, ECX), MM4 ) /* m31 | m30 */ + MOVD ( REGOFF(56, ECX), MM5 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TP3R_2 ): + + MOVQ ( REGIND(EAX), MM6 ) /* x1 | x0 */ + MOVQ ( MM6, MM7 ) /* x1 | x0 */ + + PFMUL ( MM0, MM6 ) /* x1*m10 | x0*m00 */ + PFMUL ( MM1, MM7 ) /* x1*m11 | x0*m01 */ + + PFACC ( MM7, MM6 ) /* x0*m01+x1*m11 | x0*x00+x1*m10 */ + PFADD ( MM4, MM6 ) /* x0*...*m11+m31 | x0*...*m10+m30 */ + + MOVQ ( MM6, REGIND(EDX) ) /* write r1, r0 */ + MOVQ ( REGIND(EAX), MM6 ) /* x1 | x0 */ + + MOVQ ( MM6, MM7 ) /* x1 | x0 */ + PFMUL ( MM2, MM6 ) /* x1*m12 | x0*m02 */ + + PFACC ( MM7, MM6 ) /* ***trash*** | x0*x02+x1*m12 */ + PFADD ( MM5, MM6 ) /* ***trash*** | x0*...*m12+m32 */ + + MOVD ( MM6, REGOFF(8, EDX) ) /* write r2 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP3R_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3R_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_3d_no_rot ) +GLNAME( _mesa_3dnow_transform_points2_3d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3 ), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3NRR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TP3NRR_2 ): + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + + PFADD ( MM2, MM4 ) /* x1*m11+m31 | x0*m00+m30 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + + MOVD ( MM3, REGOFF(8, EDX) ) /* write r2 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP3NRR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3NRR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_2d ) +GLNAME( _mesa_3dnow_transform_points2_2d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(2), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2R_3 ) ) + + MOVQ ( REGIND(ECX), MM0 ) /* m01 | m00 */ + MOVQ ( REGOFF(16, ECX), MM1 ) /* m11 | m10 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2R_2 ): + + MOVD ( REGIND(EAX), MM4 ) /* | x0 */ + MOVD ( REGOFF(4, EAX), MM5 ) /* | x1 */ + + PUNPCKLDQ ( MM4, MM4 ) /* x0 | x0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + PFMUL ( MM0, MM4 ) /* x0*m01 | x0*m00 */ + PUNPCKLDQ ( MM5, MM5 ) /* x1 | x1 */ + + PFMUL ( MM1, MM5 ) /* x1*m11 | x1*m10 */ + PFADD ( MM2, MM4 ) /* x...x1*m11+31 | x0*..*m10+m30 */ + + PFADD ( MM5, MM4 ) /* x0*m01+x1*m11 | x0*m00+x1*m10 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP2R_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2R_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_2d_no_rot ) +GLNAME( _mesa_3dnow_transform_points2_2d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(2), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2NRR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2NRR_2 ): + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + PFADD ( MM2, MM4 ) /* m31 | x0*m00+m30 */ + + MOVQ ( MM4, REGIND(EDX) ) /* write r1, r0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP2NRR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2NRR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points2_identity ) +GLNAME( _mesa_3dnow_transform_points2_identity ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(2), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPIR_3 ) ) + +ALIGNTEXT16 +LLBL( G3TPIR_3 ): + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + ADD_L ( EDI, EAX ) /* next vertex */ + + MOVQ ( MM0, REGIND(EDX) ) /* r1 | r0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPIR_3 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPIR_4 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET +#endif diff --git a/src/mesa/x86/3dnow_xform3.S b/src/mesa/x86/3dnow_xform3.S new file mode 100644 index 0000000..2822794 --- /dev/null +++ b/src/mesa/x86/3dnow_xform3.S @@ -0,0 +1,550 @@ +/* $Id: 3dnow_xform3.S,v 1.3 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifdef USE_3DNOW_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FRAME_OFFSET 4 + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_general ) +GLNAME( _mesa_3dnow_transform_points3_general ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPGR_2 ) ) + + PREFETCHW ( REGIND(EDX) ) + +ALIGNTEXT16 +LLBL( G3TPGR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM2 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + MOVQ ( MM0, MM1 ) /* x1 | x0 */ + PUNPCKLDQ ( MM2, MM2 ) /* x2 | x2 */ + + PUNPCKLDQ ( MM0, MM0 ) /* x0 | x0 */ + MOVQ ( MM2, MM5 ) /* x2 | x2 */ + + PUNPCKHDQ ( MM1, MM1 ) /* x1 | x1 */ + PFMUL ( REGOFF(32, ECX), MM2 ) /* x2*m9 | x2*m8 */ + + MOVQ ( MM0, MM3 ) /* x0 | x0 */ + PFMUL ( REGOFF(40, ECX), MM5 ) /* x2*m11 | x2*m10 */ + + MOVQ ( MM1, MM4 ) /* x1 | x1 */ + PFMUL ( REGIND(ECX), MM0 ) /* x0*m1 | x0*m0 */ + + PFADD ( REGOFF(48, ECX), MM2 ) /* x2*m9+m13 | x2*m8+m12 */ + PFMUL ( REGOFF(16, ECX), MM1 ) /* x1*m5 | x1*m4 */ + + PFADD ( REGOFF(56, ECX), MM5 ) /* x2*m11+m15 | x2*m10+m14 */ + PFADD ( MM0, MM1 ) /* x0*m1+x1*m5 | x0*m0+x1*m4 */ + + PFMUL ( REGOFF(8, ECX), MM3 ) /* x0*m3 | x0*m2 */ + PFADD ( MM1, MM2 ) /* r1 | r0 */ + + PFMUL ( REGOFF(24, ECX), MM4 ) /* x1*m7 | x1*m6 */ + ADD_L ( CONST(16), EDX ) /* next output vertex */ + + PFADD ( MM3, MM4 ) /* x0*m3+x1*m7 | x0*m2+x1*m6 */ + MOVQ ( MM2, REGOFF(-16, EDX) ) /* write r0, r1 */ + + PFADD ( MM4, MM5 ) /* r3 | r2 */ + MOVQ ( MM5, REGOFF(-8, EDX) ) /* write r2, r3 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPGR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPGR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_perspective ) +GLNAME( _mesa_3dnow_transform_points3_perspective ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPPR_2 ) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCHW ( REGIND(EDX) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVQ ( REGOFF(32, ECX), MM1 ) /* m21 | m20 */ + MOVD ( REGOFF(40, ECX), MM2 ) /* | m22 */ + + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + +ALIGNTEXT16 +LLBL( G3TPPR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVD ( REGOFF(8, EAX), MM5 ) /* | x2 */ + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + PXOR ( MM7, MM7 ) /* 0 | 0 */ + MOVQ ( MM5, MM6 ) /* | x2 */ + + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + PFSUB ( MM5, MM7 ) /* | -x2 */ + + PFMUL ( MM2, MM6 ) /* | x2*m22 */ + PUNPCKLDQ ( MM5, MM5 ) /* x2 | x2 */ + + ADD_L ( CONST(16), EDX ) /* next r */ + PFMUL ( MM1, MM5 ) /* x2*m21 | x2*m20 */ + + PFADD ( MM3, MM6 ) /* | x2*m22+m32 */ + PFADD ( MM4, MM5 ) /* x1*m11+x2*m21 | x0*m00+x2*m20 */ + + MOVQ ( MM5, REGOFF(-16, EDX) ) /* write r0, r1 */ + MOVD ( MM6, REGOFF(-8, EDX) ) /* write r2 */ + + MOVD ( MM7, REGOFF(-4, EDX) ) /* write r3 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPPR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPPR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_3d ) +GLNAME( _mesa_3dnow_transform_points3_3d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3R_2 ) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCH ( REGIND(EDX) ) + + MOVD ( REGOFF(8, ECX), MM7 ) /* | m2 */ + PUNPCKLDQ ( REGOFF(24, ECX), MM7 ) /* m6 | m2 */ + + +ALIGNTEXT16 +LLBL( G3TP3R_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM1 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + MOVQ ( MM0, MM2 ) /* x1 | x0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PUNPCKLDQ ( MM2, MM2 ) /* x0 | x0 */ + MOVQ ( MM0, MM3 ) /* x1 | x0 */ + + PFMUL ( REGIND(ECX), MM2 ) /* x0*m1 | x0*m0 */ + PUNPCKHDQ ( MM3, MM3 ) /* x1 | x1 */ + + MOVQ ( MM1, MM4 ) /* | x2 */ + PFMUL ( REGOFF(16, ECX), MM3 ) /* x1*m5 | x1*m4 */ + + PUNPCKLDQ ( MM4, MM4 ) /* x2 | x2 */ + PFADD ( MM2, MM3 ) /* x0*m1+x1*m5 | x0*m0+x1*m4 */ + + PFMUL ( REGOFF(32, ECX), MM4 ) /* x2*m9 | x2*m8 */ + PFADD ( REGOFF(48, ECX), MM3 ) /* x0*m1+...+m11 | x0*m0+x1*m4+m12 */ + + PFMUL ( MM7, MM0 ) /* x1*m6 | x0*m2 */ + PFADD ( MM4, MM3 ) /* r1 | r0 */ + + PFMUL ( REGOFF(40, ECX), MM1 ) /* | x2*m10 */ + PUNPCKLDQ ( REGOFF(56, ECX), MM1 ) /* m14 | x2*m10 */ + + PFACC ( MM0, MM1 ) + + MOVQ ( MM3, REGOFF(-16, EDX) ) /* write r0, r1 */ + PFACC ( MM1, MM1 ) /* | r2 */ + + MOVD ( MM1, REGOFF(-8, EDX) ) /* write r2 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP3R_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3R_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_3d_no_rot ) +GLNAME( _mesa_3dnow_transform_points3_3d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3NRR_2 ) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCHW ( REGIND(EDX) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVD ( REGOFF(40, ECX), MM2 ) /* | m22 */ + PUNPCKLDQ ( MM2, MM2 ) /* m22 | m22 */ + + MOVQ ( REGOFF(48, ECX), MM1 ) /* m31 | m30 */ + MOVD ( REGOFF(56, ECX), MM3 ) /* | m32 */ + + PUNPCKLDQ ( MM3, MM3 ) /* m32 | m32 */ + + +ALIGNTEXT16 +LLBL( G3TP3NRR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM5 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCHW ( REGIND(EAX) ) + + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + + PFADD ( MM1, MM4 ) /* x1*m11+m31 | x0*m00+m30 */ + PFMUL ( MM2, MM5 ) /* | x2*m22 */ + + PFADD ( MM3, MM5 ) /* | x2*m22+m32 */ + MOVQ ( MM4, REGIND(EDX) ) /* write r0, r1 */ + + ADD_L ( CONST(16), EDX ) /* next r */ + DEC_L ( ESI ) /* decrement vertex counter */ + + MOVD ( MM5, REGOFF(-8, EDX) ) /* write r2 */ + JNZ ( LLBL( G3TP3NRR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3NRR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_2d ) +GLNAME( _mesa_3dnow_transform_points3_2d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2R_3) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCHW ( REGIND(EDX) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(16, ECX), MM0 ) /* m10 | m00 */ + + MOVD ( REGOFF(4, ECX), MM1 ) /* | m01 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM1 ) /* m11 | m01 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2R_2 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM3 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM5 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + MOVQ ( MM3, MM4 ) /* x1 | x0 */ + PFMUL ( MM0, MM3 ) /* x1*m10 | x0*m00 */ + + ADD_L ( CONST(16), EDX ) /* next r */ + PFMUL ( MM1, MM4 ) /* x1*m11 | x0*m01 */ + + PFACC ( MM4, MM3 ) /* x0*m00+x1*m10 | x0*m01+x1*m11 */ + MOVD ( MM5, REGOFF(-8, EDX) ) /* write r2 (=x2) */ + + PFADD ( MM2, MM3 ) /* x0*...*m10+m30 | x0*...*m11+m31 */ + MOVQ ( MM3, REGOFF(-16, EDX) ) /* write r0, r1 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP2R_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2R_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_2d_no_rot ) +GLNAME( _mesa_3dnow_transform_points3_2d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2NRR_2 ) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCHW ( REGIND(EDX) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVQ ( REGOFF(48, ECX), MM1 ) /* m31 | m30 */ + + +ALIGNTEXT16 +LLBL( G3TP2NRR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM5 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PFADD ( MM1, MM4 ) /* x1*m11+m31 | x0*m00+m30 */ + + MOVQ ( MM4, REGOFF(-16, EDX) ) /* write r0, r1 */ + MOVD ( MM5, REGOFF(-8, EDX) ) /* write r2 (=x2) */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP2NRR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2NRR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points3_identity ) +GLNAME( _mesa_3dnow_transform_points3_identity ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(3), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPIR_2 ) ) + + PREFETCHW ( REGIND(EDX) ) + +ALIGNTEXT16 +LLBL( G3TPIR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVD ( REGOFF(8, EAX), MM1 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + ADD_L ( CONST(16), EDX ) /* next r */ + + DEC_L ( ESI ) /* decrement vertex counter */ + MOVQ ( MM0, REGOFF(-16, EDX) ) /* r1 | r0 */ + + MOVD ( MM1, REGOFF(-8, EDX) ) /* | r2 */ + JNZ ( LLBL( G3TPIR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPIR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET +#endif diff --git a/src/mesa/x86/3dnow_xform4.S b/src/mesa/x86/3dnow_xform4.S new file mode 100644 index 0000000..6692529 --- /dev/null +++ b/src/mesa/x86/3dnow_xform4.S @@ -0,0 +1,559 @@ +/* $Id: 3dnow_xform4.S,v 1.3 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifdef USE_3DNOW_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FRAME_OFFSET 4 + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_general ) +GLNAME( _mesa_3dnow_transform_points4_general ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPGR_2 ) ) + + PREFETCHW ( REGIND(EDX) ) + +ALIGNTEXT16 +LLBL( G3TPGR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM4 ) /* x3 | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + MOVQ ( MM0, MM2 ) /* x1 | x0 */ + MOVQ ( MM4, MM6 ) /* x3 | x2 */ + + PUNPCKLDQ ( MM0, MM0 ) /* x0 | x0 */ + PUNPCKHDQ ( MM2, MM2 ) /* x1 | x1 */ + + MOVQ ( MM0, MM1 ) /* x0 | x0 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PFMUL ( REGIND(ECX), MM0 ) /* x0*m1 | x0*m0 */ + MOVQ ( MM2, MM3 ) /* x1 | x1 */ + + PFMUL ( REGOFF(8, ECX), MM1 ) /* x0*m3 | x0*m2 */ + PUNPCKLDQ ( MM4, MM4 ) /* x2 | x2 */ + + PFMUL ( REGOFF(16, ECX), MM2 ) /* x1*m5 | x1*m4 */ + MOVQ ( MM4, MM5 ) /* x2 | x2 */ + + PFMUL ( REGOFF(24, ECX), MM3 ) /* x1*m7 | x1*m6 */ + PUNPCKHDQ ( MM6, MM6 ) /* x3 | x3 */ + + PFMUL ( REGOFF(32, ECX), MM4 ) /* x2*m9 | x2*m8 */ + MOVQ ( MM6, MM7 ) /* x3 | x3 */ + + PFMUL ( REGOFF(40, ECX), MM5 ) /* x2*m11 | x2*m10 */ + PFADD ( MM0, MM2 ) + + PFMUL ( REGOFF(48, ECX), MM6 ) /* x3*m13 | x3*m12 */ + PFADD ( MM1, MM3 ) + + PFMUL ( REGOFF(56, ECX), MM7 ) /* x3*m15 | x3*m14 */ + PFADD ( MM4, MM6 ) + + PFADD ( MM5, MM7 ) + PFADD ( MM2, MM6 ) + + PFADD ( MM3, MM7 ) + MOVQ ( MM6, REGOFF(-16, EDX) ) + + MOVQ ( MM7, REGOFF(-8, EDX) ) + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPGR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPGR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_perspective ) +GLNAME( _mesa_3dnow_transform_points4_perspective ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPPR_2 ) ) + + PREFETCH ( REGIND(EAX) ) + PREFETCHW ( REGIND(EDX) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVD ( REGOFF(40, ECX), MM1 ) /* | m22 */ + PUNPCKLDQ ( REGOFF(56, ECX), MM1 ) /* m32 | m22 */ + + MOVQ ( REGOFF(32, ECX), MM2 ) /* m21 | m20 */ + PXOR ( MM7, MM7 ) /* 0 | 0 */ + +ALIGNTEXT16 +LLBL( G3TPPR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM5 ) /* x3 | x2 */ + MOVD ( REGOFF(8, EAX), MM3 ) /* | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGOFF(32, EAX) ) /* hopefully stride is zero */ + + MOVQ ( MM5, MM6 ) /* x3 | x2 */ + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + + PUNPCKLDQ ( MM5, MM5 ) /* x2 | x2 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PFMUL ( MM2, MM5 ) /* x2*m21 | x2*m20 */ + PFSUBR ( MM7, MM3 ) /* | -x2 */ + + PFMUL ( MM1, MM6 ) /* x3*m32 | x2*m22 */ + PFADD ( MM4, MM5 ) /* x1*m11+x2*m21 | x0*m00+x2*m20 */ + + PFACC ( MM3, MM6 ) /* -x2 | x2*m22+x3*m32 */ + MOVQ ( MM5, REGOFF(-16, EDX) ) /* write r0, r1 */ + + MOVQ ( MM6, REGOFF(-8, EDX) ) /* write r2, r3 */ + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TPPR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPPR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_3d ) +GLNAME( _mesa_3dnow_transform_points4_3d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3R_2 ) ) + + MOVD ( REGOFF(8, ECX), MM6 ) /* | m2 */ + PUNPCKLDQ ( REGOFF(24, ECX), MM6 ) /* m6 | m2 */ + + MOVD ( REGOFF(40, ECX), MM7 ) /* | m10 */ + PUNPCKLDQ ( REGOFF(56, ECX), MM7 ) /* m14 | m10 */ + +ALIGNTEXT16 +LLBL( G3TP3R_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + PREFETCH ( REGOFF(32, EAX) ) /* hopefully array is tightly packed */ + + MOVQ ( REGIND(EAX), MM2 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM3 ) /* x3 | x2 */ + + MOVQ ( MM2, MM0 ) /* x1 | x0 */ + MOVQ ( MM3, MM4 ) /* x3 | x2 */ + + MOVQ ( MM0, MM1 ) /* x1 | x0 */ + MOVQ ( MM4, MM5 ) /* x3 | x2 */ + + PUNPCKLDQ ( MM0, MM0 ) /* x0 | x0 */ + PUNPCKHDQ ( MM1, MM1 ) /* x1 | x1 */ + + PFMUL ( REGIND(ECX), MM0 ) /* x0*m1 | x0*m0 */ + PUNPCKLDQ ( MM3, MM3 ) /* x2 | x2 */ + + PFMUL ( REGOFF(16, ECX), MM1 ) /* x1*m5 | x1*m4 */ + PUNPCKHDQ ( MM4, MM4 ) /* x3 | x3 */ + + PFMUL ( MM6, MM2 ) /* x1*m6 | x0*m2 */ + PFADD ( MM0, MM1 ) /* x0*m1+x1*m5 | x0*m0+x1*m4 */ + + PFMUL ( REGOFF(32, ECX), MM3 ) /* x2*m9 | x2*m8 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PFMUL ( REGOFF(48, ECX), MM4 ) /* x3*m13 | x3*m12 */ + PFADD ( MM1, MM3 ) /* x0*m1+..+x2*m9 | x0*m0+...+x2*m8 */ + + PFMUL ( MM7, MM5 ) /* x3*m14 | x2*m10 */ + PFADD ( MM3, MM4 ) /* r1 | r0 */ + + PFACC ( MM2, MM5 ) /* x0*m2+x1*m6 | x2*m10+x3*m14 */ + MOVD ( REGOFF(12, EAX), MM0 ) /* | x3 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PFACC ( MM0, MM5 ) /* r3 | r2 */ + + MOVQ ( MM4, REGOFF(-16, EDX) ) /* write r0, r1 */ + MOVQ ( MM5, REGOFF(-8, EDX) ) /* write r2, r3 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP3R_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3R_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_3d_no_rot ) +GLNAME( _mesa_3dnow_transform_points4_3d_no_rot ): + + PUSH_L ( ESI ) + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP3NRR_2 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVD ( REGOFF(40, ECX), MM2 ) /* | m22 */ + PUNPCKLDQ ( REGOFF(56, ECX), MM2 ) /* m32 | m22 */ + + MOVQ ( REGOFF(48, ECX), MM1 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP3NRR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM5 ) /* x3 | x2 */ + MOVD ( REGOFF(12, EAX), MM7 ) /* | x3 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGOFF(32, EAX) ) /* hopefully stride is zero */ + + MOVQ ( MM5, MM6 ) /* x3 | x2 */ + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + + PUNPCKHDQ ( MM6, MM6 ) /* x3 | x3 */ + PFMUL ( MM2, MM5 ) /* x3*m32 | x2*m22 */ + + PFMUL ( MM1, MM6 ) /* x3*m31 | x3*m30 */ + PFACC ( MM7, MM5 ) /* x3 | x2*m22+x3*m32 */ + + PFADD ( MM6, MM4 ) /* x1*m11+x3*m31 | x0*m00+x3*m30 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + MOVQ ( MM4, REGOFF(-16, EDX) ) /* write r0, r1 */ + MOVQ ( MM5, REGOFF(-8, EDX) ) /* write r2, r3 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP3NRR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP3NRR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_2d ) +GLNAME( _mesa_3dnow_transform_points4_2d ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2R_2 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(16, ECX), MM0 ) /* m10 | m00 */ + + MOVD ( REGOFF(4, ECX), MM1 ) /* | m01 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM1 ) /* m11 | m01 */ + + MOVQ ( REGOFF(48, ECX), MM2 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2R_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM3 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM5 ) /* x3 | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + MOVQ ( MM3, MM4 ) /* x1 | x0 */ + MOVQ ( MM5, MM6 ) /* x3 | x2 */ + + PFMUL ( MM1, MM4 ) /* x1*m11 | x0*m01 */ + PUNPCKHDQ ( MM6, MM6 ) /* x3 | x3 */ + + PFMUL ( MM0, MM3 ) /* x1*m10 | x0*m00 */ + ADD_L ( CONST(16), EDX ) /* next r */ + + PFACC ( MM4, MM3 ) /* x0*m01+x1*m11 | x0*m00+x1*m10 */ + PFMUL ( MM2, MM6 ) /* x3*m31 | x3*m30 */ + + PFADD ( MM6, MM3 ) /* r1 | r0 */ + MOVQ ( MM5, REGOFF(-8, EDX) ) /* write r2, r3 */ + + MOVQ ( MM3, REGOFF(-16, EDX) ) /* write r0, r1 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TP2R_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2R_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_2d_no_rot ) +GLNAME( _mesa_3dnow_transform_points4_2d_no_rot ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TP2NRR_3 ) ) + + MOVD ( REGIND(ECX), MM0 ) /* | m00 */ + PUNPCKLDQ ( REGOFF(20, ECX), MM0 ) /* m11 | m00 */ + + MOVQ ( REGOFF(48, ECX), MM1 ) /* m31 | m30 */ + +ALIGNTEXT16 +LLBL( G3TP2NRR_2 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM4 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM5 ) /* x3 | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + PFMUL ( MM0, MM4 ) /* x1*m11 | x0*m00 */ + MOVQ ( MM5, MM6 ) /* x3 | x2 */ + + ADD_L ( CONST(16), EDX ) /* next r */ + PUNPCKHDQ ( MM6, MM6 ) /* x3 | x3 */ + + PFMUL ( MM1, MM6 ) /* x3*m31 | x3*m30 */ + PFADD ( MM4, MM6 ) /* x1*m11+x3*m31 | x0*m00+x3*m30 */ + + MOVQ ( MM6, REGOFF(-16, EDX) ) /* write r0, r1 */ + MOVQ ( MM5, REGOFF(-8, EDX) ) /* write r2, r3 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + + JNZ ( LLBL( G3TP2NRR_2 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TP2NRR_3 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_3dnow_transform_points4_identity ) +GLNAME( _mesa_3dnow_transform_points4_identity ): + + PUSH_L ( ESI ) + + MOV_L ( ARG_DEST, ECX ) + MOV_L ( ARG_MATRIX, ESI ) + MOV_L ( ARG_SOURCE, EAX ) + MOV_L ( CONST(4), REGOFF(V4F_SIZE, ECX) ) + OR_B ( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, ECX) ) + MOV_L ( REGOFF(V4F_COUNT, EAX), EDX ) + MOV_L ( EDX, REGOFF(V4F_COUNT, ECX) ) + + PUSH_L ( EDI ) + + MOV_L ( REGOFF(V4F_START, ECX), EDX ) + MOV_L ( ESI, ECX ) + MOV_L ( REGOFF(V4F_COUNT, EAX), ESI ) + MOV_L ( REGOFF(V4F_STRIDE, EAX), EDI ) + MOV_L ( REGOFF(V4F_START, EAX), EAX ) + + TEST_L ( ESI, ESI ) + JZ ( LLBL( G3TPIR_2 ) ) + +ALIGNTEXT16 +LLBL( G3TPIR_1 ): + + PREFETCHW ( REGOFF(32, EDX) ) /* prefetch 2 vertices ahead */ + + MOVQ ( REGIND(EAX), MM0 ) /* x1 | x0 */ + MOVQ ( REGOFF(8, EAX), MM1 ) /* x3 | x2 */ + + ADD_L ( EDI, EAX ) /* next vertex */ + PREFETCH ( REGIND(EAX) ) + + ADD_L ( CONST(16), EDX ) /* next r */ + MOVQ ( MM0, REGOFF(-16, EDX) ) /* r1 | r0 */ + + MOVQ ( MM1, REGOFF(-8, EDX) ) /* r3 | r2 */ + + DEC_L ( ESI ) /* decrement vertex counter */ + JNZ ( LLBL( G3TPIR_1 ) ) /* cnt > 0 ? -> process next vertex */ + +LLBL( G3TPIR_2 ): + + FEMMS + POP_L ( EDI ) + POP_L ( ESI ) + RET +#endif diff --git a/src/mesa/x86/Makefile b/src/mesa/x86/Makefile new file mode 100644 index 0000000..125f893 --- /dev/null +++ b/src/mesa/x86/Makefile @@ -0,0 +1,43 @@ +# src/mesa/x86/Makefile + +TOP = ../../.. + +include $(TOP)/configs/current + + + +INCLUDE_DIRS = \ + -I$(TOP)/include/GL \ + -I$(TOP)/include \ + -I.. \ + -I../main \ + -I../math \ + -I../glapi \ + -I../tnl + + +default: matypes.h + + +# need some special rules here, unfortunately +matypes.h: ../main/mtypes.h ../tnl/t_context.h gen_matypes.c + $(CC) $(INCLUDE_DIRS) $(CFLAGS) gen_matypes.c -o gen_matypes + ./gen_matypes > matypes.h + +common_x86_asm.o: matypes.h +3dnow_normal.o: matypes.h +3dnow_xform1.o: matypes.h +3dnow_xform2.o: matypes.h +3dnow_xform3.o: matypes.h +3dnow_xform4.o: matypes.h +mmx_blend.o: matypes.h +sse_normal.o: matypes.h +sse_xform1.o: matypes.h +sse_xform2.o: matypes.h +sse_xform3.o: matypes.h +sse_xform4.o: matypes.h +x86_cliptest.o: matypes.h +x86_xform2.o: matypes.h +x86_xform3.o: matypes.h +x86_xform4.o: matypes.h + diff --git a/src/mesa/x86/assyntax.h b/src/mesa/x86/assyntax.h new file mode 100644 index 0000000..16aaf90 --- /dev/null +++ b/src/mesa/x86/assyntax.h @@ -0,0 +1,1733 @@ + +#ifndef __ASSYNTAX_H__ +#define __ASSYNTAX_H__ + +/* + * Copyright 1992 Vrije Universiteit, The Netherlands + * + * Permission to use, copy, modify, and distribute this software and its + * documentation for any purpose and without fee is hereby granted, provided + * that the above copyright notice appear in all copies and that both that + * copyright notice and this permission notice appear in supporting + * documentation, and that the name of the Vrije Universiteit not be used in + * advertising or publicity pertaining to distribution of the software without + * specific, written prior permission. The Vrije Universiteit makes no + * representations about the suitability of this software for any purpose. + * It is provided "as is" without express or implied warranty. + * + * The Vrije Universiteit DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS + * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, + * IN NO EVENT SHALL The Vrije Universiteit BE LIABLE FOR ANY SPECIAL, + * INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM + * LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE + * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR + * PERFORMANCE OF THIS SOFTWARE. + */ + +/* + * assyntax.h + * + * Select the syntax appropriate to the 386 assembler being used + * To add support for more assemblers add more columns to the CHOICE + * macro. Note that register names must also have uppercase names + * to avoid macro recursion. e.g., #define ah %ah recurses! + * + * NB 1. Some of the macros for certain assemblers imply that the code is to + * run in protected mode!! Caveat emptor. + * + * NB 2. 486 specific instructions are not included. This is to discourage + * their accidental use in code that is intended to run on 386 and 486 + * systems. + * + * Supported assemblers: + * + * (a) AT&T SysVr4 as(1): define ATT_ASSEMBLER + * (b) GNU Assembler gas: define GNU_ASSEMBLER (default) + * (c) Amsterdam Compiler kit: define ACK_ASSEMBLER + * (d) The Netwide Assembler: define NASM_ASSEMBLER + * (e) Microsoft Assembler: define MASM_ASSEMBLER (UNTESTED!) + * + * The following naming conventions have been used to identify the various + * data types: + * _SR = segment register version + * Integer: + * _Q = quadword = 64 bits + * _L = long = 32 bits + * _W = short = 16 bits + * _B = byte = 8 bits + * Floating-point: + * _X = m80real = 80 bits + * _D = double = 64 bits + * _S = single = 32 bits + * + * Author: Gregory J. Sharp, Sept 1992 + * Vrije Universiteit, Amsterdam, The Netherlands + * + * [support for Intel syntax added by Josh Vanderhoof, 1999] + */ + +#if !(defined(NASM_ASSEMBLER) || defined(MASM_ASSEMBLER)) + +/* Default to ATT_ASSEMBLER when SVR4 or SYSV are defined */ +#if (defined(SVR4) || defined(SYSV)) && !defined(GNU_ASSEMBLER) +#define ATT_ASSEMBLER +#endif + +#if !defined(ATT_ASSEMBLER) && !defined(GNU_ASSEMBLER) && !defined(ACK_ASSEMBLER) +#define GNU_ASSEMBLER +#endif + +#if (defined(__STDC__) && !defined(UNIXCPP)) || (defined (sun) && defined (i386) && defined (SVR4) && defined (__STDC__) && !defined (__GNUC__)) +#define CONCAT(x, y) x ## y +#define CONCAT3(x, y, z) x ## y ## z +#else +#define CONCAT(x, y) x/**/y +#define CONCAT3(x, y, z) x/**/y/**/z +#endif + +#ifdef ACK_ASSEMBLER + +/* Assume we write code for 32-bit protected mode! */ + +/* Redefine register names for GAS & AT&T assemblers */ +#define AL al +#define AH ah +#define AX ax +#define EAX ax +#define BL bl +#define BH bh +#define BX bx +#define EBX bx +#define CL cl +#define CH ch +#define CX cx +#define ECX cx +#define DL dl +#define DH dh +#define DX dx +#define EDX dx +#define BP bp +#define EBP bp +#define SI si +#define ESI si +#define DI di +#define EDI di +#define SP sp +#define ESP sp +#define CS cs +#define SS ss +#define DS ds +#define ES es +#define FS fs +#define GS gs +/* Control Registers */ +#define CR0 cr0 +#define CR1 cr1 +#define CR2 cr2 +#define CR3 cr3 +/* Debug Registers */ +#define DR0 dr0 +#define DR1 dr1 +#define DR2 dr2 +#define DR3 dr3 +#define DR4 dr4 +#define DR5 dr5 +#define DR6 dr6 +#define DR7 dr7 +/* Floating-point Stack */ +#define ST st + +#define AS_BEGIN .sect .text; .sect .rom; .sect .data; .sect .bss; .sect .text + + +#define _WTOG o16 /* word toggle for _W instructions */ +#define _LTOG /* long toggle for _L instructions */ +#define ADDR_TOGGLE a16 +#define OPSZ_TOGGLE o16 +#define USE16 .use16 +#define USE32 .use32 + +#define CHOICE(a,b,c) c + +#else /* AT&T or GAS */ + +/* Redefine register names for GAS & AT&T assemblers */ +#define AL %al +#define AH %ah +#define AX %ax +#define EAX %eax +#define BL %bl +#define BH %bh +#define BX %bx +#define EBX %ebx +#define CL %cl +#define CH %ch +#define CX %cx +#define ECX %ecx +#define DL %dl +#define DH %dh +#define DX %dx +#define EDX %edx +#define BP %bp +#define EBP %ebp +#define SI %si +#define ESI %esi +#define DI %di +#define EDI %edi +#define SP %sp +#define ESP %esp +#define CS %cs +#define SS %ss +#define DS %ds +#define ES %es +#define FS %fs +#define GS %gs +/* Control Registers */ +#define CR0 %cr0 +#define CR1 %cr1 +#define CR2 %cr2 +#define CR3 %cr3 +/* Debug Registers */ +#define DR0 %db0 +#define DR1 %db1 +#define DR2 %db2 +#define DR3 %db3 +#define DR4 %db4 +#define DR5 %db5 +#define DR6 %db6 +#define DR7 %db7 +/* Floating-point Stack */ +#define _STX0 %st(0) +#define _STX1 %st(1) +#define _STX2 %st(2) +#define _STX3 %st(3) +#define _STX4 %st(4) +#define _STX5 %st(5) +#define _STX6 %st(6) +#define _STX7 %st(7) +#define ST(x) CONCAT(_STX,x) +#ifdef GNU_ASSEMBLER +#define ST0 %st(0) +#else +#define ST0 %st +#endif +/* MMX Registers */ +#define MM0 %mm0 +#define MM1 %mm1 +#define MM2 %mm2 +#define MM3 %mm3 +#define MM4 %mm4 +#define MM5 %mm5 +#define MM6 %mm6 +#define MM7 %mm7 +/* SSE Registers */ +#define XMM0 %xmm0 +#define XMM1 %xmm1 +#define XMM2 %xmm2 +#define XMM3 %xmm3 +#define XMM4 %xmm4 +#define XMM5 %xmm5 +#define XMM6 %xmm6 +#define XMM7 %xmm7 + +#define AS_BEGIN +#define USE16 +#define USE32 + +#ifdef GNU_ASSEMBLER + +#define ADDR_TOGGLE aword +#define OPSZ_TOGGLE word + +#define CHOICE(a,b,c) b + +#else +/* + * AT&T ASSEMBLER SYNTAX + * ********************* + */ +#define CHOICE(a,b,c) a + +#define ADDR_TOGGLE addr16 +#define OPSZ_TOGGLE data16 + +#endif /* GNU_ASSEMBLER */ +#endif /* ACK_ASSEMBLER */ + + +#if defined(__QNX__) || defined(Lynx) || (defined(SYSV) || defined(SVR4)) && !defined(ACK_ASSEMBLER) || defined(__ELF__) || defined(__GNU__) || defined(__GNUC__) && !defined(__DJGPP__) && !defined(__MINGW32__) +#define GLNAME(a) a +#else +#define GLNAME(a) CONCAT(_,a) +#endif + + + /****************************************/ + /* */ + /* Select the various choices */ + /* */ + /****************************************/ + + +/* Redefine assembler directives */ +/*********************************/ +#define GLOBL CHOICE(.globl, .globl, .extern) +#define GLOBAL GLOBL +#define EXTERN GLOBL +#ifndef __AOUT__ +#define ALIGNTEXT32 CHOICE(.align 32, .balign 32, .align 32) +#define ALIGNTEXT16 CHOICE(.align 16, .balign 16, .align 16) +#define ALIGNTEXT8 CHOICE(.align 8, .balign 8, .align 8) +#define ALIGNTEXT4 CHOICE(.align 4, .balign 4, .align 4) +#define ALIGNTEXT2 CHOICE(.align 2, .balign 2, .align 2) +/* ALIGNTEXT4ifNOP is the same as ALIGNTEXT4, but only if the space is + * guaranteed to be filled with NOPs. Otherwise it does nothing. + */ +#define ALIGNTEXT32ifNOP CHOICE(.align 32, .balign ARG2(32,0x90), /*can't do it*/) +#define ALIGNTEXT16ifNOP CHOICE(.align 16, .balign ARG2(16,0x90), /*can't do it*/) +#define ALIGNTEXT8ifNOP CHOICE(.align 8, .balign ARG2(8,0x90), /*can't do it*/) +#define ALIGNTEXT4ifNOP CHOICE(.align 4, .balign ARG2(4,0x90), /*can't do it*/) +#define ALIGNDATA32 CHOICE(.align 32, .balign ARG2(32,0x0), .align 32) +#define ALIGNDATA16 CHOICE(.align 16, .balign ARG2(16,0x0), .align 16) +#define ALIGNDATA8 CHOICE(.align 8, .balign ARG2(8,0x0), .align 8) +#define ALIGNDATA4 CHOICE(.align 4, .balign ARG2(4,0x0), .align 4) +#define ALIGNDATA2 CHOICE(.align 2, .balign ARG2(2,0x0), .align 2) +#else +/* 'as -aout' on FreeBSD doesn't have .balign */ +#define ALIGNTEXT32 CHOICE(.align 32, .align ARG2(5,0x90), .align 32) +#define ALIGNTEXT16 CHOICE(.align 16, .align ARG2(4,0x90), .align 16) +#define ALIGNTEXT8 CHOICE(.align 8, .align ARG2(3,0x90), .align 8) +#define ALIGNTEXT4 CHOICE(.align 4, .align ARG2(2,0x90), .align 4) +#define ALIGNTEXT2 CHOICE(.align 2, .align ARG2(1,0x90), .align 2) +/* ALIGNTEXT4ifNOP is the same as ALIGNTEXT4, but only if the space is + * guaranteed to be filled with NOPs. Otherwise it does nothing. + */ +#define ALIGNTEXT32ifNOP CHOICE(.align 32, .align ARG2(5,0x90), /*can't do it*/) +#define ALIGNTEXT16ifNOP CHOICE(.align 16, .align ARG2(4,0x90), /*can't do it*/) +#define ALIGNTEXT8ifNOP CHOICE(.align 8, .align ARG2(3,0x90), /*can't do it*/) +#define ALIGNTEXT4ifNOP CHOICE(.align 4, .align ARG2(2,0x90), /*can't do it*/) +#define ALIGNDATA32 CHOICE(.align 32, .align ARG2(5,0x0), .align 32) +#define ALIGNDATA16 CHOICE(.align 16, .align ARG2(4,0x0), .align 16) +#define ALIGNDATA8 CHOICE(.align 8, .align ARG2(3,0x0), .align 8) +#define ALIGNDATA4 CHOICE(.align 4, .align ARG2(2,0x0), .align 4) +#define ALIGNDATA2 CHOICE(.align 2, .align ARG2(1,0x0), .align 2) +#endif /* __AOUT__ */ +#define FILE(s) CHOICE(.file s, .file s, .file s) +#define STRING(s) CHOICE(.string s, .asciz s, .asciz s) +#define D_LONG CHOICE(.long, .long, .data4) +#define D_WORD CHOICE(.value, .short, .data2) +#define D_BYTE CHOICE(.byte, .byte, .data1) +#define SPACE CHOICE(.comm, .space, .space) +#define COMM CHOICE(.comm, .comm, .comm) +#define SEG_DATA CHOICE(.data, .data, .sect .data) +#define SEG_TEXT CHOICE(.text, .text, .sect .text) +#define SEG_BSS CHOICE(.bss, .bss, .sect .bss) + +#ifdef GNU_ASSEMBLER +#define D_SPACE(n) . = . + n +#else +#define D_SPACE(n) .space n +#endif + +/* Addressing Modes */ +/* Immediate Mode */ +#define ADDR(a) CHOICE(CONCAT($,a), $a, a) +#define CONST(a) CHOICE(CONCAT($,a), $a, a) + +/* Indirect Mode */ +#define CONTENT(a) CHOICE(a, a, (a)) /* take contents of variable */ +#define REGIND(a) CHOICE((a), (a), (a)) /* Register a indirect */ +/* Register b indirect plus displacement a */ +#define REGOFF(a, b) CHOICE(a(b), a(b), a(b)) +/* Reg indirect Base + Index + Displacement - this is mainly for 16-bit mode + * which has no scaling + */ +#define REGBID(b,i,d) CHOICE(d(b,i), d(b,i), d(b)(i)) +/* Reg indirect Base + (Index * Scale) */ +#define REGBIS(b,i,s) CHOICE((b,i,s), (b,i,s), (b)(i*s)) +/* Reg indirect Base + (Index * Scale) + Displacement */ +#define REGBISD(b,i,s,d) CHOICE(d(b,i,s), d(b,i,s), d(b)(i*s)) +/* Displaced Scaled Index: */ +#define REGDIS(d,i,s) CHOICE(d(,i,s), d(,i,s), d(i * s)) +/* Indexed Base: */ +#define REGBI(b,i) CHOICE((b,i), (b,i), (b)(i)) +/* Displaced Base: */ +#define REGDB(d,b) CHOICE(d(b), d(b), d(b)) +/* Variable indirect: */ +#define VARINDIRECT(var) CHOICE(*var, *var, (var)) +/* Use register contents as jump/call target: */ +#define CODEPTR(reg) CHOICE(*reg, *reg, reg) + +/* For expressions requiring bracketing + * eg. (CRT0_PM | CRT_EM) + */ + +#define EXPR(a) CHOICE([a], (a), [a]) +#define ENOT(a) CHOICE(0!a, ~a, ~a) +#define EMUL(a,b) CHOICE(a\*b, a*b, a*b) +#define EDIV(a,b) CHOICE(a\/b, a/b, a/b) + +/* + * We have to beat the problem of commas within arguments to choice. + * eg. choice (add a,b, add b,a) will get argument mismatch. Luckily ANSI + * and other known cpp definitions evaluate arguments before substitution + * so the following works. + */ +#define ARG2(a, b) a,b +#define ARG3(a,b,c) a,b,c + +/* Redefine assembler commands */ +#define AAA CHOICE(aaa, aaa, aaa) +#define AAD CHOICE(aad, aad, aad) +#define AAM CHOICE(aam, aam, aam) +#define AAS CHOICE(aas, aas, aas) +#define ADC_L(a, b) CHOICE(adcl ARG2(a,b), adcl ARG2(a,b), _LTOG adc ARG2(b,a)) +#define ADC_W(a, b) CHOICE(adcw ARG2(a,b), adcw ARG2(a,b), _WTOG adc ARG2(b,a)) +#define ADC_B(a, b) CHOICE(adcb ARG2(a,b), adcb ARG2(a,b), adcb ARG2(b,a)) +#define ADD_L(a, b) CHOICE(addl ARG2(a,b), addl ARG2(a,b), _LTOG add ARG2(b,a)) +#define ADD_W(a, b) CHOICE(addw ARG2(a,b), addw ARG2(a,b), _WTOG add ARG2(b,a)) +#define ADD_B(a, b) CHOICE(addb ARG2(a,b), addb ARG2(a,b), addb ARG2(b,a)) +#define AND_L(a, b) CHOICE(andl ARG2(a,b), andl ARG2(a,b), _LTOG and ARG2(b,a)) +#define AND_W(a, b) CHOICE(andw ARG2(a,b), andw ARG2(a,b), _WTOG and ARG2(b,a)) +#define AND_B(a, b) CHOICE(andb ARG2(a,b), andb ARG2(a,b), andb ARG2(b,a)) +#define ARPL(a,b) CHOICE(arpl ARG2(a,b), arpl ARG2(a,b), arpl ARG2(b,a)) +#define BOUND_L(a, b) CHOICE(boundl ARG2(a,b), boundl ARG2(b,a), _LTOG bound ARG2(b,a)) +#define BOUND_W(a, b) CHOICE(boundw ARG2(a,b), boundw ARG2(b,a), _WTOG bound ARG2(b,a)) +#define BSF_L(a, b) CHOICE(bsfl ARG2(a,b), bsfl ARG2(a,b), _LTOG bsf ARG2(b,a)) +#define BSF_W(a, b) CHOICE(bsfw ARG2(a,b), bsfw ARG2(a,b), _WTOG bsf ARG2(b,a)) +#define BSR_L(a, b) CHOICE(bsrl ARG2(a,b), bsrl ARG2(a,b), _LTOG bsr ARG2(b,a)) +#define BSR_W(a, b) CHOICE(bsrw ARG2(a,b), bsrw ARG2(a,b), _WTOG bsr ARG2(b,a)) +#define BT_L(a, b) CHOICE(btl ARG2(a,b), btl ARG2(a,b), _LTOG bt ARG2(b,a)) +#define BT_W(a, b) CHOICE(btw ARG2(a,b), btw ARG2(a,b), _WTOG bt ARG2(b,a)) +#define BTC_L(a, b) CHOICE(btcl ARG2(a,b), btcl ARG2(a,b), _LTOG btc ARG2(b,a)) +#define BTC_W(a, b) CHOICE(btcw ARG2(a,b), btcw ARG2(a,b), _WTOG btc ARG2(b,a)) +#define BTR_L(a, b) CHOICE(btrl ARG2(a,b), btrl ARG2(a,b), _LTOG btr ARG2(b,a)) +#define BTR_W(a, b) CHOICE(btrw ARG2(a,b), btrw ARG2(a,b), _WTOG btr ARG2(b,a)) +#define BTS_L(a, b) CHOICE(btsl ARG2(a,b), btsl ARG2(a,b), _LTOG bts ARG2(b,a)) +#define BTS_W(a, b) CHOICE(btsw ARG2(a,b), btsw ARG2(a,b), _WTOG bts ARG2(b,a)) +#define CALL(a) CHOICE(call a, call a, call a) +#define CALLF(s,a) CHOICE(lcall ARG2(s,a), lcall ARG2(s,a), callf s:a) +#define CBW CHOICE(cbtw, cbw, cbw) +#define CWDE CHOICE(cwtd, cwde, cwde) +#define CLC CHOICE(clc, clc, clc) +#define CLD CHOICE(cld, cld, cld) +#define CLI CHOICE(cli, cli, cli) +#define CLTS CHOICE(clts, clts, clts) +#define CMC CHOICE(cmc, cmc, cmc) +#define CMP_L(a, b) CHOICE(cmpl ARG2(a,b), cmpl ARG2(a,b), _LTOG cmp ARG2(b,a)) +#define CMP_W(a, b) CHOICE(cmpw ARG2(a,b), cmpw ARG2(a,b), _WTOG cmp ARG2(b,a)) +#define CMP_B(a, b) CHOICE(cmpb ARG2(a,b), cmpb ARG2(a,b), cmpb ARG2(b,a)) +#define CMPS_L CHOICE(cmpsl, cmpsl, _LTOG cmps) +#define CMPS_W CHOICE(cmpsw, cmpsw, _WTOG cmps) +#define CMPS_B CHOICE(cmpsb, cmpsb, cmpsb) +#define CWD CHOICE(cwtl, cwd, cwd) +#define CDQ CHOICE(cltd, cdq, cdq) +#define DAA CHOICE(daa, daa, daa) +#define DAS CHOICE(das, das, das) +#define DEC_L(a) CHOICE(decl a, decl a, _LTOG dec a) +#define DEC_W(a) CHOICE(decw a, decw a, _WTOG dec a) +#define DEC_B(a) CHOICE(decb a, decb a, decb a) +#define DIV_L(a) CHOICE(divl a, divl a, div a) +#define DIV_W(a) CHOICE(divw a, divw a, div a) +#define DIV_B(a) CHOICE(divb a, divb a, divb a) +#define ENTER(a,b) CHOICE(enter ARG2(a,b), enter ARG2(a,b), enter ARG2(b,a)) +#define HLT CHOICE(hlt, hlt, hlt) +#define IDIV_L(a) CHOICE(idivl a, idivl a, _LTOG idiv a) +#define IDIV_W(a) CHOICE(idivw a, idivw a, _WTOG idiv a) +#define IDIV_B(a) CHOICE(idivb a, idivb a, idivb a) +/* More forms than this for imul!! */ +#define IMUL_L(a, b) CHOICE(imull ARG2(a,b), imull ARG2(a,b), _LTOG imul ARG2(b,a)) +#define IMUL_W(a, b) CHOICE(imulw ARG2(a,b), imulw ARG2(a,b), _WTOG imul ARG2(b,a)) +#define IMUL_B(a) CHOICE(imulb a, imulb a, imulb a) +#define IN_L CHOICE(inl (DX), inl ARG2(DX,EAX), _LTOG in DX) +#define IN_W CHOICE(inw (DX), inw ARG2(DX,AX), _WTOG in DX) +#define IN_B CHOICE(inb (DX), inb ARG2(DX,AL), inb DX) +/* Please AS code writer: use the following ONLY, if you refer to ports<256 + * directly, but not in IN1_W(DX), for instance, even if IN1_ looks nicer + */ +#if defined (sun) +#define IN1_L(a) CHOICE(inl (a), inl ARG2(a,EAX), _LTOG in a) +#define IN1_W(a) CHOICE(inw (a), inw ARG2(a,AX), _WTOG in a) +#define IN1_B(a) CHOICE(inb (a), inb ARG2(a,AL), inb a) +#else +#define IN1_L(a) CHOICE(inl a, inl ARG2(a,EAX), _LTOG in a) +#define IN1_W(a) CHOICE(inw a, inw ARG2(a,AX), _WTOG in a) +#define IN1_B(a) CHOICE(inb a, inb ARG2(a,AL), inb a) +#endif +#define INC_L(a) CHOICE(incl a, incl a, _LTOG inc a) +#define INC_W(a) CHOICE(incw a, incw a, _WTOG inc a) +#define INC_B(a) CHOICE(incb a, incb a, incb a) +#define INS_L CHOICE(insl, insl, _LTOG ins) +#define INS_W CHOICE(insw, insw, _WTOG ins) +#define INS_B CHOICE(insb, insb, insb) +#define INT(a) CHOICE(int a, int a, int a) +#define INT3 CHOICE(int CONST(3), int3, int CONST(3)) +#define INTO CHOICE(into, into, into) +#define IRET CHOICE(iret, iret, iret) +#define IRETD CHOICE(iret, iret, iretd) +#define JA(a) CHOICE(ja a, ja a, ja a) +#define JAE(a) CHOICE(jae a, jae a, jae a) +#define JB(a) CHOICE(jb a, jb a, jb a) +#define JBE(a) CHOICE(jbe a, jbe a, jbe a) +#define JC(a) CHOICE(jc a, jc a, jc a) +#define JE(a) CHOICE(je a, je a, je a) +#define JG(a) CHOICE(jg a, jg a, jg a) +#define JGE(a) CHOICE(jge a, jge a, jge a) +#define JL(a) CHOICE(jl a, jl a, jl a) +#define JLE(a) CHOICE(jle a, jle a, jle a) +#define JNA(a) CHOICE(jna a, jna a, jna a) +#define JNAE(a) CHOICE(jnae a, jnae a, jnae a) +#define JNB(a) CHOICE(jnb a, jnb a, jnb a) +#define JNBE(a) CHOICE(jnbe a, jnbe a, jnbe a) +#define JNC(a) CHOICE(jnc a, jnc a, jnc a) +#define JNE(a) CHOICE(jne a, jne a, jne a) +#define JNG(a) CHOICE(jng a, jng a, jng a) +#define JNGE(a) CHOICE(jnge a, jnge a, jnge a) +#define JNL(a) CHOICE(jnl a, jnl a, jnl a) +#define JNLE(a) CHOICE(jnle a, jnle a, jnle a) +#define JNO(a) CHOICE(jno a, jno a, jno a) +#define JNP(a) CHOICE(jnp a, jnp a, jnp a) +#define JNS(a) CHOICE(jns a, jns a, jns a) +#define JNZ(a) CHOICE(jnz a, jnz a, jnz a) +#define JO(a) CHOICE(jo a, jo a, jo a) +#define JP(a) CHOICE(jp a, jp a, jp a) +#define JPE(a) CHOICE(jpe a, jpe a, jpe a) +#define JPO(a) CHOICE(jpo a, jpo a, jpo a) +#define JS(a) CHOICE(js a, js a, js a) +#define JZ(a) CHOICE(jz a, jz a, jz a) +#define JMP(a) CHOICE(jmp a, jmp a, jmp a) +#define JMPF(s,a) CHOICE(ljmp ARG2(s,a), ljmp ARG2(s,a), jmpf s:a) +#define LAHF CHOICE(lahf, lahf, lahf) +#if !defined(_REAL_MODE) && !defined(_V86_MODE) +#define LAR(a, b) CHOICE(lar ARG2(a, b), lar ARG2(a, b), lar ARG2(b, a)) +#endif +#define LEA_L(a, b) CHOICE(leal ARG2(a,b), leal ARG2(a,b), _LTOG lea ARG2(b,a)) +#define LEA_W(a, b) CHOICE(leaw ARG2(a,b), leaw ARG2(a,b), _WTOG lea ARG2(b,a)) +#define LEAVE CHOICE(leave, leave, leave) +#define LGDT(a) CHOICE(lgdt a, lgdt a, lgdt a) +#define LIDT(a) CHOICE(lidt a, lidt a, lidt a) +#define LDS(a, b) CHOICE(ldsl ARG2(a,b), lds ARG2(a,b), lds ARG2(b,a)) +#define LES(a, b) CHOICE(lesl ARG2(a,b), les ARG2(a,b), les ARG2(b,a)) +#define LFS(a, b) CHOICE(lfsl ARG2(a,b), lfs ARG2(a,b), lfs ARG2(b,a)) +#define LGS(a, b) CHOICE(lgsl ARG2(a,b), lgs ARG2(a,b), lgs ARG2(b,a)) +#define LSS(a, b) CHOICE(lssl ARG2(a,b), lss ARG2(a,b), lss ARG2(b,a)) +#define LLDT(a) CHOICE(lldt a, lldt a, lldt a) +#define LMSW(a) CHOICE(lmsw a, lmsw a, lmsw a) +#define LOCK CHOICE(lock, lock, lock) +#define LODS_L CHOICE(lodsl, lodsl, _LTOG lods) +#define LODS_W CHOICE(lodsw, lodsw, _WTOG lods) +#define LODS_B CHOICE(lodsb, lodsb, lodsb) +#define LOOP(a) CHOICE(loop a, loop a, loop a) +#define LOOPE(a) CHOICE(loope a, loope a, loope a) +#define LOOPZ(a) CHOICE(loopz a, loopz a, loopz a) +#define LOOPNE(a) CHOICE(loopne a, loopne a, loopne a) +#define LOOPNZ(a) CHOICE(loopnz a, loopnz a, loopnz a) +#if !defined(_REAL_MODE) && !defined(_V86_MODE) +#define LSL(a, b) CHOICE(lsl ARG2(a,b), lsl ARG2(a,b), lsl ARG2(b,a)) +#endif +#define LTR(a) CHOICE(ltr a, ltr a, ltr a) +#define MOV_SR(a, b) CHOICE(movw ARG2(a,b), mov ARG2(a,b), mov ARG2(b,a)) +#define MOV_L(a, b) CHOICE(movl ARG2(a,b), movl ARG2(a,b), _LTOG mov ARG2(b,a)) +#define MOV_W(a, b) CHOICE(movw ARG2(a,b), movw ARG2(a,b), _WTOG mov ARG2(b,a)) +#define MOV_B(a, b) CHOICE(movb ARG2(a,b), movb ARG2(a,b), movb ARG2(b,a)) +#define MOVS_L CHOICE(movsl, movsl, _LTOG movs) +#define MOVS_W CHOICE(movsw, movsw, _WTOG movs) +#define MOVS_B CHOICE(movsb, movsb, movsb) +#define MOVSX_BL(a, b) CHOICE(movsbl ARG2(a,b), movsbl ARG2(a,b), movsx ARG2(b,a)) +#define MOVSX_BW(a, b) CHOICE(movsbw ARG2(a,b), movsbw ARG2(a,b), movsx ARG2(b,a)) +#define MOVSX_WL(a, b) CHOICE(movswl ARG2(a,b), movswl ARG2(a,b), movsx ARG2(b,a)) +#define MOVZX_BL(a, b) CHOICE(movzbl ARG2(a,b), movzbl ARG2(a,b), movzx ARG2(b,a)) +#define MOVZX_BW(a, b) CHOICE(movzbw ARG2(a,b), movzbw ARG2(a,b), movzx ARG2(b,a)) +#define MOVZX_WL(a, b) CHOICE(movzwl ARG2(a,b), movzwl ARG2(a,b), movzx ARG2(b,a)) +#define MUL_L(a) CHOICE(mull a, mull a, _LTOG mul a) +#define MUL_W(a) CHOICE(mulw a, mulw a, _WTOG mul a) +#define MUL_B(a) CHOICE(mulb a, mulb a, mulb a) +#define NEG_L(a) CHOICE(negl a, negl a, _LTOG neg a) +#define NEG_W(a) CHOICE(negw a, negw a, _WTOG neg a) +#define NEG_B(a) CHOICE(negb a, negb a, negb a) +#define NOP CHOICE(nop, nop, nop) +#define NOT_L(a) CHOICE(notl a, notl a, _LTOG not a) +#define NOT_W(a) CHOICE(notw a, notw a, _WTOG not a) +#define NOT_B(a) CHOICE(notb a, notb a, notb a) +#define OR_L(a,b) CHOICE(orl ARG2(a,b), orl ARG2(a,b), _LTOG or ARG2(b,a)) +#define OR_W(a,b) CHOICE(orw ARG2(a,b), orw ARG2(a,b), _WTOG or ARG2(b,a)) +#define OR_B(a,b) CHOICE(orb ARG2(a,b), orb ARG2(a,b), orb ARG2(b,a)) +#define OUT_L CHOICE(outl (DX), outl ARG2(EAX,DX), _LTOG out DX) +#define OUT_W CHOICE(outw (DX), outw ARG2(AX,DX), _WTOG out DX) +#define OUT_B CHOICE(outb (DX), outb ARG2(AL,DX), outb DX) +/* Please AS code writer: use the following ONLY, if you refer to ports<256 + * directly, but not in OUT1_W(DX), for instance, even if OUT1_ looks nicer + */ +#define OUT1_L(a) CHOICE(outl (a), outl ARG2(EAX,a), _LTOG out a) +#define OUT1_W(a) CHOICE(outw (a), outw ARG2(AX,a), _WTOG out a) +#define OUT1_B(a) CHOICE(outb (a), outb ARG2(AL,a), outb a) +#define OUTS_L CHOICE(outsl, outsl, _LTOG outs) +#define OUTS_W CHOICE(outsw, outsw, _WTOG outs) +#define OUTS_B CHOICE(outsb, outsb, outsb) +#define POP_SR(a) CHOICE(pop a, pop a, pop a) +#define POP_L(a) CHOICE(popl a, popl a, _LTOG pop a) +#define POP_W(a) CHOICE(popw a, popw a, _WTOG pop a) +#define POPA_L CHOICE(popal, popal, _LTOG popa) +#define POPA_W CHOICE(popaw, popaw, _WTOG popa) +#define POPF_L CHOICE(popfl, popfl, _LTOG popf) +#define POPF_W CHOICE(popfw, popfw, _WTOG popf) +#define PUSH_SR(a) CHOICE(push a, push a, push a) +#define PUSH_L(a) CHOICE(pushl a, pushl a, _LTOG push a) +#define PUSH_W(a) CHOICE(pushw a, pushw a, _WTOG push a) +#define PUSH_B(a) CHOICE(push a, pushb a, push a) +#define PUSHA_L CHOICE(pushal, pushal, _LTOG pusha) +#define PUSHA_W CHOICE(pushaw, pushaw, _WTOG pusha) +#define PUSHF_L CHOICE(pushfl, pushfl, _LTOG pushf) +#define PUSHF_W CHOICE(pushfw, pushfw, _WTOG pushf) +#define RCL_L(a, b) CHOICE(rcll ARG2(a,b), rcll ARG2(a,b), _LTOG rcl ARG2(b,a)) +#define RCL_W(a, b) CHOICE(rclw ARG2(a,b), rclw ARG2(a,b), _WTOG rcl ARG2(b,a)) +#define RCL_B(a, b) CHOICE(rclb ARG2(a,b), rclb ARG2(a,b), rclb ARG2(b,a)) +#define RCR_L(a, b) CHOICE(rcrl ARG2(a,b), rcrl ARG2(a,b), _LTOG rcr ARG2(b,a)) +#define RCR_W(a, b) CHOICE(rcrw ARG2(a,b), rcrw ARG2(a,b), _WTOG rcr ARG2(b,a)) +#define RCR_B(a, b) CHOICE(rcrb ARG2(a,b), rcrb ARG2(a,b), rcrb ARG2(b,a)) +#define ROL_L(a, b) CHOICE(roll ARG2(a,b), roll ARG2(a,b), _LTOG rol ARG2(b,a)) +#define ROL_W(a, b) CHOICE(rolw ARG2(a,b), rolw ARG2(a,b), _WTOG rol ARG2(b,a)) +#define ROL_B(a, b) CHOICE(rolb ARG2(a,b), rolb ARG2(a,b), rolb ARG2(b,a)) +#define ROR_L(a, b) CHOICE(rorl ARG2(a,b), rorl ARG2(a,b), _LTOG ror ARG2(b,a)) +#define ROR_W(a, b) CHOICE(rorw ARG2(a,b), rorw ARG2(a,b), _WTOG ror ARG2(b,a)) +#define ROR_B(a, b) CHOICE(rorb ARG2(a,b), rorb ARG2(a,b), rorb ARG2(b,a)) +#define REP CHOICE(rep ;, rep ;, repe) +#define REPE CHOICE(repz ;, repe ;, repe) +#define REPNE CHOICE(repnz ;, repne ;, repne) +#define REPNZ REPNE +#define REPZ REPE +#define RET CHOICE(ret, ret, ret) +#define SAHF CHOICE(sahf, sahf, sahf) +#define SAL_L(a, b) CHOICE(sall ARG2(a,b), sall ARG2(a,b), _LTOG sal ARG2(b,a)) +#define SAL_W(a, b) CHOICE(salw ARG2(a,b), salw ARG2(a,b), _WTOG sal ARG2(b,a)) +#define SAL_B(a, b) CHOICE(salb ARG2(a,b), salb ARG2(a,b), salb ARG2(b,a)) +#define SAR_L(a, b) CHOICE(sarl ARG2(a,b), sarl ARG2(a,b), _LTOG sar ARG2(b,a)) +#define SAR_W(a, b) CHOICE(sarw ARG2(a,b), sarw ARG2(a,b), _WTOG sar ARG2(b,a)) +#define SAR_B(a, b) CHOICE(sarb ARG2(a,b), sarb ARG2(a,b), sarb ARG2(b,a)) +#define SBB_L(a, b) CHOICE(sbbl ARG2(a,b), sbbl ARG2(a,b), _LTOG sbb ARG2(b,a)) +#define SBB_W(a, b) CHOICE(sbbw ARG2(a,b), sbbw ARG2(a,b), _WTOG sbb ARG2(b,a)) +#define SBB_B(a, b) CHOICE(sbbb ARG2(a,b), sbbb ARG2(a,b), sbbb ARG2(b,a)) +#define SCAS_L CHOICE(scasl, scasl, _LTOG scas) +#define SCAS_W CHOICE(scasw, scasw, _WTOG scas) +#define SCAS_B CHOICE(scasb, scasb, scasb) +#define SETA(a) CHOICE(seta a, seta a, seta a) +#define SETAE(a) CHOICE(setae a, setae a, setae a) +#define SETB(a) CHOICE(setb a, setb a, setb a) +#define SETBE(a) CHOICE(setbe a, setbe a, setbe a) +#define SETC(a) CHOICE(setc a, setb a, setb a) +#define SETE(a) CHOICE(sete a, sete a, sete a) +#define SETG(a) CHOICE(setg a, setg a, setg a) +#define SETGE(a) CHOICE(setge a, setge a, setge a) +#define SETL(a) CHOICE(setl a, setl a, setl a) +#define SETLE(a) CHOICE(setle a, setle a, setle a) +#define SETNA(a) CHOICE(setna a, setna a, setna a) +#define SETNAE(a) CHOICE(setnae a, setnae a, setnae a) +#define SETNB(a) CHOICE(setnb a, setnb a, setnb a) +#define SETNBE(a) CHOICE(setnbe a, setnbe a, setnbe a) +#define SETNC(a) CHOICE(setnc a, setnb a, setnb a) +#define SETNE(a) CHOICE(setne a, setne a, setne a) +#define SETNG(a) CHOICE(setng a, setng a, setng a) +#define SETNGE(a) CHOICE(setnge a, setnge a, setnge a) +#define SETNL(a) CHOICE(setnl a, setnl a, setnl a) +#define SETNLE(a) CHOICE(setnle a, setnle a, setnle a) +#define SETNO(a) CHOICE(setno a, setno a, setno a) +#define SETNP(a) CHOICE(setnp a, setnp a, setnp a) +#define SETNS(a) CHOICE(setns a, setns a, setna a) +#define SETNZ(a) CHOICE(setnz a, setnz a, setnz a) +#define SETO(a) CHOICE(seto a, seto a, seto a) +#define SETP(a) CHOICE(setp a, setp a, setp a) +#define SETPE(a) CHOICE(setpe a, setpe a, setpe a) +#define SETPO(a) CHOICE(setpo a, setpo a, setpo a) +#define SETS(a) CHOICE(sets a, sets a, seta a) +#define SETZ(a) CHOICE(setz a, setz a, setz a) +#define SGDT(a) CHOICE(sgdt a, sgdt a, sgdt a) +#define SIDT(a) CHOICE(sidt a, sidt a, sidt a) +#define SHL_L(a, b) CHOICE(shll ARG2(a,b), shll ARG2(a,b), _LTOG shl ARG2(b,a)) +#define SHL_W(a, b) CHOICE(shlw ARG2(a,b), shlw ARG2(a,b), _WTOG shl ARG2(b,a)) +#define SHL_B(a, b) CHOICE(shlb ARG2(a,b), shlb ARG2(a,b), shlb ARG2(b,a)) +#define SHLD_L(a,b,c) CHOICE(shldl ARG3(a,b,c), shldl ARG3(a,b,c), _LTOG shld ARG3(c,b,a)) +#define SHLD2_L(a,b) CHOICE(shldl ARG2(a,b), shldl ARG3(CL,a,b), _LTOG shld ARG3(b,a,CL)) +#define SHLD_W(a,b,c) CHOICE(shldw ARG3(a,b,c), shldw ARG3(a,b,c), _WTOG shld ARG3(c,b,a)) +#define SHLD2_W(a,b) CHOICE(shldw ARG2(a,b), shldw ARG3(CL,a,b), _WTOG shld ARG3(b,a,CL)) +#define SHR_L(a, b) CHOICE(shrl ARG2(a,b), shrl ARG2(a,b), _LTOG shr ARG2(b,a)) +#define SHR_W(a, b) CHOICE(shrw ARG2(a,b), shrw ARG2(a,b), _WTOG shr ARG2(b,a)) +#define SHR_B(a, b) CHOICE(shrb ARG2(a,b), shrb ARG2(a,b), shrb ARG2(b,a)) +#define SHRD_L(a,b,c) CHOICE(shrdl ARG3(a,b,c), shrdl ARG3(a,b,c), _LTOG shrd ARG3(c,b,a)) +#define SHRD2_L(a,b) CHOICE(shrdl ARG2(a,b), shrdl ARG3(CL,a,b), _LTOG shrd ARG3(b,a,CL)) +#define SHRD_W(a,b,c) CHOICE(shrdw ARG3(a,b,c), shrdw ARG3(a,b,c), _WTOG shrd ARG3(c,b,a)) +#define SHRD2_W(a,b) CHOICE(shrdw ARG2(a,b), shrdw ARG3(CL,a,b), _WTOG shrd ARG3(b,a,CL)) +#define SLDT(a) CHOICE(sldt a, sldt a, sldt a) +#define SMSW(a) CHOICE(smsw a, smsw a, smsw a) +#define STC CHOICE(stc, stc, stc) +#define STD CHOICE(std, std, std) +#define STI CHOICE(sti, sti, sti) +#define STOS_L CHOICE(stosl, stosl, _LTOG stos) +#define STOS_W CHOICE(stosw, stosw, _WTOG stos) +#define STOS_B CHOICE(stosb, stosb, stosb) +#define STR(a) CHOICE(str a, str a, str a) +#define SUB_L(a, b) CHOICE(subl ARG2(a,b), subl ARG2(a,b), _LTOG sub ARG2(b,a)) +#define SUB_W(a, b) CHOICE(subw ARG2(a,b), subw ARG2(a,b), _WTOG sub ARG2(b,a)) +#define SUB_B(a, b) CHOICE(subb ARG2(a,b), subb ARG2(a,b), subb ARG2(b,a)) +#define TEST_L(a, b) CHOICE(testl ARG2(a,b), testl ARG2(a,b), _LTOG test ARG2(b,a)) +#define TEST_W(a, b) CHOICE(testw ARG2(a,b), testw ARG2(a,b), _WTOG test ARG2(b,a)) +#define TEST_B(a, b) CHOICE(testb ARG2(a,b), testb ARG2(a,b), testb ARG2(b,a)) +#define VERR(a) CHOICE(verr a, verr a, verr a) +#define VERW(a) CHOICE(verw a, verw a, verw a) +#define WAIT CHOICE(wait, wait, wait) +#define XCHG_L(a, b) CHOICE(xchgl ARG2(a,b), xchgl ARG2(a,b), _LTOG xchg ARG2(b,a)) +#define XCHG_W(a, b) CHOICE(xchgw ARG2(a,b), xchgw ARG2(a,b), _WTOG xchg ARG2(b,a)) +#define XCHG_B(a, b) CHOICE(xchgb ARG2(a,b), xchgb ARG2(a,b), xchgb ARG2(b,a)) +#define XLAT CHOICE(xlat, xlat, xlat) +#define XOR_L(a, b) CHOICE(xorl ARG2(a,b), xorl ARG2(a,b), _LTOG xor ARG2(b,a)) +#define XOR_W(a, b) CHOICE(xorw ARG2(a,b), xorw ARG2(a,b), _WTOG xor ARG2(b,a)) +#define XOR_B(a, b) CHOICE(xorb ARG2(a,b), xorb ARG2(a,b), xorb ARG2(b,a)) + + +/* Floating Point Instructions */ +#define F2XM1 CHOICE(f2xm1, f2xm1, f2xm1) +#define FABS CHOICE(fabs, fabs, fabs) +#define FADD_D(a) CHOICE(faddl a, faddl a, faddd a) +#define FADD_S(a) CHOICE(fadds a, fadds a, fadds a) +#define FADD2(a, b) CHOICE(fadd ARG2(a,b), fadd ARG2(a,b), fadd ARG2(b,a)) +#define FADDP(a, b) CHOICE(faddp ARG2(a,b), faddp ARG2(a,b), faddp ARG2(b,a)) +#define FIADD_L(a) CHOICE(fiaddl a, fiaddl a, fiaddl a) +#define FIADD_W(a) CHOICE(fiadd a, fiadds a, fiadds a) +#define FBLD(a) CHOICE(fbld a, fbld a, fbld a) +#define FBSTP(a) CHOICE(fbstp a, fbstp a, fbstp a) +#define FCHS CHOICE(fchs, fchs, fchs) +#define FCLEX CHOICE(fclex, wait; fnclex, wait; fclex) +#define FNCLEX CHOICE(fnclex, fnclex, fclex) +#define FCOM(a) CHOICE(fcom a, fcom a, fcom a) +#define FCOM_D(a) CHOICE(fcoml a, fcoml a, fcomd a) +#define FCOM_S(a) CHOICE(fcoms a, fcoms a, fcoms a) +#define FCOMP(a) CHOICE(fcomp a, fcomp a, fcomp a) +#define FCOMP_D(a) CHOICE(fcompl a, fcompl a, fcompd a) +#define FCOMP_S(a) CHOICE(fcomps a, fcomps a, fcomps a) +#define FCOMPP CHOICE(fcompp, fcompp, fcompp) +#define FCOS CHOICE(fcos, fcos, fcos) +#define FDECSTP CHOICE(fdecstp, fdecstp, fdecstp) +#define FDIV_D(a) CHOICE(fdivl a, fdivl a, fdivd a) +#define FDIV_S(a) CHOICE(fdivs a, fdivs a, fdivs a) +#define FDIV2(a, b) CHOICE(fdiv ARG2(a,b), fdiv ARG2(a,b), fdiv ARG2(b,a)) +#define FDIVP(a, b) CHOICE(fdivp ARG2(a,b), fdivp ARG2(a,b), fdivp ARG2(b,a)) +#define FIDIV_L(a) CHOICE(fidivl a, fidivl a, fidivl a) +#define FIDIV_W(a) CHOICE(fidiv a, fidivs a, fidivs a) +#define FDIVR_D(a) CHOICE(fdivrl a, fdivrl a, fdivrd a) +#define FDIVR_S(a) CHOICE(fdivrs a, fdivrs a, fdivrs a) +#define FDIVR2(a, b) CHOICE(fdivr ARG2(a,b), fdivr ARG2(a,b), fdivr ARG2(b,a)) +#define FDIVRP(a, b) CHOICE(fdivrp ARG2(a,b), fdivrp ARG2(a,b), fdivrp ARG2(b,a)) +#define FIDIVR_L(a) CHOICE(fidivrl a, fidivrl a, fidivrl a) +#define FIDIVR_W(a) CHOICE(fidivr a, fidivrs a, fidivrs a) +#define FFREE(a) CHOICE(ffree a, ffree a, ffree a) +#define FICOM_L(a) CHOICE(ficoml a, ficoml a, ficoml a) +#define FICOM_W(a) CHOICE(ficom a, ficoms a, ficoms a) +#define FICOMP_L(a) CHOICE(ficompl a, ficompl a, ficompl a) +#define FICOMP_W(a) CHOICE(ficomp a, ficomps a, ficomps a) +#define FILD_Q(a) CHOICE(fildll a, fildq a, fildq a) +#define FILD_L(a) CHOICE(fildl a, fildl a, fildl a) +#define FILD_W(a) CHOICE(fild a, filds a, filds a) +#define FINCSTP CHOICE(fincstp, fincstp, fincstp) +#define FINIT CHOICE(finit, wait; fninit, wait; finit) +#define FNINIT CHOICE(fninit, fninit, finit) +#define FIST_L(a) CHOICE(fistl a, fistl a, fistl a) +#define FIST_W(a) CHOICE(fist a, fists a, fists a) +#define FISTP_Q(a) CHOICE(fistpll a, fistpq a, fistpq a) +#define FISTP_L(a) CHOICE(fistpl a, fistpl a, fistpl a) +#define FISTP_W(a) CHOICE(fistp a, fistps a, fistps a) +#define FLD_X(a) CHOICE(fldt a, fldt a, fldx a) /* 80 bit data type! */ +#define FLD_D(a) CHOICE(fldl a, fldl a, fldd a) +#define FLD_S(a) CHOICE(flds a, flds a, flds a) +#define FLD1 CHOICE(fld1, fld1, fld1) +#define FLDL2T CHOICE(fldl2t, fldl2t, fldl2t) +#define FLDL2E CHOICE(fldl2e, fldl2e, fldl2e) +#define FLDPI CHOICE(fldpi, fldpi, fldpi) +#define FLDLG2 CHOICE(fldlg2, fldlg2, fldlg2) +#define FLDLN2 CHOICE(fldln2, fldln2, fldln2) +#define FLDZ CHOICE(fldz, fldz, fldz) +#define FLDCW(a) CHOICE(fldcw a, fldcw a, fldcw a) +#define FLDENV(a) CHOICE(fldenv a, fldenv a, fldenv a) +#define FMUL_S(a) CHOICE(fmuls a, fmuls a, fmuls a) +#define FMUL_D(a) CHOICE(fmull a, fmull a, fmuld a) +#define FMUL2(a, b) CHOICE(fmul ARG2(a,b), fmul ARG2(a,b), fmul ARG2(b,a)) +#define FMULP(a, b) CHOICE(fmulp ARG2(a,b), fmulp ARG2(a,b), fmulp ARG2(b,a)) +#define FIMUL_L(a) CHOICE(fimull a, fimull a, fimull a) +#define FIMUL_W(a) CHOICE(fimul a, fimuls a, fimuls a) +#define FNOP CHOICE(fnop, fnop, fnop) +#define FPATAN CHOICE(fpatan, fpatan, fpatan) +#define FPREM CHOICE(fprem, fprem, fprem) +#define FPREM1 CHOICE(fprem1, fprem1, fprem1) +#define FPTAN CHOICE(fptan, fptan, fptan) +#define FRNDINT CHOICE(frndint, frndint, frndint) +#define FRSTOR(a) CHOICE(frstor a, frstor a, frstor a) +#define FSAVE(a) CHOICE(fsave a, wait; fnsave a, wait; fsave a) +#define FNSAVE(a) CHOICE(fnsave a, fnsave a, fsave a) +#define FSCALE CHOICE(fscale, fscale, fscale) +#define FSIN CHOICE(fsin, fsin, fsin) +#define FSINCOS CHOICE(fsincos, fsincos, fsincos) +#define FSQRT CHOICE(fsqrt, fsqrt, fsqrt) +#define FST_D(a) CHOICE(fstl a, fstl a, fstd a) +#define FST_S(a) CHOICE(fsts a, fsts a, fsts a) +#define FSTP_X(a) CHOICE(fstpt a, fstpt a, fstpx a) +#define FSTP_D(a) CHOICE(fstpl a, fstpl a, fstpd a) +#define FSTP_S(a) CHOICE(fstps a, fstps a, fstps a) +#define FSTP(a) CHOICE(fstp a, fstp a, fstp a) +#define FSTCW(a) CHOICE(fstcw a, wait; fnstcw a, wait; fstcw a) +#define FNSTCW(a) CHOICE(fnstcw a, fnstcw a, fstcw a) +#define FSTENV(a) CHOICE(fstenv a, wait; fnstenv a, fstenv a) +#define FNSTENV(a) CHOICE(fnstenv a, fnstenv a, fstenv a) +#define FSTSW(a) CHOICE(fstsw a, wait; fnstsw a, wait; fstsw a) +#define FNSTSW(a) CHOICE(fnstsw a, fnstsw a, fstsw a) +#define FSUB_S(a) CHOICE(fsubs a, fsubs a, fsubs a) +#define FSUB_D(a) CHOICE(fsubl a, fsubl a, fsubd a) +#define FSUB2(a, b) CHOICE(fsub ARG2(a,b), fsub ARG2(a,b), fsub ARG2(b,a)) +#define FSUBP(a, b) CHOICE(fsubp ARG2(a,b), fsubp ARG2(a,b), fsubp ARG2(b,a)) +#define FISUB_L(a) CHOICE(fisubl a, fisubl a, fisubl a) +#define FISUB_W(a) CHOICE(fisub a, fisubs a, fisubs a) +#define FSUBR_S(a) CHOICE(fsubrs a, fsubrs a, fsubrs a) +#define FSUBR_D(a) CHOICE(fsubrl a, fsubrl a, fsubrd a) +#define FSUBR2(a, b) CHOICE(fsubr ARG2(a,b), fsubr ARG2(a,b), fsubr ARG2(b,a)) +#define FSUBRP(a, b) CHOICE(fsubrp ARG2(a,b), fsubrp ARG2(a,b), fsubrp ARG2(b,a)) +#define FISUBR_L(a) CHOICE(fisubrl a, fisubrl a, fisubrl a) +#define FISUBR_W(a) CHOICE(fisubr a, fisubrs a, fisubrs a) +#define FTST CHOICE(ftst, ftst, ftst) +#define FUCOM(a) CHOICE(fucom a, fucom a, fucom a) +#define FUCOMP(a) CHOICE(fucomp a, fucomp a, fucomp a) +#define FUCOMPP CHOICE(fucompp, fucompp, fucompp) +#define FWAIT CHOICE(wait, wait, wait) +#define FXAM CHOICE(fxam, fxam, fxam) +#define FXCH(a) CHOICE(fxch a, fxch a, fxch a) +#define FXTRACT CHOICE(fxtract, fxtract, fxtract) +#define FYL2X CHOICE(fyl2x, fyl2x, fyl2x) +#define FYL2XP1 CHOICE(fyl2xp1, fyl2xp1, fyl2xp1) + +/* New instructions */ +#define CPUID CHOICE(D_BYTE ARG2(15, 162), cpuid, D_BYTE ARG2(15, 162)) +#define RDTSC CHOICE(D_BYTE ARG2(15, 49), rdtsc, D_BYTE ARG2(15, 49)) + +#else /* NASM_ASSEMBLER || MASM_ASSEMBLER is defined */ + + /****************************************/ + /* */ + /* Intel style assemblers. */ + /* (NASM and MASM) */ + /* */ + /****************************************/ + +#define P_EAX EAX +#define L_EAX EAX +#define W_AX AX +#define B_AH AH +#define B_AL AL + +#define P_EBX EBX +#define L_EBX EBX +#define W_BX BX +#define B_BH BH +#define B_BL BL + +#define P_ECX ECX +#define L_ECX ECX +#define W_CX CX +#define B_CH CH +#define B_CL CL + +#define P_EDX EDX +#define L_EDX EDX +#define W_DX DX +#define B_DH DH +#define B_DL DL + +#define P_EBP EBP +#define L_EBP EBP +#define W_BP BP + +#define P_ESI ESI +#define L_ESI ESI +#define W_SI SI + +#define P_EDI EDI +#define L_EDI EDI +#define W_DI DI + +#define P_ESP ESP +#define L_ESP ESP +#define W_SP SP + +#define W_CS CS +#define W_SS SS +#define W_DS DS +#define W_ES ES +#define W_FS FS +#define W_GS GS + +#define X_ST ST +#define D_ST ST +#define L_ST ST + +#define P_MM0 mm0 +#define P_MM1 mm1 +#define P_MM2 mm2 +#define P_MM3 mm3 +#define P_MM4 mm4 +#define P_MM5 mm5 +#define P_MM6 mm6 +#define P_MM7 mm7 + +#define P_XMM0 xmm0 +#define P_XMM1 xmm1 +#define P_XMM2 xmm2 +#define P_XMM3 xmm3 +#define P_XMM4 xmm4 +#define P_XMM5 xmm5 +#define P_XMM6 xmm6 +#define P_XMM7 xmm7 + +#define CONCAT(x, y) x ## y +#define CONCAT3(x, y, z) x ## y ## z + +#if defined(NASM_ASSEMBLER) + +#define ST(n) st ## n +#define ST0 st0 + +#define TBYTE_PTR tword +#define QWORD_PTR qword +#define DWORD_PTR dword +#define WORD_PTR word +#define BYTE_PTR byte + +#define OFFSET + +#define GLOBL GLOBAL +#define ALIGNTEXT32 ALIGN 32 +#define ALIGNTEXT16 ALIGN 16 +#define ALIGNTEXT8 ALIGN 8 +#define ALIGNTEXT4 ALIGN 4 +#define ALIGNTEXT2 ALIGN 2 +#define ALIGNTEXT32ifNOP ALIGN 32 +#define ALIGNTEXT16ifNOP ALIGN 16 +#define ALIGNTEXT8ifNOP ALIGN 8 +#define ALIGNTEXT4ifNOP ALIGN 4 +#define ALIGNDATA32 ALIGN 32 +#define ALIGNDATA16 ALIGN 16 +#define ALIGNDATA8 ALIGN 8 +#define ALIGNDATA4 ALIGN 4 +#define ALIGNDATA2 ALIGN 2 +#define FILE(s) +#define STRING(s) db s +#define D_LONG dd +#define D_WORD dw +#define D_BYTE db +/* #define SPACE */ +/* #define COMM */ +#if defined(__WATCOMC__) +SECTION _TEXT public align=16 class=CODE use32 flat +SECTION _DATA public align=16 class=DATA use32 flat +#define SEG_TEXT SECTION _TEXT +#define SEG_DATA SECTION _DATA +#define SEG_BSS SECTION .bss +#else +#define SEG_DATA SECTION .data +#define SEG_TEXT SECTION .text +#define SEG_BSS SECTION .bss +#endif + +#define D_SPACE(n) db n REP 0 + +#define AS_BEGIN + +/* Jcc's should be handled better than this... */ +#define NEAR near + +#else /* MASM */ + +#define TBYTE_PTR tbyte ptr +#define QWORD_PTR qword ptr +#define DWORD_PTR dword ptr +#define WORD_PTR word ptr +#define BYTE_PTR byte ptr + +#define OFFSET offset + +#define GLOBL GLOBAL +#define ALIGNTEXT32 ALIGN 32 +#define ALIGNTEXT16 ALIGN 16 +#define ALIGNTEXT8 ALIGN 8 +#define ALIGNTEXT4 ALIGN 4 +#define ALIGNTEXT2 ALIGN 2 +#define ALIGNTEXT32ifNOP ALIGN 32 +#define ALIGNTEXT16ifNOP ALIGN 16 +#define ALIGNTEXT8ifNOP ALIGN 8 +#define ALIGNTEXT4ifNOP ALIGN 4 +#define ALIGNDATA32 ALIGN 32 +#define ALIGNDATA16 ALIGN 16 +#define ALIGNDATA8 ALIGN 8 +#define ALIGNDATA4 ALIGN 4 +#define ALIGNDATA2 ALIGN 2 +#define FILE(s) +#define STRING(s) db s +#define D_LONG dd +#define D_WORD dw +#define D_BYTE db +/* #define SPACE */ +/* #define COMM */ +#define SEG_DATA .DATA +#define SEG_TEXT .CODE +#define SEG_BSS .DATA + +#define D_SPACE(n) db n REP 0 + +#define AS_BEGIN + +#define NEAR + +#endif + +#if defined(Lynx) || (defined(SYSV) || defined(SVR4)) \ + || (defined(__linux__) || defined(__OS2ELF__)) && defined(__ELF__) \ + || defined(__FreeBSD__) && __FreeBSD__ >= 3 +#define GLNAME(a) a +#else +#define GLNAME(a) CONCAT(_, a) +#endif + +/* + * Addressing Modes + */ + +/* Immediate Mode */ +#define P_ADDR(a) OFFSET a +#define X_ADDR(a) OFFSET a +#define D_ADDR(a) OFFSET a +#define L_ADDR(a) OFFSET a +#define W_ADDR(a) OFFSET a +#define B_ADDR(a) OFFSET a + +#define P_CONST(a) a +#define X_CONST(a) a +#define D_CONST(a) a +#define L_CONST(a) a +#define W_CONST(a) a +#define B_CONST(a) a + +/* Indirect Mode */ +#ifdef NASM_ASSEMBLER +#define P_CONTENT(a) [a] +#define X_CONTENT(a) TBYTE_PTR [a] +#define D_CONTENT(a) QWORD_PTR [a] +#define L_CONTENT(a) DWORD_PTR [a] +#define W_CONTENT(a) WORD_PTR [a] +#define B_CONTENT(a) BYTE_PTR [a] +#else +#define P_CONTENT(a) a +#define X_CONTENT(a) TBYTE_PTR a +#define D_CONTENT(a) QWORD_PTR a +#define L_CONTENT(a) DWORD_PTR a +#define W_CONTENT(a) WORD_PTR a +#define B_CONTENT(a) BYTE_PTR a +#endif + +/* Register a indirect */ +#define P_REGIND(a) [a] +#define X_REGIND(a) TBYTE_PTR [a] +#define D_REGIND(a) QWORD_PTR [a] +#define L_REGIND(a) DWORD_PTR [a] +#define W_REGIND(a) WORD_PTR [a] +#define B_REGIND(a) BYTE_PTR [a] + +/* Register b indirect plus displacement a */ +#define P_REGOFF(a, b) [b + a] +#define X_REGOFF(a, b) TBYTE_PTR [b + a] +#define D_REGOFF(a, b) QWORD_PTR [b + a] +#define L_REGOFF(a, b) DWORD_PTR [b + a] +#define W_REGOFF(a, b) WORD_PTR [b + a] +#define B_REGOFF(a, b) BYTE_PTR [b + a] + +/* Reg indirect Base + Index + Displacement - this is mainly for 16-bit mode + * which has no scaling + */ +#define P_REGBID(b, i, d) [b + i + d] +#define X_REGBID(b, i, d) TBYTE_PTR [b + i + d] +#define D_REGBID(b, i, d) QWORD_PTR [b + i + d] +#define L_REGBID(b, i, d) DWORD_PTR [b + i + d] +#define W_REGBID(b, i, d) WORD_PTR [b + i + d] +#define B_REGBID(b, i, d) BYTE_PTR [b + i + d] + +/* Reg indirect Base + (Index * Scale) */ +#define P_REGBIS(b, i, s) [b + i * s] +#define X_REGBIS(b, i, s) TBYTE_PTR [b + i * s] +#define D_REGBIS(b, i, s) QWORD_PTR [b + i * s] +#define L_REGBIS(b, i, s) DWORD_PTR [b + i * s] +#define W_REGBIS(b, i, s) WORD_PTR [b + i * s] +#define B_REGBIS(b, i, s) BYTE_PTR [b + i * s] + +/* Reg indirect Base + (Index * Scale) + Displacement */ +#define P_REGBISD(b, i, s, d) [b + i * s + d] +#define X_REGBISD(b, i, s, d) TBYTE_PTR [b + i * s + d] +#define D_REGBISD(b, i, s, d) QWORD_PTR [b + i * s + d] +#define L_REGBISD(b, i, s, d) DWORD_PTR [b + i * s + d] +#define W_REGBISD(b, i, s, d) WORD_PTR [b + i * s + d] +#define B_REGBISD(b, i, s, d) BYTE_PTR [b + i * s + d] + +/* Displaced Scaled Index: */ +#define P_REGDIS(d, i, s) [i * s + d] +#define X_REGDIS(d, i, s) TBYTE_PTR [i * s + d] +#define D_REGDIS(d, i, s) QWORD_PTR [i * s + d] +#define L_REGDIS(d, i, s) DWORD_PTR [i * s + d] +#define W_REGDIS(d, i, s) WORD_PTR [i * s + d] +#define B_REGDIS(d, i, s) BYTE_PTR [i * s + d] + +/* Indexed Base: */ +#define P_REGBI(b, i) [b + i] +#define X_REGBI(b, i) TBYTE_PTR [b + i] +#define D_REGBI(b, i) QWORD_PTR [b + i] +#define L_REGBI(b, i) DWORD_PTR [b + i] +#define W_REGBI(b, i) WORD_PTR [b + i] +#define B_REGBI(b, i) BYTE_PTR [b + i] + +/* Displaced Base: */ +#define P_REGDB(d, b) [b + d] +#define X_REGDB(d, b) TBYTE_PTR [b + d] +#define D_REGDB(d, b) QWORD_PTR [b + d] +#define L_REGDB(d, b) DWORD_PTR [b + d] +#define W_REGDB(d, b) WORD_PTR [b + d] +#define B_REGDB(d, b) BYTE_PTR [b + d] + +/* Variable indirect: */ +#define VARINDIRECT(var) [var] + +/* Use register contents as jump/call target: */ +#define CODEPTR(reg) P_(reg) + +/* + * Redefine assembler commands + */ + +#define P_(a) P_ ## a +#define X_(a) X_ ## a +#define D_(a) D_ ## a +#define SR_(a) W_ ## a +#define S_(a) L_ ## a +#define L_(a) L_ ## a +#define W_(a) W_ ## a +#define B_(a) B_ ## a + +#define AAA aaa +#define AAD aad +#define AAM aam +#define AAS aas +#define ADC_L(a, b) adc L_(b), L_(a) +#define ADC_W(a, b) adc W_(b), W_(a) +#define ADC_B(a, b) adc B_(b), B_(a) +#define ADD_L(a, b) add L_(b), L_(a) +#define ADD_W(a, b) add W_(b), W_(a) +#define ADD_B(a, b) add B_(b), B_(a) +#define AND_L(a, b) and L_(b), L_(a) +#define AND_W(a, b) and W_(b), W_(a) +#define AND_B(a, b) and B_(b), B_(a) +#define ARPL(a,b) arpl W_(b), a +#define BOUND_L(a, b) bound L_(b), L_(a) +#define BOUND_W(a, b) bound W_(b), W_(a) +#define BSF_L(a, b) bsf L_(b), L_(a) +#define BSF_W(a, b) bsf W_(b), W_(a) +#define BSR_L(a, b) bsr L_(b), L_(a) +#define BSR_W(a, b) bsr W_(b), W_(a) +#define BT_L(a, b) bt L_(b), L_(a) +#define BT_W(a, b) bt W_(b), W_(a) +#define BTC_L(a, b) btc L_(b), L_(a) +#define BTC_W(a, b) btc W_(b), W_(a) +#define BTR_L(a, b) btr L_(b), L_(a) +#define BTR_W(a, b) btr W_(b), W_(a) +#define BTS_L(a, b) bts L_(b), L_(a) +#define BTS_W(a, b) bts W_(b), W_(a) +#define CALL(a) call a +#define CALLF(s,a) call far s:a +#define CBW cbw +#define CWDE cwde +#define CLC clc +#define CLD cld +#define CLI cli +#define CLTS clts +#define CMC cmc +#define CMP_L(a, b) cmp L_(b), L_(a) +#define CMP_W(a, b) cmp W_(b), W_(a) +#define CMP_B(a, b) cmp B_(b), B_(a) +#define CMPS_L cmpsd +#define CMPS_W cmpsw +#define CMPS_B cmpsb +#define CPUID cpuid +#define CWD cwd +#define CDQ cdq +#define DAA daa +#define DAS das +#define DEC_L(a) dec L_(a) +#define DEC_W(a) dec W_(a) +#define DEC_B(a) dec B_(a) +#define DIV_L(a) div L_(a) +#define DIV_W(a) div W_(a) +#define DIV_B(a) div B_(a) +#define ENTER(a,b) enter b, a +#define HLT hlt +#define IDIV_L(a) idiv L_(a) +#define IDIV_W(a) idiv W_(a) +#define IDIV_B(a) idiv B_(a) +#define IMUL_L(a, b) imul L_(b), L_(a) +#define IMUL_W(a, b) imul W_(b), W_(a) +#define IMUL_B(a) imul B_(a) +#define IN_L in EAX, DX +#define IN_W in AX, DX +#define IN_B in AL, DX +#define IN1_L(a) in1 L_(a) +#define IN1_W(a) in1 W_(a) +#define IN1_B(a) in1 B_(a) +#define INC_L(a) inc L_(a) +#define INC_W(a) inc W_(a) +#define INC_B(a) inc B_(a) +#define INS_L ins +#define INS_W ins +#define INS_B ins +#define INT(a) int B_(a) +#define INT3 int3 +#define INTO into +#define IRET iret +#define IRETD iretd +#define JA(a) ja NEAR a +#define JAE(a) jae NEAR a +#define JB(a) jb NEAR a +#define JBE(a) jbe NEAR a +#define JC(a) jc NEAR a +#define JE(a) je NEAR a +#define JG(a) jg NEAR a +#define JGE(a) jge NEAR a +#define JL(a) jl NEAR a +#define JLE(a) jle NEAR a +#define JNA(a) jna NEAR a +#define JNAE(a) jnae NEAR a +#define JNB(a) jnb NEAR a +#define JNBE(a) jnbe NEAR a +#define JNC(a) jnc NEAR a +#define JNE(a) jne NEAR a +#define JNG(a) jng NEAR a +#define JNGE(a) jnge NEAR a +#define JNL(a) jnl NEAR a +#define JNLE(a) jnle NEAR a +#define JNO(a) jno NEAR a +#define JNP(a) jnp NEAR a +#define JNS(a) jns NEAR a +#define JNZ(a) jnz NEAR a +#define JO(a) jo NEAR a +#define JP(a) jp NEAR a +#define JPE(a) jpe NEAR a +#define JPO(a) jpo NEAR a +#define JS(a) js NEAR a +#define JZ(a) jz NEAR a +#define JMP(a) jmp a +#define JMPF(s,a) jmp far s:a +#define LAHF lahf +#define LAR(a, b) lar b, a +#define LEA_L(a, b) lea P_(b), P_(a) +#define LEA_W(a, b) lea P_(b), P_(a) +#define LEAVE leave +#define LGDT(a) lgdt a +#define LIDT(a) lidt a +#define LDS(a, b) lds b, P_(a) +#define LES(a, b) les b, P_(a) +#define LFS(a, b) lfs b, P_(a) +#define LGS(a, b) lgs b, P_(a) +#define LSS(a, b) lss b, P_(a) +#define LLDT(a) lldt a +#define LMSW(a) lmsw a +#define LOCK lock +#define LODS_L lodsd +#define LODS_W lodsw +#define LODS_B lodsb +#define LOOP(a) loop a +#define LOOPE(a) loope a +#define LOOPZ(a) loopz a +#define LOOPNE(a) loopne a +#define LOOPNZ(a) loopnz a +#define LSL(a, b) lsl b, a +#define LTR(a) ltr a +#define MOV_SR(a, b) mov SR_(b), SR_(a) +#define MOV_L(a, b) mov L_(b), L_(a) +#define MOV_W(a, b) mov W_(b), W_(a) +#define MOV_B(a, b) mov B_(b), B_(a) +#define MOVS_L movsd +#define MOVS_W movsw +#define MOVS_B movsb +#define MOVSX_BL(a, b) movsx B_(b), B_(a) +#define MOVSX_BW(a, b) movsx B_(b), B_(a) +#define MOVSX_WL(a, b) movsx W_(b), W_(a) +#define MOVZX_BL(a, b) movzx B_(b), B_(a) +#define MOVZX_BW(a, b) movzx B_(b), B_(a) +#define MOVZX_WL(a, b) movzx W_(b), W_(a) +#define MUL_L(a) mul L_(a) +#define MUL_W(a) mul W_(a) +#define MUL_B(a) mul B_(a) +#define NEG_L(a) neg L_(a) +#define NEG_W(a) neg W_(a) +#define NEG_B(a) neg B_(a) +#define NOP nop +#define NOT_L(a) not L_(a) +#define NOT_W(a) not W_(a) +#define NOT_B(a) not B_(a) +#define OR_L(a,b) or L_(b), L_(a) +#define OR_W(a,b) or W_(b), W_(a) +#define OR_B(a,b) or B_(b), B_(a) +#define OUT_L out DX, EAX +#define OUT_W out DX, AX +#define OUT_B out DX, AL +#define OUT1_L(a) out1 L_(a) +#define OUT1_W(a) out1 W_(a) +#define OUT1_B(a) out1 B_(a) +#define OUTS_L outsd +#define OUTS_W outsw +#define OUTS_B outsb +#define POP_SR(a) pop SR_(a) +#define POP_L(a) pop L_(a) +#define POP_W(a) pop W_(a) +#define POPA_L popad +#define POPA_W popa +#define POPF_L popfd +#define POPF_W popf +#define PUSH_SR(a) push SR_(a) +#define PUSH_L(a) push L_(a) +#define PUSH_W(a) push W_(a) +#define PUSH_B(a) push B_(a) +#define PUSHA_L pushad +#define PUSHA_W pusha +#define PUSHF_L pushfd +#define PUSHF_W pushf +#define RCL_L(a, b) rcl L_(b), L_(a) +#define RCL_W(a, b) rcl W_(b), W_(a) +#define RCL_B(a, b) rcl B_(b), B_(a) +#define RCR_L(a, b) rcr L_(b), L_(a) +#define RCR_W(a, b) rcr W_(b), W_(a) +#define RCR_B(a, b) rcr B_(b), B_(a) +#define RDTSC rdtsc +#define ROL_L(a, b) rol L_(b), L_(a) +#define ROL_W(a, b) rol W_(b), W_(a) +#define ROL_B(a, b) rol B_(b), B_(a) +#define ROR_L(a, b) ror L_(b), L_(a) +#define ROR_W(a, b) ror W_(b), W_(a) +#define ROR_B(a, b) ror B_(b), B_(a) +#define REP rep +#define REPE repe +#define REPNE repne +#define REPNZ REPNE +#define REPZ REPE +#define RET ret +#define SAHF sahf +#define SAL_L(a, b) sal L_(b), B_(a) +#define SAL_W(a, b) sal W_(b), B_(a) +#define SAL_B(a, b) sal B_(b), B_(a) +#define SAR_L(a, b) sar L_(b), B_(a) +#define SAR_W(a, b) sar W_(b), B_(a) +#define SAR_B(a, b) sar B_(b), B_(a) +#define SBB_L(a, b) sbb L_(b), L_(a) +#define SBB_W(a, b) sbb W_(b), W_(a) +#define SBB_B(a, b) sbb B_(b), B_(a) +#define SCAS_L scas +#define SCAS_W scas +#define SCAS_B scas +#define SETA(a) seta a +#define SETAE(a) setae a +#define SETB(a) setb a +#define SETBE(a) setbe a +#define SETC(a) setc a +#define SETE(a) sete a +#define SETG(a) setg a +#define SETGE(a) setge a +#define SETL(a) setl a +#define SETLE(a) setle a +#define SETNA(a) setna a +#define SETNAE(a) setnae a +#define SETNB(a) setnb a +#define SETNBE(a) setnbe a +#define SETNC(a) setnc a +#define SETNE(a) setne a +#define SETNG(a) setng a +#define SETNGE(a) setnge a +#define SETNL(a) setnl a +#define SETNLE(a) setnle a +#define SETNO(a) setno a +#define SETNP(a) setnp a +#define SETNS(a) setns a +#define SETNZ(a) setnz a +#define SETO(a) seto a +#define SETP(a) setp a +#define SETPE(a) setpe a +#define SETPO(a) setpo a +#define SETS(a) sets a +#define SETZ(a) setz a +#define SGDT(a) sgdt a +#define SIDT(a) sidt a +#define SHL_L(a, b) shl L_(b), B_(a) +#define SHL_W(a, b) shl W_(b), B_(a) +#define SHL_B(a, b) shl B_(b), B_(a) +#define SHLD_L(a,b,c) shld +#define SHLD2_L(a,b) shld L_(b), L_(a) +#define SHLD_W(a,b,c) shld +#define SHLD2_W(a,b) shld W_(b), W_(a) +#define SHR_L(a, b) shr L_(b), B_(a) +#define SHR_W(a, b) shr W_(b), B_(a) +#define SHR_B(a, b) shr B_(b), B_(a) +#define SHRD_L(a,b,c) shrd +#define SHRD2_L(a,b) shrd L_(b), L_(a) +#define SHRD_W(a,b,c) shrd +#define SHRD2_W(a,b) shrd W_(b), W_(a) +#define SLDT(a) sldt a +#define SMSW(a) smsw a +#define STC stc +#define STD std +#define STI sti +#define STOS_L stosd +#define STOS_W stosw +#define STOS_B stosb +#define STR(a) str a +#define SUB_L(a, b) sub L_(b), L_(a) +#define SUB_W(a, b) sub W_(b), W_(a) +#define SUB_B(a, b) sub B_(b), B_(a) +#define TEST_L(a, b) test L_(b), L_(a) +#define TEST_W(a, b) test W_(b), W_(a) +#define TEST_B(a, b) test B_(b), B_(a) +#define VERR(a) verr a +#define VERW(a) verw a +#define WAIT wait +#define XCHG_L(a, b) xchg L_(b), L_(a) +#define XCHG_W(a, b) xchg W_(b), W_(a) +#define XCHG_B(a, b) xchg B_(b), B_(a) +#define XLAT xlat +#define XOR_L(a, b) xor L_(b), L_(a) +#define XOR_W(a, b) xor W_(b), W_(a) +#define XOR_B(a, b) xor B_(b), B_(a) + + +/* Floating Point Instructions */ +#define F2XM1 f2xm1 +#define FABS fabs +#define FADD_D(a) fadd D_(a) +#define FADD_S(a) fadd S_(a) +#define FADD2(a, b) fadd b, a +#define FADDP(a, b) faddp b, a +#define FIADD_L(a) fiadd L_(a) +#define FIADD_W(a) fiadd W_(a) +#define FBLD(a) fbld a +#define FBSTP(a) fbstp a +#define FCHS fchs +#define FCLEX fclex +#define FNCLEX fnclex +#define FCOM(a) fcom a +#define FCOM_D(a) fcom D_(a) +#define FCOM_S(a) fcom S_(a) +#define FCOMP(a) fcomp a +#define FCOMP_D(a) fcomp D_(a) +#define FCOMP_S(a) fcomp S_(a) +#define FCOMPP fcompp +#define FCOS fcos +#define FDECSTP fdecstp +#define FDIV_D(a) fdiv D_(a) +#define FDIV_S(a) fdiv S_(a) +#define FDIV2(a, b) fdiv b, a +#define FDIVP(a, b) fdivp b, a +#define FIDIV_L(a) fidiv L_(a) +#define FIDIV_W(a) fidiv W_(a) +#define FDIVR_D(a) fdivr D_(a) +#define FDIVR_S(a) fdivr S_(a) +#define FDIVR2(a, b) fdivr b, a +#define FDIVRP(a, b) fdivrp b, a +#define FIDIVR_L(a) fidivr L_(a) +#define FIDIVR_W(a) fidivr W_(a) +#define FFREE(a) ffree a +#define FICOM_L(a) ficom L_(a) +#define FICOM_W(a) ficom W_(a) +#define FICOMP_L(a) ficomp L_(a) +#define FICOMP_W(a) ficomp W_(a) +#define FILD_Q(a) fild D_(a) +#define FILD_L(a) fild L_(a) +#define FILD_W(a) fild W_(a) +#define FINCSTP fincstp +#define FINIT finit +#define FNINIT fninit +#define FIST_L(a) fist L_(a) +#define FIST_W(a) fist W_(a) +#define FISTP_Q(a) fistp D_(a) +#define FISTP_L(a) fistp L_(a) +#define FISTP_W(a) fistp W_(a) +#define FLD_X(a) fld X_(a) +#define FLD_D(a) fld D_(a) +#define FLD_S(a) fld S_(a) +#define FLD1 fld1 +#define FLDL2T fldl2t +#define FLDL2E fldl2e +#define FLDPI fldpi +#define FLDLG2 fldlg2 +#define FLDLN2 fldln2 +#define FLDZ fldz +#define FLDCW(a) fldcw a +#define FLDENV(a) fldenv a +#define FMUL_S(a) fmul S_(a) +#define FMUL_D(a) fmul D_(a) +#define FMUL2(a, b) fmul b, a +#define FMULP(a, b) fmulp b, a +#define FIMUL_L(a) fimul L_(a) +#define FIMUL_W(a) fimul W_(a) +#define FNOP fnop +#define FPATAN fpatan +#define FPREM fprem +#define FPREM1 fprem1 +#define FPTAN fptan +#define FRNDINT frndint +#define FRSTOR(a) frstor a +#define FSAVE(a) fsave a +#define FNSAVE(a) fnsave a +#define FSCALE fscale +#define FSIN fsin +#define FSINCOS fsincos +#define FSQRT fsqrt +#define FST_D(a) fst D_(a) +#define FST_S(a) fst S_(a) +#define FSTP_X(a) fstp X_(a) +#define FSTP_D(a) fstp D_(a) +#define FSTP_S(a) fstp S_(a) +#define FSTP(a) fstp a +#define FSTCW(a) fstcw a +#define FNSTCW(a) fnstcw a +#define FSTENV(a) fstenv a +#define FNSTENV(a) fnstenv a +#define FSTSW(a) fstsw a +#define FNSTSW(a) fnstsw a +#define FSUB_S(a) fsub S_(a) +#define FSUB_D(a) fsub D_(a) +#define FSUB2(a, b) fsub b, a +#define FSUBP(a, b) fsubp b, a +#define FISUB_L(a) fisub L_(a) +#define FISUB_W(a) fisub W_(a) +#define FSUBR_S(a) fsubr S_(a) +#define FSUBR_D(a) fsubr D_(a) +#define FSUBR2(a, b) fsubr b, a +#define FSUBRP(a, b) fsubrp b, a +#define FISUBR_L(a) fisubr L_(a) +#define FISUBR_W(a) fisubr W_(a) +#define FTST ftst +#define FUCOM(a) fucom a +#define FUCOMP(a) fucomp a +#define FUCOMPP fucompp +#define FWAIT fwait +#define FXAM fxam +#define FXCH(a) fxch a +#define FXTRACT fxtract +#define FYL2X fyl2x +#define FYL2XP1 fyl2xp1 + +#endif /* NASM_ASSEMBLER, MASM_ASSEMBLER */ + + /****************************************/ + /* */ + /* Extensions to x86 insn set - */ + /* MMX, 3DNow! */ + /* */ + /****************************************/ + +#if defined(NASM_ASSEMBLER) || defined(MASM_ASSEMBLER) +#define P_ARG1(a) P_ ## a +#define P_ARG2(a, b) P_ ## b, P_ ## a +#define P_ARG3(a, b, c) P_ ## c, P_ ## b, P_ ## a +#else +#define P_ARG1(a) a +#define P_ARG2(a, b) a, b +#define P_ARG3(a, b, c) a, b, c +#endif + +/* MMX */ +#define MOVD(a, b) movd P_ARG2(a, b) +#define MOVQ(a, b) movq P_ARG2(a, b) + +#define PADDB(a, b) paddb P_ARG2(a, b) +#define PADDW(a, b) paddw P_ARG2(a, b) +#define PADDD(a, b) paddd P_ARG2(a, b) + +#define PADDSB(a, b) paddsb P_ARG2(a, b) +#define PADDSW(a, b) paddsw P_ARG2(a, b) + +#define PADDUSB(a, b) paddusb P_ARG2(a, b) +#define PADDUSW(a, b) paddusw P_ARG2(a, b) + +#define PSUBB(a, b) psubb P_ARG2(a, b) +#define PSUBW(a, b) psubw P_ARG2(a, b) +#define PSUBD(a, b) psubd P_ARG2(a, b) + +#define PSUBSB(a, b) psubsb P_ARG2(a, b) +#define PSUBSW(a, b) psubsw P_ARG2(a, b) + +#define PSUBUSB(a, b) psubusb P_ARG2(a, b) +#define PSUBUSW(a, b) psubusw P_ARG2(a, b) + +#define PCMPEQB(a, b) pcmpeqb P_ARG2(a, b) +#define PCMPEQW(a, b) pcmpeqw P_ARG2(a, b) +#define PCMPEQD(a, b) pcmpeqd P_ARG2(a, b) + +#define PCMPGTB(a, b) pcmpgtb P_ARG2(a, b) +#define PCMPGTW(a, b) pcmpgtw P_ARG2(a, b) +#define PCMPGTD(a, b) pcmpgtd P_ARG2(a, b) + +#define PMULHW(a, b) pmulhw P_ARG2(a, b) +#define PMULLW(a, b) pmullw P_ARG2(a, b) + +#define PMADDWD(a, b) pmaddwd P_ARG2(a, b) + +#define PAND(a, b) pand P_ARG2(a, b) + +#define PANDN(a, b) pandn P_ARG2(a, b) + +#define POR(a, b) por P_ARG2(a, b) + +#define PXOR(a, b) pxor P_ARG2(a, b) + +#define PSRAW(a, b) psraw P_ARG2(a, b) +#define PSRAD(a, b) psrad P_ARG2(a, b) + +#define PSRLW(a, b) psrlw P_ARG2(a, b) +#define PSRLD(a, b) psrld P_ARG2(a, b) +#define PSRLQ(a, b) psrlq P_ARG2(a, b) + +#define PSLLW(a, b) psllw P_ARG2(a, b) +#define PSLLD(a, b) pslld P_ARG2(a, b) +#define PSLLQ(a, b) psllq P_ARG2(a, b) + +#define PACKSSWB(a, b) packsswb P_ARG2(a, b) +#define PACKSSDW(a, b) packssdw P_ARG2(a, b) +#define PACKUSWB(a, b) packuswb P_ARG2(a, b) + +#define PUNPCKHBW(a, b) punpckhbw P_ARG2(a, b) +#define PUNPCKHWD(a, b) punpckhwd P_ARG2(a, b) +#define PUNPCKHDQ(a, b) punpckhdq P_ARG2(a, b) +#define PUNPCKLBW(a, b) punpcklbw P_ARG2(a, b) +#define PUNPCKLWD(a, b) punpcklwd P_ARG2(a, b) +#define PUNPCKLDQ(a, b) punpckldq P_ARG2(a, b) + +#define EMMS emms + +/* AMD 3DNow! */ +#define PAVGUSB(a, b) pavgusb P_ARG2(a, b) +#define PFADD(a, b) pfadd P_ARG2(a, b) +#define PFSUB(a, b) pfsub P_ARG2(a, b) +#define PFSUBR(a, b) pfsubr P_ARG2(a, b) +#define PFACC(a, b) pfacc P_ARG2(a, b) +#define PFCMPGE(a, b) pfcmpge P_ARG2(a, b) +#define PFCMPGT(a, b) pfcmpgt P_ARG2(a, b) +#define PFCMPEQ(a, b) pfcmpeq P_ARG2(a, b) +#define PFMIN(a, b) pfmin P_ARG2(a, b) +#define PFMAX(a, b) pfmax P_ARG2(a, b) +#define PI2FD(a, b) pi2fd P_ARG2(a, b) +#define PF2ID(a, b) pf2id P_ARG2(a, b) +#define PFRCP(a, b) pfrcp P_ARG2(a, b) +#define PFRSQRT(a, b) pfrsqrt P_ARG2(a, b) +#define PFMUL(a, b) pfmul P_ARG2(a, b) +#define PFRCPIT1(a, b) pfrcpit1 P_ARG2(a, b) +#define PFRSQIT1(a, b) pfrsqit1 P_ARG2(a, b) +#define PFRCPIT2(a, b) pfrcpit2 P_ARG2(a, b) +#define PMULHRW(a, b) pmulhrw P_ARG2(a, b) + +#define FEMMS femms +#define PREFETCH(a) prefetch P_ARG1(a) +#define PREFETCHW(a) prefetchw P_ARG1(a) + +/* Intel SSE */ +#define ADDPS(a, b) addps P_ARG2(a, b) +#define ADDSS(a, b) addss P_ARG2(a, b) +#define ANDNPS(a, b) andnps P_ARG2(a, b) +#define ANDPS(a, b) andps P_ARG2(a, b) +/* NASM only knows the pseudo ops for these. +#define CMPPS(a, b, c) cmpps P_ARG3(a, b, c) +#define CMPSS(a, b, c) cmpss P_ARG3(a, b, c) +*/ +#define CMPEQPS(a, b) cmpeqps P_ARG2(a, b) +#define CMPLTPS(a, b) cmpltps P_ARG2(a, b) +#define CMPLEPS(a, b) cmpleps P_ARG2(a, b) +#define CMPUNORDPS(a, b) cmpunordps P_ARG2(a, b) +#define CMPNEQPS(a, b) cmpneqps P_ARG2(a, b) +#define CMPNLTPS(a, b) cmpnltps P_ARG2(a, b) +#define CMPNLEPS(a, b) cmpnleps P_ARG2(a, b) +#define CMPORDPS(a, b) cmpordps P_ARG2(a, b) +#define CMPEQSS(a, b) cmpeqss P_ARG2(a, b) +#define CMPLTSS(a, b) cmpltss P_ARG2(a, b) +#define CMPLESS(a, b) cmpless P_ARG2(a, b) +#define CMPUNORDSS(a, b) cmpunordss P_ARG2(a, b) +#define CMPNEQSS(a, b) cmpneqss P_ARG2(a, b) +#define CMPNLTSS(a, b) cmpnltss P_ARG2(a, b) +#define CMPNLESS(a, b) cmpnless P_ARG2(a, b) +#define CMPORDSS(a, b) cmpordss P_ARG2(a, b) +#define COMISS(a, b) comiss P_ARG2(a, b) +#define CVTPI2PS(a, b) cvtpi2ps P_ARG2(a, b) +#define CVTPS2PI(a, b) cvtps2pi P_ARG2(a, b) +#define CVTSI2SS(a, b) cvtsi2ss P_ARG2(a, b) +#define CVTSS2SI(a, b) cvtss2si P_ARG2(a, b) +#define CVTTPS2PI(a, b) cvttps2pi P_ARG2(a, b) +#define CVTTSS2SI(a, b) cvttss2si P_ARG2(a, b) +#define DIVPS(a, b) divps P_ARG2(a, b) +#define DIVSS(a, b) divss P_ARG2(a, b) +#define FXRSTOR(a) fxrstor P_ARG1(a) +#define FXSAVE(a) fxsave P_ARG1(a) +#define LDMXCSR(a) ldmxcsr P_ARG1(a) +#define MAXPS(a, b) maxps P_ARG2(a, b) +#define MAXSS(a, b) maxss P_ARG2(a, b) +#define MINPS(a, b) minps P_ARG2(a, b) +#define MINSS(a, b) minss P_ARG2(a, b) +#define MOVAPS(a, b) movaps P_ARG2(a, b) +#define MOVHLPS(a, b) movhlps P_ARG2(a, b) +#define MOVHPS(a, b) movhps P_ARG2(a, b) +#define MOVLHPS(a, b) movlhps P_ARG2(a, b) +#define MOVLPS(a, b) movlps P_ARG2(a, b) +#define MOVMSKPS(a, b) movmskps P_ARG2(a, b) +#define MOVNTPS(a, b) movntps P_ARG2(a, b) +#define MOVNTQ(a, b) movntq P_ARG2(a, b) +#define MOVSS(a, b) movss P_ARG2(a, b) +#define MOVUPS(a, b) movups P_ARG2(a, b) +#define MULPS(a, b) mulps P_ARG2(a, b) +#define MULSS(a, b) mulss P_ARG2(a, b) +#define ORPS(a, b) orps P_ARG2(a, b) +#define RCPPS(a, b) rcpps P_ARG2(a, b) +#define RCPSS(a, b) rcpss P_ARG2(a, b) +#define RSQRTPS(a, b) rsqrtps P_ARG2(a, b) +#define RSQRTSS(a, b) rsqrtss P_ARG2(a, b) +#define SHUFPS(a, b, c) shufps P_ARG3(a, b, c) +#define SQRTPS(a, b) sqrtps P_ARG2(a, b) +#define SQRTSS(a, b) sqrtss P_ARG2(a, b) +#define STMXCSR(a) stmxcsr P_ARG1(a) +#define SUBPS(a, b) subps P_ARG2(a, b) +#define UCOMISS(a, b) ucomiss P_ARG2(a, b) +#define UNPCKHPS(a, b) unpckhps P_ARG2(a, b) +#define UNPCKLPS(a, b) unpcklps P_ARG2(a, b) +#define XORPS(a, b) xorps P_ARG2(a, b) + +#define PREFETCHNTA(a) prefetchnta P_ARG1(a) +#define PREFETCHT0(a) prefetcht0 P_ARG1(a) +#define PREFETCHT1(a) prefetcht1 P_ARG1(a) +#define PREFETCHT2(a) prefetcht2 P_ARG1(a) +#define SFENCE sfence + +/* Added by BrianP for FreeBSD (per David Dawes) */ +#if !defined(NASM_ASSEMBLER) && !defined(MASM_ASSEMBLER) && !defined(__bsdi__) +#define LLBL(a) CONCAT(.L,a) +#define LLBL2(a,b) CONCAT3(.L,a,b) +#else +#define LLBL(a) a +#define LLBL2(a,b) CONCAT(a,b) +#endif + +/* Segment overrides */ +#define SEGCS D_BYTE 46 +#define SEGDS D_BYTE 62 +#define SEGES D_BYTE 38 +#define SEGFS D_BYTE 100 +#define SEGGS D_BYTE 101 + +/* Temporary labels: valid until next non-local label */ +#ifdef NASM_ASSEMBLER +#define TLBL(a) CONCAT(.,a) +#else +#define TLBL(a) CONCAT(a,$) +#endif + +#endif /* __ASSYNTAX_H__ */ diff --git a/src/mesa/x86/clip_args.h b/src/mesa/x86/clip_args.h new file mode 100644 index 0000000..cccf801 --- /dev/null +++ b/src/mesa/x86/clip_args.h @@ -0,0 +1,60 @@ +/* $Id: clip_args.h,v 1.5 2002/10/29 20:28:57 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Clip test function interface for assembly code. Simply define + * FRAME_OFFSET to the number of bytes pushed onto the stack before + * using the ARG_* argument macros. + * + * Gareth Hughes + */ + +#ifndef __CLIP_ARGS_H__ +#define __CLIP_ARGS_H__ + +/* + * Offsets for clip_func arguments + * + * typedef GLvector4f *(*clip_func)( GLvector4f *clip_vec, + * GLvector4f *proj_vec, + * GLubyte clipMask[], + * GLubyte *orMask, + * GLubyte *andMask ); + */ + +#define OFFSET_SOURCE 4 +#define OFFSET_DEST 8 +#define OFFSET_CLIP 12 +#define OFFSET_OR 16 +#define OFFSET_AND 20 + +#define ARG_SOURCE REGOFF(FRAME_OFFSET+OFFSET_SOURCE, ESP) +#define ARG_DEST REGOFF(FRAME_OFFSET+OFFSET_DEST, ESP) +#define ARG_CLIP REGOFF(FRAME_OFFSET+OFFSET_CLIP, ESP) +#define ARG_OR REGOFF(FRAME_OFFSET+OFFSET_OR, ESP) +#define ARG_AND REGOFF(FRAME_OFFSET+OFFSET_AND, ESP) + +#endif diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c new file mode 100644 index 0000000..221e5db --- /dev/null +++ b/src/mesa/x86/common_x86.c @@ -0,0 +1,369 @@ +/* + * Mesa 3-D graphics library + * Version: 6.0.1 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Check CPU capabilities & initialize optimized funtions for this particular + * processor. + * + * Written by Holger Waechtler <holger@akaflieg.extern.tu-berlin.de> + * Changed by Andre Werthmann <wertmann@cs.uni-potsdam.de> for using the + * new SSE functions. + */ + +/* XXX these includes should probably go into imports.h or glheader.h */ +#if defined(USE_SSE_ASM) && defined(__linux__) +#include <signal.h> +#endif +#if defined(USE_SSE_ASM) && defined(__FreeBSD__) +#include <sys/types.h> +#include <sys/sysctl.h> +#endif + +#include "common_x86_asm.h" +#include "imports.h" + + +int _mesa_x86_cpu_features = 0; + +/* No reason for this to be public. + */ +extern GLuint _ASMAPI _mesa_x86_has_cpuid(void); +extern void _ASMAPI _mesa_x86_cpuid(GLuint op, GLuint *reg_eax, GLuint *reg_ebx, GLuint *reg_ecx, GLuint *reg_edx); +extern GLuint _ASMAPI _mesa_x86_cpuid_eax(GLuint op); +extern GLuint _ASMAPI _mesa_x86_cpuid_ebx(GLuint op); +extern GLuint _ASMAPI _mesa_x86_cpuid_ecx(GLuint op); +extern GLuint _ASMAPI _mesa_x86_cpuid_edx(GLuint op); + +static void message( const char *msg ) +{ + GLboolean debug; +#ifdef DEBUG + debug = GL_TRUE; +#else + if ( _mesa_getenv( "MESA_DEBUG" ) ) { + debug = GL_TRUE; + } else { + debug = GL_FALSE; + } +#endif + if ( debug ) { + fprintf( stderr, "%s", msg ); + } +} + +#if defined(USE_SSE_ASM) +/* + * We must verify that the Streaming SIMD Extensions are truly supported + * on this processor before we go ahead and hook out the optimized code. + * Unfortunately, the CPUID bit isn't enough, as the OS must set the + * OSFXSR bit in CR4 if it supports the extended FPU save and restore + * required to use SSE. Unfortunately, we can't just go ahead and read + * this register, as only the kernel can do that. Similarly, we must + * verify that the OSXMMEXCPT bit in CR4 has been set by the OS, + * signifying that it supports unmasked SIMD FPU exceptions. If we take + * an unmasked exception and the OS doesn't correctly support them, the + * best we'll get is a SIGILL and the worst we'll get is an infinite + * loop in the signal delivery from the kernel as we can't interact with + * the SIMD FPU state to clear the exception bits. Either way, this is + * not good. + */ + +extern void _mesa_test_os_sse_support( void ); +extern void _mesa_test_os_sse_exception_support( void ); + +#if defined(__linux__) && defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC) +static void sigill_handler( int signal, struct sigcontext sc ) +{ + message( "SIGILL, " ); + + /* Both the "xorps %%xmm0,%%xmm0" and "divps %xmm0,%%xmm1" + * instructions are 3 bytes long. We must increment the instruction + * pointer manually to avoid repeated execution of the offending + * instruction. + * + * If the SIGILL is caused by a divide-by-zero when unmasked + * exceptions aren't supported, the SIMD FPU status and control + * word will be restored at the end of the test, so we don't need + * to worry about doing it here. Besides, we may not be able to... + */ + sc.eip += 3; + + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); +} + +static void sigfpe_handler( int signal, struct sigcontext sc ) +{ + message( "SIGFPE, " ); + + if ( sc.fpstate->magic != 0xffff ) { + /* Our signal context has the extended FPU state, so reset the + * divide-by-zero exception mask and clear the divide-by-zero + * exception bit. + */ + sc.fpstate->mxcsr |= 0x00000200; + sc.fpstate->mxcsr &= 0xfffffffb; + } else { + /* If we ever get here, we're completely hosed. + */ + message( "\n\n" ); + _mesa_problem( NULL, "SSE enabling test failed badly!" ); + } +} +#endif /* __linux__ && _POSIX_SOURCE && X86_FXSR_MAGIC */ + +/* If we're running on a processor that can do SSE, let's see if we + * are allowed to or not. This will catch 2.4.0 or later kernels that + * haven't been configured for a Pentium III but are running on one, + * and RedHat patched 2.2 kernels that have broken exception handling + * support for user space apps that do SSE. + * + * GH: Isn't this just awful? + */ +static void check_os_sse_support( void ) +{ +#if defined(__linux__) +#if defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC) + struct sigaction saved_sigill; + struct sigaction saved_sigfpe; + + /* Save the original signal handlers. + */ + sigaction( SIGILL, NULL, &saved_sigill ); + sigaction( SIGFPE, NULL, &saved_sigfpe ); + + signal( SIGILL, (void (*)(int))sigill_handler ); + signal( SIGFPE, (void (*)(int))sigfpe_handler ); + + /* Emulate test for OSFXSR in CR4. The OS will set this bit if it + * supports the extended FPU save and restore required for SSE. If + * we execute an SSE instruction on a PIII and get a SIGILL, the OS + * doesn't support Streaming SIMD Exceptions, even if the processor + * does. + */ + if ( cpu_has_xmm ) { + message( "Testing OS support for SSE... " ); + + _mesa_test_os_sse_support(); + + if ( cpu_has_xmm ) { + message( "yes.\n" ); + } else { + message( "no!\n" ); + } + } + + /* Emulate test for OSXMMEXCPT in CR4. The OS will set this bit if + * it supports unmasked SIMD FPU exceptions. If we unmask the + * exceptions, do a SIMD divide-by-zero and get a SIGILL, the OS + * doesn't support unmasked SIMD FPU exceptions. If we get a SIGFPE + * as expected, we're okay but we need to clean up after it. + * + * Are we being too stringent in our requirement that the OS support + * unmasked exceptions? Certain RedHat 2.2 kernels enable SSE by + * setting CR4.OSFXSR but don't support unmasked exceptions. Win98 + * doesn't even support them. We at least know the user-space SSE + * support is good in kernels that do support unmasked exceptions, + * and therefore to be safe I'm going to leave this test in here. + */ + if ( cpu_has_xmm ) { + message( "Testing OS support for SSE unmasked exceptions... " ); + + _mesa_test_os_sse_exception_support(); + + if ( cpu_has_xmm ) { + message( "yes.\n" ); + } else { + message( "no!\n" ); + } + } + + /* Restore the original signal handlers. + */ + sigaction( SIGILL, &saved_sigill, NULL ); + sigaction( SIGFPE, &saved_sigfpe, NULL ); + + /* If we've gotten to here and the XMM CPUID bit is still set, we're + * safe to go ahead and hook out the SSE code throughout Mesa. + */ + if ( cpu_has_xmm ) { + message( "Tests of OS support for SSE passed.\n" ); + } else { + message( "Tests of OS support for SSE failed!\n" ); + } +#else + /* We can't use POSIX signal handling to test the availability of + * SSE, so we disable it by default. + */ + message( "Cannot test OS support for SSE, disabling to be safe.\n" ); + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); +#endif /* _POSIX_SOURCE && X86_FXSR_MAGIC */ +#elif defined(__FreeBSD__) + { + int ret, enabled; + unsigned int len; + len = sizeof(enabled); + ret = sysctlbyname("hw.instruction_sse", &enabled, &len, NULL, 0); + if (ret || !enabled) + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); + } +#else + /* Do nothing on other platforms for now. + */ + message( "Not testing OS support for SSE, leaving enabled.\n" ); +#endif /* __linux__ */ +} + +#endif /* USE_SSE_ASM */ + + +void _mesa_init_all_x86_transform_asm( void ) +{ + (void) message; /* silence warning */ +#ifdef USE_X86_ASM + _mesa_x86_cpu_features = 0; + + if (!_mesa_x86_has_cpuid()) { + message("CPUID not detected"); + } + else { + GLuint cpu_features; + GLuint cpu_ext_features; + GLuint cpu_ext_info; + char cpu_vendor[13]; + GLuint result; + + /* get vendor name */ + _mesa_x86_cpuid(0, &result, (GLuint *)(cpu_vendor + 0), (GLuint *)(cpu_vendor + 8), (GLuint *)(cpu_vendor + 4)); + cpu_vendor[12] = '\0'; + + message("cpu vendor: "); + message(cpu_vendor); + message("\n"); + + /* get cpu features */ + cpu_features = _mesa_x86_cpuid_edx(1); + + if (cpu_features & X86_CPU_FPU) + _mesa_x86_cpu_features |= X86_FEATURE_FPU; + if (cpu_features & X86_CPU_CMOV) + _mesa_x86_cpu_features |= X86_FEATURE_CMOV; + +#ifdef USE_MMX_ASM + if (cpu_features & X86_CPU_MMX) + _mesa_x86_cpu_features |= X86_FEATURE_MMX; +#endif + +#ifdef USE_SSE_ASM + if (cpu_features & X86_CPU_XMM) + _mesa_x86_cpu_features |= X86_FEATURE_XMM; + if (cpu_features & X86_CPU_XMM2) + _mesa_x86_cpu_features |= X86_FEATURE_XMM2; +#endif + + /* query extended cpu features */ + if ((cpu_ext_info = _mesa_x86_cpuid_eax(0x80000000)) > 0x80000000) { + if (cpu_ext_info >= 0x80000001) { + + cpu_ext_features = _mesa_x86_cpuid_edx(0x80000001); + + if (cpu_features & X86_CPU_MMX) { + +#ifdef USE_3DNOW_ASM + if (cpu_ext_features & X86_CPUEXT_3DNOW) + _mesa_x86_cpu_features |= X86_FEATURE_3DNOW; + if (cpu_ext_features & X86_CPUEXT_3DNOW_EXT) + _mesa_x86_cpu_features |= X86_FEATURE_3DNOWEXT; +#endif + +#ifdef USE_MMX_ASM + if (cpu_ext_features & X86_CPUEXT_MMX_EXT) + _mesa_x86_cpu_features |= X86_FEATURE_MMXEXT; +#endif + } + } + + /* query cpu name */ + if (cpu_ext_info >= 0x80000002) { + GLuint ofs; + char cpu_name[49]; + for (ofs = 0; ofs < 3; ofs++) + _mesa_x86_cpuid(0x80000002+ofs, (GLuint *)(cpu_name + (16*ofs)+0), (GLuint *)(cpu_name + (16*ofs)+4), (GLuint *)(cpu_name + (16*ofs)+8), (GLuint *)(cpu_name + (16*ofs)+12)); + cpu_name[48] = '\0'; /* the name should be NULL terminated, but just to be sure */ + + message("cpu name: "); + message(cpu_name); + message("\n"); + } + } + + } + + if ( _mesa_getenv( "MESA_NO_ASM" ) ) { + _mesa_x86_cpu_features = 0; + } + + if ( _mesa_x86_cpu_features ) { + _mesa_init_x86_transform_asm(); + } + +#ifdef USE_MMX_ASM + if ( cpu_has_mmx ) { + if ( _mesa_getenv( "MESA_NO_MMX" ) == 0 ) { + message( "MMX cpu detected.\n" ); + } else { + _mesa_x86_cpu_features &= ~(X86_FEATURE_MMX); + } + } +#endif + +#ifdef USE_3DNOW_ASM + if ( cpu_has_3dnow ) { + if ( _mesa_getenv( "MESA_NO_3DNOW" ) == 0 ) { + message( "3DNow! cpu detected.\n" ); + _mesa_init_3dnow_transform_asm(); + } else { + _mesa_x86_cpu_features &= ~(X86_FEATURE_3DNOW); + } + } +#endif + +#ifdef USE_SSE_ASM + if ( cpu_has_xmm ) { + if ( _mesa_getenv( "MESA_NO_SSE" ) == 0 ) { + message( "SSE cpu detected.\n" ); + if ( _mesa_getenv( "MESA_FORCE_SSE" ) == 0 ) { + check_os_sse_support(); + } + if ( cpu_has_xmm ) { + _mesa_init_sse_transform_asm(); + } + } else { + message( "SSE cpu detected, but switched off by user.\n" ); + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); + } + } +#endif +#endif +} + diff --git a/src/mesa/x86/common_x86_asm.S b/src/mesa/x86/common_x86_asm.S new file mode 100644 index 0000000..9deceea --- /dev/null +++ b/src/mesa/x86/common_x86_asm.S @@ -0,0 +1,203 @@ +/* $Id: common_x86_asm.S,v 1.17 2003/01/21 16:13:58 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2003 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Check extended CPU capabilities. Now justs returns the raw CPUID + * feature information, allowing the higher level code to interpret the + * results. + * + * Written by Holger Waechtler <holger@akaflieg.extern.tu-berlin.de> + * + * Cleaned up and simplified by Gareth Hughes <gareth@valinux.com> + * + */ + +/* + * NOTE: Avoid using spaces in between '(' ')' and arguments, especially + * with macros like CONST, LLBL that expand to CONCAT(...). Putting spaces + * in there will break the build on some platforms. + */ + +#include "matypes.h" +#include "common_x86_features.h" + + SEG_TEXT + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_has_cpuid) +GLNAME(_mesa_x86_has_cpuid): + + /* Test for the CPUID command. If the ID Flag bit in EFLAGS + * (bit 21) is writable, the CPUID command is present */ + PUSHF_L + POP_L (EAX) + MOV_L (EAX, ECX) + XOR_L (CONST(0x00200000), EAX) + PUSH_L (EAX) + POPF_L + PUSHF_L + POP_L (EAX) + + /* Verify the ID Flag bit has been written. */ + CMP_L (ECX, EAX) + SETNE (AL) + XOR_L (CONST(0xff), EAX) + + RET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_cpuid) +GLNAME(_mesa_x86_cpuid): + + MOV_L (REGOFF(4, ESP), EAX) /* cpuid op */ + PUSH_L (EDI) + PUSH_L (EBX) + + CPUID + + MOV_L (REGOFF(16, ESP), EDI) /* *eax */ + MOV_L (EAX, REGIND(EDI)) + MOV_L (REGOFF(20, ESP), EDI) /* *ebx */ + MOV_L (EBX, REGIND(EDI)) + MOV_L (REGOFF(24, ESP), EDI) /* *ecx */ + MOV_L (ECX, REGIND(EDI)) + MOV_L (REGOFF(28, ESP), EDI) /* *edx */ + MOV_L (EDX, REGIND(EDI)) + + POP_L (EBX) + POP_L (EDI) + RET + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_cpuid_eax) +GLNAME(_mesa_x86_cpuid_eax): + + MOV_L (REGOFF(4, ESP), EAX) /* cpuid op */ + PUSH_L (EBX) + + CPUID + + POP_L (EBX) + RET + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_cpuid_ebx) +GLNAME(_mesa_x86_cpuid_ebx): + + MOV_L (REGOFF(4, ESP), EAX) /* cpuid op */ + PUSH_L (EBX) + + CPUID + MOV_L (EBX, EAX) /* return EBX */ + + POP_L (EBX) + RET + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_cpuid_ecx) +GLNAME(_mesa_x86_cpuid_ecx): + + MOV_L (REGOFF(4, ESP), EAX) /* cpuid op */ + PUSH_L (EBX) + + CPUID + MOV_L (ECX, EAX) /* return ECX */ + + POP_L (EBX) + RET + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_x86_cpuid_edx) +GLNAME(_mesa_x86_cpuid_edx): + + MOV_L (REGOFF(4, ESP), EAX) /* cpuid op */ + PUSH_L (EBX) + + CPUID + MOV_L (EDX, EAX) /* return EDX */ + + POP_L (EBX) + RET + +#ifdef USE_SSE_ASM +/* Execute an SSE instruction to see if the operating system correctly + * supports SSE. A signal handler for SIGILL should have been set + * before calling this function, otherwise this could kill the client + * application. + */ +ALIGNTEXT4 +GLOBL GLNAME( _mesa_test_os_sse_support ) +GLNAME( _mesa_test_os_sse_support ): + + XORPS ( XMM0, XMM0 ) + + RET + + +/* Perform an SSE divide-by-zero to see if the operating system + * correctly supports unmasked SIMD FPU exceptions. Signal handlers for + * SIGILL and SIGFPE should have been set before calling this function, + * otherwise this could kill the client application. + */ +ALIGNTEXT4 +GLOBL GLNAME( _mesa_test_os_sse_exception_support ) +GLNAME( _mesa_test_os_sse_exception_support ): + + PUSH_L ( EBP ) + MOV_L ( ESP, EBP ) + SUB_L ( CONST( 8 ), ESP ) + + /* Save the original MXCSR register value. + */ + STMXCSR ( REGOFF( -4, EBP ) ) + + /* Unmask the divide-by-zero exception and perform one. + */ + STMXCSR ( REGOFF( -8, EBP ) ) + AND_L ( CONST( 0xfffffdff ), REGOFF( -8, EBP ) ) + LDMXCSR ( REGOFF( -8, EBP ) ) + + XORPS ( XMM0, XMM0 ) + + PUSH_L ( CONST( 0x3f800000 ) ) + PUSH_L ( CONST( 0x3f800000 ) ) + PUSH_L ( CONST( 0x3f800000 ) ) + PUSH_L ( CONST( 0x3f800000 ) ) + + MOVUPS ( REGIND( ESP ), XMM1 ) + + DIVPS ( XMM0, XMM1 ) + + /* Restore the original MXCSR register value. + */ + LDMXCSR ( REGOFF( -4, EBP ) ) + + LEAVE + RET + +#endif + diff --git a/src/mesa/x86/common_x86_asm.h b/src/mesa/x86/common_x86_asm.h new file mode 100644 index 0000000..a913bb1 --- /dev/null +++ b/src/mesa/x86/common_x86_asm.h @@ -0,0 +1,64 @@ +/* $Id: common_x86_asm.h,v 1.11 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Check CPU capabilities & initialize optimized funtions for this particular + * processor. + * + * Written by Holger Waechtler <holger@akaflieg.extern.tu-berlin.de> + * Changed by Andre Werthmann <wertmann@cs.uni-potsdam.de> for using the + * new SSE functions + * + * Reimplemented by Gareth Hughes in a more + * future-proof manner, based on code in the Linux kernel. + */ + +#ifndef __COMMON_X86_ASM_H__ +#define __COMMON_X86_ASM_H__ + +/* Do not reference mtypes.h from this file. + */ +#include "common_x86_features.h" + +#ifdef HAVE_CONFIG_H +#include "conf.h" +#endif + +#ifdef USE_X86_ASM +#include "x86.h" +#ifdef USE_3DNOW_ASM +#include "3dnow.h" +#endif +#ifdef USE_SSE_ASM +#include "sse.h" +#endif +#endif + +extern int _mesa_x86_cpu_features; + +extern void _mesa_init_all_x86_transform_asm( void ); + +#endif diff --git a/src/mesa/x86/common_x86_features.h b/src/mesa/x86/common_x86_features.h new file mode 100644 index 0000000..9050977 --- /dev/null +++ b/src/mesa/x86/common_x86_features.h @@ -0,0 +1,68 @@ +/* $Id: common_x86_features.h,v 1.6 2003/01/21 16:14:00 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * x86 CPUID feature information. The raw data is returned by + * _mesa_identify_x86_cpu_features() and interpreted with the cpu_has_* + * helper macros. + * + * Gareth Hughes + */ + +#ifndef __COMMON_X86_FEATURES_H__ +#define __COMMON_X86_FEATURES_H__ + +#define X86_FEATURE_FPU (1<<0) +#define X86_FEATURE_CMOV (1<<1) +#define X86_FEATURE_MMXEXT (1<<2) +#define X86_FEATURE_MMX (1<<3) +#define X86_FEATURE_FXSR (1<<4) +#define X86_FEATURE_XMM (1<<5) +#define X86_FEATURE_XMM2 (1<<6) +#define X86_FEATURE_3DNOWEXT (1<<7) +#define X86_FEATURE_3DNOW (1<<8) + +/* standard X86 CPU features */ +#define X86_CPU_FPU (1<<0) +#define X86_CPU_CMOV (1<<15) +#define X86_CPU_MMX (1<<23) +#define X86_CPU_XMM (1<<25) +#define X86_CPU_XMM2 (1<<26) + +/* extended X86 CPU features */ +#define X86_CPUEXT_MMX_EXT (1<<22) +#define X86_CPUEXT_3DNOW_EXT (1<<30) +#define X86_CPUEXT_3DNOW (1<<31) + +#define cpu_has_mmx (_mesa_x86_cpu_features & X86_FEATURE_MMX) +#define cpu_has_mmxext (_mesa_x86_cpu_features & X86_FEATURE_MMXEXT) +#define cpu_has_xmm (_mesa_x86_cpu_features & X86_FEATURE_XMM) +#define cpu_has_xmm2 (_mesa_x86_cpu_features & X86_FEATURE_XMM2) +#define cpu_has_3dnow (_mesa_x86_cpu_features & X86_FEATURE_3DNOW) +#define cpu_has_3dnowext (_mesa_x86_cpu_features & X86_FEATURE_3DNOWEXT) + +#endif + diff --git a/src/mesa/x86/common_x86_macros.h b/src/mesa/x86/common_x86_macros.h new file mode 100644 index 0000000..ba155ca --- /dev/null +++ b/src/mesa/x86/common_x86_macros.h @@ -0,0 +1,107 @@ +/* $Id: common_x86_macros.h,v 1.3 2002/10/29 20:28:58 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Gareth Hughes + */ + +#ifndef __COMMON_X86_MACROS_H__ +#define __COMMON_X86_MACROS_H__ + + +/* ============================================================= + * Transformation function declarations: + */ + +#define XFORM_ARGS GLvector4f *to_vec, \ + const GLfloat m[16], \ + const GLvector4f *from_vec + +#define DECLARE_XFORM_GROUP( pfx, sz ) \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_general( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_identity( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d_no_rot( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_perspective( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_2d_no_rot( XFORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_points##sz##_3d( XFORM_ARGS ); + +#define ASSIGN_XFORM_GROUP( pfx, sz ) \ + _mesa_transform_tab[sz][MATRIX_GENERAL] = \ + _mesa_##pfx##_transform_points##sz##_general; \ + _mesa_transform_tab[sz][MATRIX_IDENTITY] = \ + _mesa_##pfx##_transform_points##sz##_identity; \ + _mesa_transform_tab[sz][MATRIX_3D_NO_ROT] = \ + _mesa_##pfx##_transform_points##sz##_3d_no_rot; \ + _mesa_transform_tab[sz][MATRIX_PERSPECTIVE] = \ + _mesa_##pfx##_transform_points##sz##_perspective; \ + _mesa_transform_tab[sz][MATRIX_2D] = \ + _mesa_##pfx##_transform_points##sz##_2d; \ + _mesa_transform_tab[sz][MATRIX_2D_NO_ROT] = \ + _mesa_##pfx##_transform_points##sz##_2d_no_rot; \ + _mesa_transform_tab[sz][MATRIX_3D] = \ + _mesa_##pfx##_transform_points##sz##_3d; + + +/* ============================================================= + * Normal transformation function declarations: + */ + +#define NORM_ARGS const GLmatrix *mat, \ + GLfloat scale, \ + const GLvector4f *in, \ + const GLfloat *lengths, \ + GLvector4f *dest + +#define DECLARE_NORM_GROUP( pfx ) \ +extern void _ASMAPI _mesa_##pfx##_rescale_normals( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_normalize_normals( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_normals( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_normals_no_rot( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_rescale_normals_no_rot( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals( NORM_ARGS ); \ +extern void _ASMAPI _mesa_##pfx##_transform_normalize_normals_no_rot( NORM_ARGS ); + +#define ASSIGN_NORM_GROUP( pfx ) \ + _mesa_normal_tab[NORM_RESCALE] = \ + _mesa_##pfx##_rescale_normals; \ + _mesa_normal_tab[NORM_NORMALIZE] = \ + _mesa_##pfx##_normalize_normals; \ + _mesa_normal_tab[NORM_TRANSFORM] = \ + _mesa_##pfx##_transform_normals; \ + _mesa_normal_tab[NORM_TRANSFORM_NO_ROT] = \ + _mesa_##pfx##_transform_normals_no_rot; \ + _mesa_normal_tab[NORM_TRANSFORM | NORM_RESCALE] = \ + _mesa_##pfx##_transform_rescale_normals; \ + _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_RESCALE] = \ + _mesa_##pfx##_transform_rescale_normals_no_rot; \ + _mesa_normal_tab[NORM_TRANSFORM | NORM_NORMALIZE] = \ + _mesa_##pfx##_transform_normalize_normals; \ + _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE] = \ + _mesa_##pfx##_transform_normalize_normals_no_rot; + + +#endif diff --git a/src/mesa/x86/gen_matypes.c b/src/mesa/x86/gen_matypes.c new file mode 100644 index 0000000..30642e4 --- /dev/null +++ b/src/mesa/x86/gen_matypes.c @@ -0,0 +1,251 @@ +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2003 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Gareth Hughes + */ + +/* + * This generates an asm version of mtypes.h (called matypes.h), so that + * Mesa's x86 assembly code can access the internal structures easily. + * This will be particularly useful when developing new x86 asm code for + * Mesa, including lighting, clipping, texture image conversion etc. + */ + +#include "main/glheader.h" +#include "main/mtypes.h" +#include "tnl/t_context.h" + + +#undef offsetof +#define offsetof( type, member ) ((size_t) &((type *)0)->member) + + +#define OFFSET_HEADER( x ) \ +do { \ + printf( "\n" ); \ + printf( "\n" ); \ + printf( "/* =====================================================" \ + "========\n" ); \ + printf( " * Offsets for %s\n", x ); \ + printf( " */\n" ); \ + printf( "\n" ); \ +} while (0) + +#define DEFINE_HEADER( x ) \ +do { \ + printf( "\n" ); \ + printf( "/*\n" ); \ + printf( " * Flags for %s\n", x ); \ + printf( " */\n" ); \ + printf( "\n" ); \ +} while (0) + +#if defined(__BEOS__) +#define OFFSET( s, t, m ) \ + printf( "#define %s\t%ld\n", s, offsetof( t, m ) ); +#else +#define OFFSET( s, t, m ) \ + printf( "#define %s\t%d\n", s, offsetof( t, m ) ); +#endif + +#if defined(__BEOS__) +#define SIZEOF( s, t ) \ + printf( "#define %s\t%ld\n", s, sizeof(t) ); +#else +#define SIZEOF( s, t ) \ + printf( "#define %s\t%d\n", s, sizeof(t) ); +#endif + +#define DEFINE( s, d ) \ + printf( "#define %s\t0x%x\n", s, d ); + + + +int main( int argc, char **argv ) +{ + printf( "/*\n" ); + printf( " * This file is automatically generated from the Mesa internal type\n" ); + printf( " * definitions. Do not edit directly.\n" ); + printf( " */\n" ); + printf( "\n" ); + printf( "#ifndef __ASM_TYPES_H__\n" ); + printf( "#define __ASM_TYPES_H__\n" ); + printf( "\n" ); + printf( "#include \"assyntax.h\"\n" ); + + + /* GLcontext offsets: + */ + OFFSET_HEADER( "GLcontext" ); + + OFFSET( "CTX_DRIVER_CTX ", GLcontext, DriverCtx ); + printf( "\n" ); + OFFSET( "CTX_LIGHT_ENABLED ", GLcontext, Light.Enabled ); + OFFSET( "CTX_LIGHT_SHADE_MODEL ", GLcontext, Light.ShadeModel ); + OFFSET( "CTX_LIGHT_COLOR_MAT_FACE ", GLcontext, Light.ColorMaterialFace ); + OFFSET( "CTX_LIGHT_COLOR_MAT_MODE ", GLcontext, Light.ColorMaterialMode ); + OFFSET( "CTX_LIGHT_COLOR_MAT_MASK ", GLcontext, Light.ColorMaterialBitmask ); + OFFSET( "CTX_LIGHT_COLOR_MAT_ENABLED ", GLcontext, Light.ColorMaterialEnabled ); + OFFSET( "CTX_LIGHT_ENABLED_LIST ", GLcontext, Light.EnabledList ); + OFFSET( "CTX_LIGHT_NEED_VERTS ", GLcontext, Light._NeedVertices ); + OFFSET( "CTX_LIGHT_FLAGS ", GLcontext, Light._Flags ); + OFFSET( "CTX_LIGHT_BASE_COLOR ", GLcontext, Light._BaseColor ); + + + /* struct vertex_buffer offsets: + */ + OFFSET_HEADER( "struct vertex_buffer" ); + + OFFSET( "VB_SIZE ", struct vertex_buffer, Size ); + OFFSET( "VB_COUNT ", struct vertex_buffer, Count ); + printf( "\n" ); + OFFSET( "VB_ELTS ", struct vertex_buffer, Elts ); + OFFSET( "VB_OBJ_PTR ", struct vertex_buffer, ObjPtr ); + OFFSET( "VB_EYE_PTR ", struct vertex_buffer, EyePtr ); + OFFSET( "VB_CLIP_PTR ", struct vertex_buffer, ClipPtr ); + OFFSET( "VB_PROJ_CLIP_PTR ", struct vertex_buffer, NdcPtr ); + OFFSET( "VB_CLIP_OR_MASK ", struct vertex_buffer, ClipOrMask ); + OFFSET( "VB_CLIP_MASK ", struct vertex_buffer, ClipMask ); + OFFSET( "VB_NORMAL_PTR ", struct vertex_buffer, NormalPtr ); + OFFSET( "VB_EDGE_FLAG ", struct vertex_buffer, EdgeFlag ); + OFFSET( "VB_TEX0_COORD_PTR ", struct vertex_buffer, TexCoordPtr[0] ); + OFFSET( "VB_TEX1_COORD_PTR ", struct vertex_buffer, TexCoordPtr[1] ); + OFFSET( "VB_TEX2_COORD_PTR ", struct vertex_buffer, TexCoordPtr[2] ); + OFFSET( "VB_TEX3_COORD_PTR ", struct vertex_buffer, TexCoordPtr[3] ); + OFFSET( "VB_INDEX_PTR ", struct vertex_buffer, IndexPtr ); + OFFSET( "VB_COLOR_PTR ", struct vertex_buffer, ColorPtr ); + OFFSET( "VB_SECONDARY_COLOR_PTR ", struct vertex_buffer, SecondaryColorPtr ); + OFFSET( "VB_FOG_COORD_PTR ", struct vertex_buffer, FogCoordPtr ); + OFFSET( "VB_POINT_SIZE_PTR ", struct vertex_buffer, PointSizePtr ); + OFFSET( "VB_PRIMITIVE ", struct vertex_buffer, Primitive ); + printf( "\n" ); + OFFSET( "VB_LAST_CLIPPED ", struct vertex_buffer, LastClipped ); + + DEFINE_HEADER( "struct vertex_buffer" ); + + /* XXX use new labels here someday after vertex proram is done */ + DEFINE( "VERT_BIT_OBJ ", VERT_BIT_POS ); + DEFINE( "VERT_BIT_NORM ", VERT_BIT_NORMAL ); + DEFINE( "VERT_BIT_RGBA ", VERT_BIT_COLOR0 ); + DEFINE( "VERT_BIT_SPEC_RGB ", VERT_BIT_COLOR1 ); + DEFINE( "VERT_BIT_FOG_COORD ", VERT_BIT_FOG ); + DEFINE( "VERT_BIT_TEX0 ", VERT_BIT_TEX0 ); + DEFINE( "VERT_BIT_TEX1 ", VERT_BIT_TEX1 ); + DEFINE( "VERT_BIT_TEX2 ", VERT_BIT_TEX2 ); + DEFINE( "VERT_BIT_TEX3 ", VERT_BIT_TEX3 ); + + + /* GLvector4f offsets: + */ + OFFSET_HEADER( "GLvector4f" ); + + OFFSET( "V4F_DATA ", GLvector4f, data ); + OFFSET( "V4F_START ", GLvector4f, start ); + OFFSET( "V4F_COUNT ", GLvector4f, count ); + OFFSET( "V4F_STRIDE ", GLvector4f, stride ); + OFFSET( "V4F_SIZE ", GLvector4f, size ); + OFFSET( "V4F_FLAGS ", GLvector4f, flags ); + + DEFINE_HEADER( "GLvector4f" ); + + DEFINE( "VEC_MALLOC ", VEC_MALLOC ); + DEFINE( "VEC_NOT_WRITEABLE ", VEC_NOT_WRITEABLE ); + DEFINE( "VEC_BAD_STRIDE ", VEC_BAD_STRIDE ); + printf( "\n" ); + DEFINE( "VEC_SIZE_1 ", VEC_SIZE_1 ); + DEFINE( "VEC_SIZE_2 ", VEC_SIZE_2 ); + DEFINE( "VEC_SIZE_3 ", VEC_SIZE_3 ); + DEFINE( "VEC_SIZE_4 ", VEC_SIZE_4 ); + + + /* GLmatrix offsets: + */ + OFFSET_HEADER( "GLmatrix" ); + + OFFSET( "MATRIX_DATA ", GLmatrix, m ); + OFFSET( "MATRIX_INV ", GLmatrix, inv ); + OFFSET( "MATRIX_FLAGS ", GLmatrix, flags ); + OFFSET( "MATRIX_TYPE ", GLmatrix, type ); + + + /* struct gl_light offsets: + */ + OFFSET_HEADER( "struct gl_light" ); + + OFFSET( "LIGHT_NEXT ", struct gl_light, next ); + OFFSET( "LIGHT_PREV ", struct gl_light, prev ); + printf( "\n" ); + OFFSET( "LIGHT_AMBIENT ", struct gl_light, Ambient ); + OFFSET( "LIGHT_DIFFUSE ", struct gl_light, Diffuse ); + OFFSET( "LIGHT_SPECULAR ", struct gl_light, Specular ); + OFFSET( "LIGHT_EYE_POSITION ", struct gl_light, EyePosition ); + OFFSET( "LIGHT_EYE_DIRECTION ", struct gl_light, EyeDirection ); + OFFSET( "LIGHT_SPOT_EXPONENT ", struct gl_light, SpotExponent ); + OFFSET( "LIGHT_SPOT_CUTOFF ", struct gl_light, SpotCutoff ); + OFFSET( "LIGHT_COS_CUTOFF ", struct gl_light, _CosCutoff ); + OFFSET( "LIGHT_CONST_ATTEN ", struct gl_light, ConstantAttenuation ); + OFFSET( "LIGHT_LINEAR_ATTEN ", struct gl_light, LinearAttenuation ); + OFFSET( "LIGHT_QUADRATIC_ATTEN ", struct gl_light, QuadraticAttenuation ); + OFFSET( "LIGHT_ENABLED ", struct gl_light, Enabled ); + printf( "\n" ); + OFFSET( "LIGHT_FLAGS ", struct gl_light, _Flags ); + printf( "\n" ); + OFFSET( "LIGHT_POSITION ", struct gl_light, _Position ); + OFFSET( "LIGHT_VP_INF_NORM ", struct gl_light, _VP_inf_norm ); + OFFSET( "LIGHT_H_INF_NORM ", struct gl_light, _h_inf_norm ); + OFFSET( "LIGHT_NORM_DIRECTION ", struct gl_light, _NormDirection ); + OFFSET( "LIGHT_VP_INF_SPOT_ATTEN ", struct gl_light, _VP_inf_spot_attenuation ); + printf( "\n" ); + OFFSET( "LIGHT_SPOT_EXP_TABLE ", struct gl_light, _SpotExpTable ); + OFFSET( "LIGHT_MAT_AMBIENT ", struct gl_light, _MatAmbient ); + OFFSET( "LIGHT_MAT_DIFFUSE ", struct gl_light, _MatDiffuse ); + OFFSET( "LIGHT_MAT_SPECULAR ", struct gl_light, _MatSpecular ); + printf( "\n" ); + SIZEOF( "SIZEOF_GL_LIGHT ", struct gl_light ); + + DEFINE_HEADER( "struct gl_light" ); + + DEFINE( "LIGHT_SPOT ", LIGHT_SPOT ); + DEFINE( "LIGHT_LOCAL_VIEWER ", LIGHT_LOCAL_VIEWER ); + DEFINE( "LIGHT_POSITIONAL ", LIGHT_POSITIONAL ); + printf( "\n" ); + DEFINE( "LIGHT_NEED_VERTICES ", LIGHT_NEED_VERTICES ); + + + /* struct gl_lightmodel offsets: + */ + OFFSET_HEADER( "struct gl_lightmodel" ); + + OFFSET( "LIGHT_MODEL_AMBIENT ", struct gl_lightmodel, Ambient ); + OFFSET( "LIGHT_MODEL_LOCAL_VIEWER ", struct gl_lightmodel, LocalViewer ); + OFFSET( "LIGHT_MODEL_TWO_SIDE ", struct gl_lightmodel, TwoSide ); + OFFSET( "LIGHT_MODEL_COLOR_CONTROL ", struct gl_lightmodel, ColorControl ); + + + printf( "\n" ); + printf( "\n" ); + printf( "#endif /* __ASM_TYPES_H__ */\n" ); + + return 0; +} diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S new file mode 100644 index 0000000..0d4d19d --- /dev/null +++ b/src/mesa/x86/glapi_x86.S @@ -0,0 +1,1030 @@ +/* DO NOT EDIT - This file generated automatically by gl_x86_asm.py (from Mesa) script */ + +/* + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * (C) Copyright IBM Corporation 2004 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL, IBM, + * AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "assyntax.h" +#include "glapioffsets.h" + +#ifndef __WIN32__ + +#if defined(STDCALL_API) +# if defined(USE_MGL_NAMESPACE) +# define GL_PREFIX(n,n2) GLNAME(CONCAT(mgl,n2)) +# else +# define GL_PREFIX(n,n2) GLNAME(CONCAT(gl,n2)) +# endif +#else +# if defined(USE_MGL_NAMESPACE) +# define GL_PREFIX(n,n2) GLNAME(CONCAT(mgl,n)) +# else +# define GL_PREFIX(n,n2) GLNAME(CONCAT(gl,n)) +# endif +#endif + +#define GL_OFFSET(x) CODEPTR(REGOFF(4 * x, EAX)) + +#if defined(GNU_ASSEMBLER) && !defined(__DJGPP__) && !defined(__MINGW32__) +#define GLOBL_FN(x) GLOBL x ; .type x, function +#else +#define GLOBL_FN(x) GLOBL x +#endif + +#if defined(PTHREADS) || defined(XTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS) +# define THREADS +#endif + +#if defined(PTHREADS) +# define GL_STUB(fn,off,fn_alt) \ +ALIGNTEXT16; \ +GLOBL_FN(GL_PREFIX(fn, fn_alt)); \ +GL_PREFIX(fn, fn_alt): \ + MOV_L(CONTENT(GLNAME(_glapi_DispatchTSD)), EAX) ; \ + TEST_L(EAX, EAX) ; \ + JE(1f) ; \ + JMP(GL_OFFSET(off)) ; \ +1: CALL(_x86_get_dispatch) ; \ + JMP(GL_OFFSET(off)) +#elif defined(THREADS) +# define GL_STUB(fn,off,fn_alt) \ +ALIGNTEXT16; \ +GLOBL_FN(GL_PREFIX(fn, fn_alt)); \ +GL_PREFIX(fn, fn_alt): \ + MOV_L(CONTENT(GLNAME(_glapi_DispatchTSD)), EAX) ; \ + TEST_L(EAX, EAX) ; \ + JE(1f) ; \ + JMP(GL_OFFSET(off)) ; \ +1: CALL(_glapi_get_dispatch) ; \ + JMP(GL_OFFSET(off)) +#else /* Non-threaded version. */ +# define GL_STUB(fn,off,fn_alt) \ +ALIGNTEXT16; \ +GLOBL_FN(GL_PREFIX(fn, fn_alt)); \ +GL_PREFIX(fn, fn_alt): \ + MOV_L(CONTENT(GLNAME(_glapi_Dispatch)), EAX) ; \ + JMP(GL_OFFSET(off)) +#endif + +SEG_TEXT + +#ifdef PTHREADS +EXTERN GLNAME(_glapi_Dispatch) +EXTERN GLNAME(_gl_DispatchTSD) +EXTERN GLNAME(pthread_getspecific) + +ALIGNTEXT16 +GLNAME(_x86_get_dispatch): + SUB_L(CONST(24), ESP) + PUSH_L(GLNAME(_gl_DispatchTSD)) + CALL(GLNAME(pthread_getspecific)) + ADD_L(CONST(28), ESP) + RET +#elif defined(THREADS) +EXTERN GLNAME(_glapi_get_dispatch) +#endif + + ALIGNTEXT16 ; GLOBL GLNAME(gl_dispatch_functions_start) +GLNAME(gl_dispatch_functions_start): + + GL_STUB(NewList, _gloffset_NewList, NewList@8) + GL_STUB(EndList, _gloffset_EndList, EndList@0) + GL_STUB(CallList, _gloffset_CallList, CallList@4) + GL_STUB(CallLists, _gloffset_CallLists, CallLists@12) + GL_STUB(DeleteLists, _gloffset_DeleteLists, DeleteLists@8) + GL_STUB(GenLists, _gloffset_GenLists, GenLists@4) + GL_STUB(ListBase, _gloffset_ListBase, ListBase@4) + GL_STUB(Begin, _gloffset_Begin, Begin@4) + GL_STUB(Bitmap, _gloffset_Bitmap, Bitmap@28) + GL_STUB(Color3b, _gloffset_Color3b, Color3b@12) + GL_STUB(Color3bv, _gloffset_Color3bv, Color3bv@4) + GL_STUB(Color3d, _gloffset_Color3d, Color3d@24) + GL_STUB(Color3dv, _gloffset_Color3dv, Color3dv@4) + GL_STUB(Color3f, _gloffset_Color3f, Color3f@12) + GL_STUB(Color3fv, _gloffset_Color3fv, Color3fv@4) + GL_STUB(Color3i, _gloffset_Color3i, Color3i@12) + GL_STUB(Color3iv, _gloffset_Color3iv, Color3iv@4) + GL_STUB(Color3s, _gloffset_Color3s, Color3s@12) + GL_STUB(Color3sv, _gloffset_Color3sv, Color3sv@4) + GL_STUB(Color3ub, _gloffset_Color3ub, Color3ub@12) + GL_STUB(Color3ubv, _gloffset_Color3ubv, Color3ubv@4) + GL_STUB(Color3ui, _gloffset_Color3ui, Color3ui@12) + GL_STUB(Color3uiv, _gloffset_Color3uiv, Color3uiv@4) + GL_STUB(Color3us, _gloffset_Color3us, Color3us@12) + GL_STUB(Color3usv, _gloffset_Color3usv, Color3usv@4) + GL_STUB(Color4b, _gloffset_Color4b, Color4b@16) + GL_STUB(Color4bv, _gloffset_Color4bv, Color4bv@4) + GL_STUB(Color4d, _gloffset_Color4d, Color4d@32) + GL_STUB(Color4dv, _gloffset_Color4dv, Color4dv@4) + GL_STUB(Color4f, _gloffset_Color4f, Color4f@16) + GL_STUB(Color4fv, _gloffset_Color4fv, Color4fv@4) + GL_STUB(Color4i, _gloffset_Color4i, Color4i@16) + GL_STUB(Color4iv, _gloffset_Color4iv, Color4iv@4) + GL_STUB(Color4s, _gloffset_Color4s, Color4s@16) + GL_STUB(Color4sv, _gloffset_Color4sv, Color4sv@4) + GL_STUB(Color4ub, _gloffset_Color4ub, Color4ub@16) + GL_STUB(Color4ubv, _gloffset_Color4ubv, Color4ubv@4) + GL_STUB(Color4ui, _gloffset_Color4ui, Color4ui@16) + GL_STUB(Color4uiv, _gloffset_Color4uiv, Color4uiv@4) + GL_STUB(Color4us, _gloffset_Color4us, Color4us@16) + GL_STUB(Color4usv, _gloffset_Color4usv, Color4usv@4) + GL_STUB(EdgeFlag, _gloffset_EdgeFlag, EdgeFlag@4) + GL_STUB(EdgeFlagv, _gloffset_EdgeFlagv, EdgeFlagv@4) + GL_STUB(End, _gloffset_End, End@0) + GL_STUB(Indexd, _gloffset_Indexd, Indexd@8) + GL_STUB(Indexdv, _gloffset_Indexdv, Indexdv@4) + GL_STUB(Indexf, _gloffset_Indexf, Indexf@4) + GL_STUB(Indexfv, _gloffset_Indexfv, Indexfv@4) + GL_STUB(Indexi, _gloffset_Indexi, Indexi@4) + GL_STUB(Indexiv, _gloffset_Indexiv, Indexiv@4) + GL_STUB(Indexs, _gloffset_Indexs, Indexs@4) + GL_STUB(Indexsv, _gloffset_Indexsv, Indexsv@4) + GL_STUB(Normal3b, _gloffset_Normal3b, Normal3b@12) + GL_STUB(Normal3bv, _gloffset_Normal3bv, Normal3bv@4) + GL_STUB(Normal3d, _gloffset_Normal3d, Normal3d@24) + GL_STUB(Normal3dv, _gloffset_Normal3dv, Normal3dv@4) + GL_STUB(Normal3f, _gloffset_Normal3f, Normal3f@12) + GL_STUB(Normal3fv, _gloffset_Normal3fv, Normal3fv@4) + GL_STUB(Normal3i, _gloffset_Normal3i, Normal3i@12) + GL_STUB(Normal3iv, _gloffset_Normal3iv, Normal3iv@4) + GL_STUB(Normal3s, _gloffset_Normal3s, Normal3s@12) + GL_STUB(Normal3sv, _gloffset_Normal3sv, Normal3sv@4) + GL_STUB(RasterPos2d, _gloffset_RasterPos2d, RasterPos2d@16) + GL_STUB(RasterPos2dv, _gloffset_RasterPos2dv, RasterPos2dv@4) + GL_STUB(RasterPos2f, _gloffset_RasterPos2f, RasterPos2f@8) + GL_STUB(RasterPos2fv, _gloffset_RasterPos2fv, RasterPos2fv@4) + GL_STUB(RasterPos2i, _gloffset_RasterPos2i, RasterPos2i@8) + GL_STUB(RasterPos2iv, _gloffset_RasterPos2iv, RasterPos2iv@4) + GL_STUB(RasterPos2s, _gloffset_RasterPos2s, RasterPos2s@8) + GL_STUB(RasterPos2sv, _gloffset_RasterPos2sv, RasterPos2sv@4) + GL_STUB(RasterPos3d, _gloffset_RasterPos3d, RasterPos3d@24) + GL_STUB(RasterPos3dv, _gloffset_RasterPos3dv, RasterPos3dv@4) + GL_STUB(RasterPos3f, _gloffset_RasterPos3f, RasterPos3f@12) + GL_STUB(RasterPos3fv, _gloffset_RasterPos3fv, RasterPos3fv@4) + GL_STUB(RasterPos3i, _gloffset_RasterPos3i, RasterPos3i@12) + GL_STUB(RasterPos3iv, _gloffset_RasterPos3iv, RasterPos3iv@4) + GL_STUB(RasterPos3s, _gloffset_RasterPos3s, RasterPos3s@12) + GL_STUB(RasterPos3sv, _gloffset_RasterPos3sv, RasterPos3sv@4) + GL_STUB(RasterPos4d, _gloffset_RasterPos4d, RasterPos4d@32) + GL_STUB(RasterPos4dv, _gloffset_RasterPos4dv, RasterPos4dv@4) + GL_STUB(RasterPos4f, _gloffset_RasterPos4f, RasterPos4f@16) + GL_STUB(RasterPos4fv, _gloffset_RasterPos4fv, RasterPos4fv@4) + GL_STUB(RasterPos4i, _gloffset_RasterPos4i, RasterPos4i@16) + GL_STUB(RasterPos4iv, _gloffset_RasterPos4iv, RasterPos4iv@4) + GL_STUB(RasterPos4s, _gloffset_RasterPos4s, RasterPos4s@16) + GL_STUB(RasterPos4sv, _gloffset_RasterPos4sv, RasterPos4sv@4) + GL_STUB(Rectd, _gloffset_Rectd, Rectd@32) + GL_STUB(Rectdv, _gloffset_Rectdv, Rectdv@8) + GL_STUB(Rectf, _gloffset_Rectf, Rectf@16) + GL_STUB(Rectfv, _gloffset_Rectfv, Rectfv@8) + GL_STUB(Recti, _gloffset_Recti, Recti@16) + GL_STUB(Rectiv, _gloffset_Rectiv, Rectiv@8) + GL_STUB(Rects, _gloffset_Rects, Rects@16) + GL_STUB(Rectsv, _gloffset_Rectsv, Rectsv@8) + GL_STUB(TexCoord1d, _gloffset_TexCoord1d, TexCoord1d@8) + GL_STUB(TexCoord1dv, _gloffset_TexCoord1dv, TexCoord1dv@4) + GL_STUB(TexCoord1f, _gloffset_TexCoord1f, TexCoord1f@4) + GL_STUB(TexCoord1fv, _gloffset_TexCoord1fv, TexCoord1fv@4) + GL_STUB(TexCoord1i, _gloffset_TexCoord1i, TexCoord1i@4) + GL_STUB(TexCoord1iv, _gloffset_TexCoord1iv, TexCoord1iv@4) + GL_STUB(TexCoord1s, _gloffset_TexCoord1s, TexCoord1s@4) + GL_STUB(TexCoord1sv, _gloffset_TexCoord1sv, TexCoord1sv@4) + GL_STUB(TexCoord2d, _gloffset_TexCoord2d, TexCoord2d@16) + GL_STUB(TexCoord2dv, _gloffset_TexCoord2dv, TexCoord2dv@4) + GL_STUB(TexCoord2f, _gloffset_TexCoord2f, TexCoord2f@8) + GL_STUB(TexCoord2fv, _gloffset_TexCoord2fv, TexCoord2fv@4) + GL_STUB(TexCoord2i, _gloffset_TexCoord2i, TexCoord2i@8) + GL_STUB(TexCoord2iv, _gloffset_TexCoord2iv, TexCoord2iv@4) + GL_STUB(TexCoord2s, _gloffset_TexCoord2s, TexCoord2s@8) + GL_STUB(TexCoord2sv, _gloffset_TexCoord2sv, TexCoord2sv@4) + GL_STUB(TexCoord3d, _gloffset_TexCoord3d, TexCoord3d@24) + GL_STUB(TexCoord3dv, _gloffset_TexCoord3dv, TexCoord3dv@4) + GL_STUB(TexCoord3f, _gloffset_TexCoord3f, TexCoord3f@12) + GL_STUB(TexCoord3fv, _gloffset_TexCoord3fv, TexCoord3fv@4) + GL_STUB(TexCoord3i, _gloffset_TexCoord3i, TexCoord3i@12) + GL_STUB(TexCoord3iv, _gloffset_TexCoord3iv, TexCoord3iv@4) + GL_STUB(TexCoord3s, _gloffset_TexCoord3s, TexCoord3s@12) + GL_STUB(TexCoord3sv, _gloffset_TexCoord3sv, TexCoord3sv@4) + GL_STUB(TexCoord4d, _gloffset_TexCoord4d, TexCoord4d@32) + GL_STUB(TexCoord4dv, _gloffset_TexCoord4dv, TexCoord4dv@4) + GL_STUB(TexCoord4f, _gloffset_TexCoord4f, TexCoord4f@16) + GL_STUB(TexCoord4fv, _gloffset_TexCoord4fv, TexCoord4fv@4) + GL_STUB(TexCoord4i, _gloffset_TexCoord4i, TexCoord4i@16) + GL_STUB(TexCoord4iv, _gloffset_TexCoord4iv, TexCoord4iv@4) + GL_STUB(TexCoord4s, _gloffset_TexCoord4s, TexCoord4s@16) + GL_STUB(TexCoord4sv, _gloffset_TexCoord4sv, TexCoord4sv@4) + GL_STUB(Vertex2d, _gloffset_Vertex2d, Vertex2d@16) + GL_STUB(Vertex2dv, _gloffset_Vertex2dv, Vertex2dv@4) + GL_STUB(Vertex2f, _gloffset_Vertex2f, Vertex2f@8) + GL_STUB(Vertex2fv, _gloffset_Vertex2fv, Vertex2fv@4) + GL_STUB(Vertex2i, _gloffset_Vertex2i, Vertex2i@8) + GL_STUB(Vertex2iv, _gloffset_Vertex2iv, Vertex2iv@4) + GL_STUB(Vertex2s, _gloffset_Vertex2s, Vertex2s@8) + GL_STUB(Vertex2sv, _gloffset_Vertex2sv, Vertex2sv@4) + GL_STUB(Vertex3d, _gloffset_Vertex3d, Vertex3d@24) + GL_STUB(Vertex3dv, _gloffset_Vertex3dv, Vertex3dv@4) + GL_STUB(Vertex3f, _gloffset_Vertex3f, Vertex3f@12) + GL_STUB(Vertex3fv, _gloffset_Vertex3fv, Vertex3fv@4) + GL_STUB(Vertex3i, _gloffset_Vertex3i, Vertex3i@12) + GL_STUB(Vertex3iv, _gloffset_Vertex3iv, Vertex3iv@4) + GL_STUB(Vertex3s, _gloffset_Vertex3s, Vertex3s@12) + GL_STUB(Vertex3sv, _gloffset_Vertex3sv, Vertex3sv@4) + GL_STUB(Vertex4d, _gloffset_Vertex4d, Vertex4d@32) + GL_STUB(Vertex4dv, _gloffset_Vertex4dv, Vertex4dv@4) + GL_STUB(Vertex4f, _gloffset_Vertex4f, Vertex4f@16) + GL_STUB(Vertex4fv, _gloffset_Vertex4fv, Vertex4fv@4) + GL_STUB(Vertex4i, _gloffset_Vertex4i, Vertex4i@16) + GL_STUB(Vertex4iv, _gloffset_Vertex4iv, Vertex4iv@4) + GL_STUB(Vertex4s, _gloffset_Vertex4s, Vertex4s@16) + GL_STUB(Vertex4sv, _gloffset_Vertex4sv, Vertex4sv@4) + GL_STUB(ClipPlane, _gloffset_ClipPlane, ClipPlane@8) + GL_STUB(ColorMaterial, _gloffset_ColorMaterial, ColorMaterial@8) + GL_STUB(CullFace, _gloffset_CullFace, CullFace@4) + GL_STUB(Fogf, _gloffset_Fogf, Fogf@8) + GL_STUB(Fogfv, _gloffset_Fogfv, Fogfv@8) + GL_STUB(Fogi, _gloffset_Fogi, Fogi@8) + GL_STUB(Fogiv, _gloffset_Fogiv, Fogiv@8) + GL_STUB(FrontFace, _gloffset_FrontFace, FrontFace@4) + GL_STUB(Hint, _gloffset_Hint, Hint@8) + GL_STUB(Lightf, _gloffset_Lightf, Lightf@12) + GL_STUB(Lightfv, _gloffset_Lightfv, Lightfv@12) + GL_STUB(Lighti, _gloffset_Lighti, Lighti@12) + GL_STUB(Lightiv, _gloffset_Lightiv, Lightiv@12) + GL_STUB(LightModelf, _gloffset_LightModelf, LightModelf@8) + GL_STUB(LightModelfv, _gloffset_LightModelfv, LightModelfv@8) + GL_STUB(LightModeli, _gloffset_LightModeli, LightModeli@8) + GL_STUB(LightModeliv, _gloffset_LightModeliv, LightModeliv@8) + GL_STUB(LineStipple, _gloffset_LineStipple, LineStipple@8) + GL_STUB(LineWidth, _gloffset_LineWidth, LineWidth@4) + GL_STUB(Materialf, _gloffset_Materialf, Materialf@12) + GL_STUB(Materialfv, _gloffset_Materialfv, Materialfv@12) + GL_STUB(Materiali, _gloffset_Materiali, Materiali@12) + GL_STUB(Materialiv, _gloffset_Materialiv, Materialiv@12) + GL_STUB(PointSize, _gloffset_PointSize, PointSize@4) + GL_STUB(PolygonMode, _gloffset_PolygonMode, PolygonMode@8) + GL_STUB(PolygonStipple, _gloffset_PolygonStipple, PolygonStipple@4) + GL_STUB(Scissor, _gloffset_Scissor, Scissor@16) + GL_STUB(ShadeModel, _gloffset_ShadeModel, ShadeModel@4) + GL_STUB(TexParameterf, _gloffset_TexParameterf, TexParameterf@12) + GL_STUB(TexParameterfv, _gloffset_TexParameterfv, TexParameterfv@12) + GL_STUB(TexParameteri, _gloffset_TexParameteri, TexParameteri@12) + GL_STUB(TexParameteriv, _gloffset_TexParameteriv, TexParameteriv@12) + GL_STUB(TexImage1D, _gloffset_TexImage1D, TexImage1D@32) + GL_STUB(TexImage2D, _gloffset_TexImage2D, TexImage2D@36) + GL_STUB(TexEnvf, _gloffset_TexEnvf, TexEnvf@12) + GL_STUB(TexEnvfv, _gloffset_TexEnvfv, TexEnvfv@12) + GL_STUB(TexEnvi, _gloffset_TexEnvi, TexEnvi@12) + GL_STUB(TexEnviv, _gloffset_TexEnviv, TexEnviv@12) + GL_STUB(TexGend, _gloffset_TexGend, TexGend@16) + GL_STUB(TexGendv, _gloffset_TexGendv, TexGendv@12) + GL_STUB(TexGenf, _gloffset_TexGenf, TexGenf@12) + GL_STUB(TexGenfv, _gloffset_TexGenfv, TexGenfv@12) + GL_STUB(TexGeni, _gloffset_TexGeni, TexGeni@12) + GL_STUB(TexGeniv, _gloffset_TexGeniv, TexGeniv@12) + GL_STUB(FeedbackBuffer, _gloffset_FeedbackBuffer, FeedbackBuffer@12) + GL_STUB(SelectBuffer, _gloffset_SelectBuffer, SelectBuffer@8) + GL_STUB(RenderMode, _gloffset_RenderMode, RenderMode@4) + GL_STUB(InitNames, _gloffset_InitNames, InitNames@0) + GL_STUB(LoadName, _gloffset_LoadName, LoadName@4) + GL_STUB(PassThrough, _gloffset_PassThrough, PassThrough@4) + GL_STUB(PopName, _gloffset_PopName, PopName@0) + GL_STUB(PushName, _gloffset_PushName, PushName@4) + GL_STUB(DrawBuffer, _gloffset_DrawBuffer, DrawBuffer@4) + GL_STUB(Clear, _gloffset_Clear, Clear@4) + GL_STUB(ClearAccum, _gloffset_ClearAccum, ClearAccum@16) + GL_STUB(ClearIndex, _gloffset_ClearIndex, ClearIndex@4) + GL_STUB(ClearColor, _gloffset_ClearColor, ClearColor@16) + GL_STUB(ClearStencil, _gloffset_ClearStencil, ClearStencil@4) + GL_STUB(ClearDepth, _gloffset_ClearDepth, ClearDepth@8) + GL_STUB(StencilMask, _gloffset_StencilMask, StencilMask@4) + GL_STUB(ColorMask, _gloffset_ColorMask, ColorMask@16) + GL_STUB(DepthMask, _gloffset_DepthMask, DepthMask@4) + GL_STUB(IndexMask, _gloffset_IndexMask, IndexMask@4) + GL_STUB(Accum, _gloffset_Accum, Accum@8) + GL_STUB(Disable, _gloffset_Disable, Disable@4) + GL_STUB(Enable, _gloffset_Enable, Enable@4) + GL_STUB(Finish, _gloffset_Finish, Finish@0) + GL_STUB(Flush, _gloffset_Flush, Flush@0) + GL_STUB(PopAttrib, _gloffset_PopAttrib, PopAttrib@0) + GL_STUB(PushAttrib, _gloffset_PushAttrib, PushAttrib@4) + GL_STUB(Map1d, _gloffset_Map1d, Map1d@32) + GL_STUB(Map1f, _gloffset_Map1f, Map1f@24) + GL_STUB(Map2d, _gloffset_Map2d, Map2d@56) + GL_STUB(Map2f, _gloffset_Map2f, Map2f@40) + GL_STUB(MapGrid1d, _gloffset_MapGrid1d, MapGrid1d@20) + GL_STUB(MapGrid1f, _gloffset_MapGrid1f, MapGrid1f@12) + GL_STUB(MapGrid2d, _gloffset_MapGrid2d, MapGrid2d@40) + GL_STUB(MapGrid2f, _gloffset_MapGrid2f, MapGrid2f@24) + GL_STUB(EvalCoord1d, _gloffset_EvalCoord1d, EvalCoord1d@8) + GL_STUB(EvalCoord1dv, _gloffset_EvalCoord1dv, EvalCoord1dv@4) + GL_STUB(EvalCoord1f, _gloffset_EvalCoord1f, EvalCoord1f@4) + GL_STUB(EvalCoord1fv, _gloffset_EvalCoord1fv, EvalCoord1fv@4) + GL_STUB(EvalCoord2d, _gloffset_EvalCoord2d, EvalCoord2d@16) + GL_STUB(EvalCoord2dv, _gloffset_EvalCoord2dv, EvalCoord2dv@4) + GL_STUB(EvalCoord2f, _gloffset_EvalCoord2f, EvalCoord2f@8) + GL_STUB(EvalCoord2fv, _gloffset_EvalCoord2fv, EvalCoord2fv@4) + GL_STUB(EvalMesh1, _gloffset_EvalMesh1, EvalMesh1@12) + GL_STUB(EvalPoint1, _gloffset_EvalPoint1, EvalPoint1@4) + GL_STUB(EvalMesh2, _gloffset_EvalMesh2, EvalMesh2@20) + GL_STUB(EvalPoint2, _gloffset_EvalPoint2, EvalPoint2@8) + GL_STUB(AlphaFunc, _gloffset_AlphaFunc, AlphaFunc@8) + GL_STUB(BlendFunc, _gloffset_BlendFunc, BlendFunc@8) + GL_STUB(LogicOp, _gloffset_LogicOp, LogicOp@4) + GL_STUB(StencilFunc, _gloffset_StencilFunc, StencilFunc@12) + GL_STUB(StencilOp, _gloffset_StencilOp, StencilOp@12) + GL_STUB(DepthFunc, _gloffset_DepthFunc, DepthFunc@4) + GL_STUB(PixelZoom, _gloffset_PixelZoom, PixelZoom@8) + GL_STUB(PixelTransferf, _gloffset_PixelTransferf, PixelTransferf@8) + GL_STUB(PixelTransferi, _gloffset_PixelTransferi, PixelTransferi@8) + GL_STUB(PixelStoref, _gloffset_PixelStoref, PixelStoref@8) + GL_STUB(PixelStorei, _gloffset_PixelStorei, PixelStorei@8) + GL_STUB(PixelMapfv, _gloffset_PixelMapfv, PixelMapfv@12) + GL_STUB(PixelMapuiv, _gloffset_PixelMapuiv, PixelMapuiv@12) + GL_STUB(PixelMapusv, _gloffset_PixelMapusv, PixelMapusv@12) + GL_STUB(ReadBuffer, _gloffset_ReadBuffer, ReadBuffer@4) + GL_STUB(CopyPixels, _gloffset_CopyPixels, CopyPixels@20) + GL_STUB(ReadPixels, _gloffset_ReadPixels, ReadPixels@28) + GL_STUB(DrawPixels, _gloffset_DrawPixels, DrawPixels@20) + GL_STUB(GetBooleanv, _gloffset_GetBooleanv, GetBooleanv@8) + GL_STUB(GetClipPlane, _gloffset_GetClipPlane, GetClipPlane@8) + GL_STUB(GetDoublev, _gloffset_GetDoublev, GetDoublev@8) + GL_STUB(GetError, _gloffset_GetError, GetError@0) + GL_STUB(GetFloatv, _gloffset_GetFloatv, GetFloatv@8) + GL_STUB(GetIntegerv, _gloffset_GetIntegerv, GetIntegerv@8) + GL_STUB(GetLightfv, _gloffset_GetLightfv, GetLightfv@12) + GL_STUB(GetLightiv, _gloffset_GetLightiv, GetLightiv@12) + GL_STUB(GetMapdv, _gloffset_GetMapdv, GetMapdv@12) + GL_STUB(GetMapfv, _gloffset_GetMapfv, GetMapfv@12) + GL_STUB(GetMapiv, _gloffset_GetMapiv, GetMapiv@12) + GL_STUB(GetMaterialfv, _gloffset_GetMaterialfv, GetMaterialfv@12) + GL_STUB(GetMaterialiv, _gloffset_GetMaterialiv, GetMaterialiv@12) + GL_STUB(GetPixelMapfv, _gloffset_GetPixelMapfv, GetPixelMapfv@8) + GL_STUB(GetPixelMapuiv, _gloffset_GetPixelMapuiv, GetPixelMapuiv@8) + GL_STUB(GetPixelMapusv, _gloffset_GetPixelMapusv, GetPixelMapusv@8) + GL_STUB(GetPolygonStipple, _gloffset_GetPolygonStipple, GetPolygonStipple@4) + GL_STUB(GetString, _gloffset_GetString, GetString@4) + GL_STUB(GetTexEnvfv, _gloffset_GetTexEnvfv, GetTexEnvfv@12) + GL_STUB(GetTexEnviv, _gloffset_GetTexEnviv, GetTexEnviv@12) + GL_STUB(GetTexGendv, _gloffset_GetTexGendv, GetTexGendv@12) + GL_STUB(GetTexGenfv, _gloffset_GetTexGenfv, GetTexGenfv@12) + GL_STUB(GetTexGeniv, _gloffset_GetTexGeniv, GetTexGeniv@12) + GL_STUB(GetTexImage, _gloffset_GetTexImage, GetTexImage@20) + GL_STUB(GetTexParameterfv, _gloffset_GetTexParameterfv, GetTexParameterfv@12) + GL_STUB(GetTexParameteriv, _gloffset_GetTexParameteriv, GetTexParameteriv@12) + GL_STUB(GetTexLevelParameterfv, _gloffset_GetTexLevelParameterfv, GetTexLevelParameterfv@16) + GL_STUB(GetTexLevelParameteriv, _gloffset_GetTexLevelParameteriv, GetTexLevelParameteriv@16) + GL_STUB(IsEnabled, _gloffset_IsEnabled, IsEnabled@4) + GL_STUB(IsList, _gloffset_IsList, IsList@4) + GL_STUB(DepthRange, _gloffset_DepthRange, DepthRange@16) + GL_STUB(Frustum, _gloffset_Frustum, Frustum@48) + GL_STUB(LoadIdentity, _gloffset_LoadIdentity, LoadIdentity@0) + GL_STUB(LoadMatrixf, _gloffset_LoadMatrixf, LoadMatrixf@4) + GL_STUB(LoadMatrixd, _gloffset_LoadMatrixd, LoadMatrixd@4) + GL_STUB(MatrixMode, _gloffset_MatrixMode, MatrixMode@4) + GL_STUB(MultMatrixf, _gloffset_MultMatrixf, MultMatrixf@4) + GL_STUB(MultMatrixd, _gloffset_MultMatrixd, MultMatrixd@4) + GL_STUB(Ortho, _gloffset_Ortho, Ortho@48) + GL_STUB(PopMatrix, _gloffset_PopMatrix, PopMatrix@0) + GL_STUB(PushMatrix, _gloffset_PushMatrix, PushMatrix@0) + GL_STUB(Rotated, _gloffset_Rotated, Rotated@32) + GL_STUB(Rotatef, _gloffset_Rotatef, Rotatef@16) + GL_STUB(Scaled, _gloffset_Scaled, Scaled@24) + GL_STUB(Scalef, _gloffset_Scalef, Scalef@12) + GL_STUB(Translated, _gloffset_Translated, Translated@24) + GL_STUB(Translatef, _gloffset_Translatef, Translatef@12) + GL_STUB(Viewport, _gloffset_Viewport, Viewport@16) + GL_STUB(ArrayElement, _gloffset_ArrayElement, ArrayElement@4) + GL_STUB(BindTexture, _gloffset_BindTexture, BindTexture@8) + GL_STUB(ColorPointer, _gloffset_ColorPointer, ColorPointer@16) + GL_STUB(DisableClientState, _gloffset_DisableClientState, DisableClientState@4) + GL_STUB(DrawArrays, _gloffset_DrawArrays, DrawArrays@12) + GL_STUB(DrawElements, _gloffset_DrawElements, DrawElements@16) + GL_STUB(EdgeFlagPointer, _gloffset_EdgeFlagPointer, EdgeFlagPointer@8) + GL_STUB(EnableClientState, _gloffset_EnableClientState, EnableClientState@4) + GL_STUB(IndexPointer, _gloffset_IndexPointer, IndexPointer@12) + GL_STUB(Indexub, _gloffset_Indexub, Indexub@4) + GL_STUB(Indexubv, _gloffset_Indexubv, Indexubv@4) + GL_STUB(InterleavedArrays, _gloffset_InterleavedArrays, InterleavedArrays@12) + GL_STUB(NormalPointer, _gloffset_NormalPointer, NormalPointer@12) + GL_STUB(PolygonOffset, _gloffset_PolygonOffset, PolygonOffset@8) + GL_STUB(TexCoordPointer, _gloffset_TexCoordPointer, TexCoordPointer@16) + GL_STUB(VertexPointer, _gloffset_VertexPointer, VertexPointer@16) + GL_STUB(AreTexturesResident, _gloffset_AreTexturesResident, AreTexturesResident@12) + GL_STUB(CopyTexImage1D, _gloffset_CopyTexImage1D, CopyTexImage1D@28) + GL_STUB(CopyTexImage2D, _gloffset_CopyTexImage2D, CopyTexImage2D@32) + GL_STUB(CopyTexSubImage1D, _gloffset_CopyTexSubImage1D, CopyTexSubImage1D@24) + GL_STUB(CopyTexSubImage2D, _gloffset_CopyTexSubImage2D, CopyTexSubImage2D@32) + GL_STUB(DeleteTextures, _gloffset_DeleteTextures, DeleteTextures@8) + GL_STUB(GenTextures, _gloffset_GenTextures, GenTextures@8) + GL_STUB(GetPointerv, _gloffset_GetPointerv, GetPointerv@8) + GL_STUB(IsTexture, _gloffset_IsTexture, IsTexture@4) + GL_STUB(PrioritizeTextures, _gloffset_PrioritizeTextures, PrioritizeTextures@12) + GL_STUB(TexSubImage1D, _gloffset_TexSubImage1D, TexSubImage1D@28) + GL_STUB(TexSubImage2D, _gloffset_TexSubImage2D, TexSubImage2D@36) + GL_STUB(PopClientAttrib, _gloffset_PopClientAttrib, PopClientAttrib@0) + GL_STUB(PushClientAttrib, _gloffset_PushClientAttrib, PushClientAttrib@4) + GL_STUB(BlendColor, _gloffset_BlendColor, BlendColor@16) + GL_STUB(BlendEquation, _gloffset_BlendEquation, BlendEquation@4) + GL_STUB(DrawRangeElements, _gloffset_DrawRangeElements, DrawRangeElements@24) + GL_STUB(ColorTable, _gloffset_ColorTable, ColorTable@24) + GL_STUB(ColorTableParameterfv, _gloffset_ColorTableParameterfv, ColorTableParameterfv@12) + GL_STUB(ColorTableParameteriv, _gloffset_ColorTableParameteriv, ColorTableParameteriv@12) + GL_STUB(CopyColorTable, _gloffset_CopyColorTable, CopyColorTable@20) + GL_STUB(GetColorTable, _gloffset_GetColorTable, GetColorTable@16) + GL_STUB(GetColorTableParameterfv, _gloffset_GetColorTableParameterfv, GetColorTableParameterfv@12) + GL_STUB(GetColorTableParameteriv, _gloffset_GetColorTableParameteriv, GetColorTableParameteriv@12) + GL_STUB(ColorSubTable, _gloffset_ColorSubTable, ColorSubTable@24) + GL_STUB(CopyColorSubTable, _gloffset_CopyColorSubTable, CopyColorSubTable@20) + GL_STUB(ConvolutionFilter1D, _gloffset_ConvolutionFilter1D, ConvolutionFilter1D@24) + GL_STUB(ConvolutionFilter2D, _gloffset_ConvolutionFilter2D, ConvolutionFilter2D@28) + GL_STUB(ConvolutionParameterf, _gloffset_ConvolutionParameterf, ConvolutionParameterf@12) + GL_STUB(ConvolutionParameterfv, _gloffset_ConvolutionParameterfv, ConvolutionParameterfv@12) + GL_STUB(ConvolutionParameteri, _gloffset_ConvolutionParameteri, ConvolutionParameteri@12) + GL_STUB(ConvolutionParameteriv, _gloffset_ConvolutionParameteriv, ConvolutionParameteriv@12) + GL_STUB(CopyConvolutionFilter1D, _gloffset_CopyConvolutionFilter1D, CopyConvolutionFilter1D@20) + GL_STUB(CopyConvolutionFilter2D, _gloffset_CopyConvolutionFilter2D, CopyConvolutionFilter2D@24) + GL_STUB(GetConvolutionFilter, _gloffset_GetConvolutionFilter, GetConvolutionFilter@16) + GL_STUB(GetConvolutionParameterfv, _gloffset_GetConvolutionParameterfv, GetConvolutionParameterfv@12) + GL_STUB(GetConvolutionParameteriv, _gloffset_GetConvolutionParameteriv, GetConvolutionParameteriv@12) + GL_STUB(GetSeparableFilter, _gloffset_GetSeparableFilter, GetSeparableFilter@24) + GL_STUB(SeparableFilter2D, _gloffset_SeparableFilter2D, SeparableFilter2D@32) + GL_STUB(GetHistogram, _gloffset_GetHistogram, GetHistogram@20) + GL_STUB(GetHistogramParameterfv, _gloffset_GetHistogramParameterfv, GetHistogramParameterfv@12) + GL_STUB(GetHistogramParameteriv, _gloffset_GetHistogramParameteriv, GetHistogramParameteriv@12) + GL_STUB(GetMinmax, _gloffset_GetMinmax, GetMinmax@20) + GL_STUB(GetMinmaxParameterfv, _gloffset_GetMinmaxParameterfv, GetMinmaxParameterfv@12) + GL_STUB(GetMinmaxParameteriv, _gloffset_GetMinmaxParameteriv, GetMinmaxParameteriv@12) + GL_STUB(Histogram, _gloffset_Histogram, Histogram@16) + GL_STUB(Minmax, _gloffset_Minmax, Minmax@12) + GL_STUB(ResetHistogram, _gloffset_ResetHistogram, ResetHistogram@4) + GL_STUB(ResetMinmax, _gloffset_ResetMinmax, ResetMinmax@4) + GL_STUB(TexImage3D, _gloffset_TexImage3D, TexImage3D@40) + GL_STUB(TexSubImage3D, _gloffset_TexSubImage3D, TexSubImage3D@44) + GL_STUB(CopyTexSubImage3D, _gloffset_CopyTexSubImage3D, CopyTexSubImage3D@36) + GL_STUB(ActiveTextureARB, _gloffset_ActiveTextureARB, ActiveTextureARB@4) + GL_STUB(ClientActiveTextureARB, _gloffset_ClientActiveTextureARB, ClientActiveTextureARB@4) + GL_STUB(MultiTexCoord1dARB, _gloffset_MultiTexCoord1dARB, MultiTexCoord1dARB@12) + GL_STUB(MultiTexCoord1dvARB, _gloffset_MultiTexCoord1dvARB, MultiTexCoord1dvARB@8) + GL_STUB(MultiTexCoord1fARB, _gloffset_MultiTexCoord1fARB, MultiTexCoord1fARB@8) + GL_STUB(MultiTexCoord1fvARB, _gloffset_MultiTexCoord1fvARB, MultiTexCoord1fvARB@8) + GL_STUB(MultiTexCoord1iARB, _gloffset_MultiTexCoord1iARB, MultiTexCoord1iARB@8) + GL_STUB(MultiTexCoord1ivARB, _gloffset_MultiTexCoord1ivARB, MultiTexCoord1ivARB@8) + GL_STUB(MultiTexCoord1sARB, _gloffset_MultiTexCoord1sARB, MultiTexCoord1sARB@8) + GL_STUB(MultiTexCoord1svARB, _gloffset_MultiTexCoord1svARB, MultiTexCoord1svARB@8) + GL_STUB(MultiTexCoord2dARB, _gloffset_MultiTexCoord2dARB, MultiTexCoord2dARB@20) + GL_STUB(MultiTexCoord2dvARB, _gloffset_MultiTexCoord2dvARB, MultiTexCoord2dvARB@8) + GL_STUB(MultiTexCoord2fARB, _gloffset_MultiTexCoord2fARB, MultiTexCoord2fARB@12) + GL_STUB(MultiTexCoord2fvARB, _gloffset_MultiTexCoord2fvARB, MultiTexCoord2fvARB@8) + GL_STUB(MultiTexCoord2iARB, _gloffset_MultiTexCoord2iARB, MultiTexCoord2iARB@12) + GL_STUB(MultiTexCoord2ivARB, _gloffset_MultiTexCoord2ivARB, MultiTexCoord2ivARB@8) + GL_STUB(MultiTexCoord2sARB, _gloffset_MultiTexCoord2sARB, MultiTexCoord2sARB@12) + GL_STUB(MultiTexCoord2svARB, _gloffset_MultiTexCoord2svARB, MultiTexCoord2svARB@8) + GL_STUB(MultiTexCoord3dARB, _gloffset_MultiTexCoord3dARB, MultiTexCoord3dARB@28) + GL_STUB(MultiTexCoord3dvARB, _gloffset_MultiTexCoord3dvARB, MultiTexCoord3dvARB@8) + GL_STUB(MultiTexCoord3fARB, _gloffset_MultiTexCoord3fARB, MultiTexCoord3fARB@16) + GL_STUB(MultiTexCoord3fvARB, _gloffset_MultiTexCoord3fvARB, MultiTexCoord3fvARB@8) + GL_STUB(MultiTexCoord3iARB, _gloffset_MultiTexCoord3iARB, MultiTexCoord3iARB@16) + GL_STUB(MultiTexCoord3ivARB, _gloffset_MultiTexCoord3ivARB, MultiTexCoord3ivARB@8) + GL_STUB(MultiTexCoord3sARB, _gloffset_MultiTexCoord3sARB, MultiTexCoord3sARB@16) + GL_STUB(MultiTexCoord3svARB, _gloffset_MultiTexCoord3svARB, MultiTexCoord3svARB@8) + GL_STUB(MultiTexCoord4dARB, _gloffset_MultiTexCoord4dARB, MultiTexCoord4dARB@36) + GL_STUB(MultiTexCoord4dvARB, _gloffset_MultiTexCoord4dvARB, MultiTexCoord4dvARB@8) + GL_STUB(MultiTexCoord4fARB, _gloffset_MultiTexCoord4fARB, MultiTexCoord4fARB@20) + GL_STUB(MultiTexCoord4fvARB, _gloffset_MultiTexCoord4fvARB, MultiTexCoord4fvARB@8) + GL_STUB(MultiTexCoord4iARB, _gloffset_MultiTexCoord4iARB, MultiTexCoord4iARB@20) + GL_STUB(MultiTexCoord4ivARB, _gloffset_MultiTexCoord4ivARB, MultiTexCoord4ivARB@8) + GL_STUB(MultiTexCoord4sARB, _gloffset_MultiTexCoord4sARB, MultiTexCoord4sARB@20) + GL_STUB(MultiTexCoord4svARB, _gloffset_MultiTexCoord4svARB, MultiTexCoord4svARB@8) + GL_STUB(LoadTransposeMatrixfARB, _gloffset_LoadTransposeMatrixfARB, LoadTransposeMatrixfARB@4) + GL_STUB(LoadTransposeMatrixdARB, _gloffset_LoadTransposeMatrixdARB, LoadTransposeMatrixdARB@4) + GL_STUB(MultTransposeMatrixfARB, _gloffset_MultTransposeMatrixfARB, MultTransposeMatrixfARB@4) + GL_STUB(MultTransposeMatrixdARB, _gloffset_MultTransposeMatrixdARB, MultTransposeMatrixdARB@4) + GL_STUB(SampleCoverageARB, _gloffset_SampleCoverageARB, SampleCoverageARB@8) + GL_STUB(__unused413, _gloffset___unused413, __unused413@0) + GL_STUB(PolygonOffsetEXT, _gloffset_PolygonOffsetEXT, PolygonOffsetEXT@8) + GL_STUB(GetTexFilterFuncSGIS, _gloffset_GetTexFilterFuncSGIS, GetTexFilterFuncSGIS@12) + GL_STUB(TexFilterFuncSGIS, _gloffset_TexFilterFuncSGIS, TexFilterFuncSGIS@16) + GL_STUB(GetHistogramEXT, _gloffset_GetHistogramEXT, GetHistogramEXT@20) + GL_STUB(GetHistogramParameterfvEXT, _gloffset_GetHistogramParameterfvEXT, GetHistogramParameterfvEXT@12) + GL_STUB(GetHistogramParameterivEXT, _gloffset_GetHistogramParameterivEXT, GetHistogramParameterivEXT@12) + GL_STUB(GetMinmaxEXT, _gloffset_GetMinmaxEXT, GetMinmaxEXT@20) + GL_STUB(GetMinmaxParameterfvEXT, _gloffset_GetMinmaxParameterfvEXT, GetMinmaxParameterfvEXT@12) + GL_STUB(GetMinmaxParameterivEXT, _gloffset_GetMinmaxParameterivEXT, GetMinmaxParameterivEXT@12) + GL_STUB(GetConvolutionFilterEXT, _gloffset_GetConvolutionFilterEXT, GetConvolutionFilterEXT@16) + GL_STUB(GetConvolutionParameterfvEXT, _gloffset_GetConvolutionParameterfvEXT, GetConvolutionParameterfvEXT@12) + GL_STUB(GetConvolutionParameterivEXT, _gloffset_GetConvolutionParameterivEXT, GetConvolutionParameterivEXT@12) + GL_STUB(GetSeparableFilterEXT, _gloffset_GetSeparableFilterEXT, GetSeparableFilterEXT@24) + GL_STUB(GetColorTableSGI, _gloffset_GetColorTableSGI, GetColorTableSGI@16) + GL_STUB(GetColorTableParameterfvSGI, _gloffset_GetColorTableParameterfvSGI, GetColorTableParameterfvSGI@12) + GL_STUB(GetColorTableParameterivSGI, _gloffset_GetColorTableParameterivSGI, GetColorTableParameterivSGI@12) + GL_STUB(PixelTexGenSGIX, _gloffset_PixelTexGenSGIX, PixelTexGenSGIX@4) + GL_STUB(PixelTexGenParameteriSGIS, _gloffset_PixelTexGenParameteriSGIS, PixelTexGenParameteriSGIS@8) + GL_STUB(PixelTexGenParameterivSGIS, _gloffset_PixelTexGenParameterivSGIS, PixelTexGenParameterivSGIS@8) + GL_STUB(PixelTexGenParameterfSGIS, _gloffset_PixelTexGenParameterfSGIS, PixelTexGenParameterfSGIS@8) + GL_STUB(PixelTexGenParameterfvSGIS, _gloffset_PixelTexGenParameterfvSGIS, PixelTexGenParameterfvSGIS@8) + GL_STUB(GetPixelTexGenParameterivSGIS, _gloffset_GetPixelTexGenParameterivSGIS, GetPixelTexGenParameterivSGIS@8) + GL_STUB(GetPixelTexGenParameterfvSGIS, _gloffset_GetPixelTexGenParameterfvSGIS, GetPixelTexGenParameterfvSGIS@8) + GL_STUB(TexImage4DSGIS, _gloffset_TexImage4DSGIS, TexImage4DSGIS@44) + GL_STUB(TexSubImage4DSGIS, _gloffset_TexSubImage4DSGIS, TexSubImage4DSGIS@52) + GL_STUB(AreTexturesResidentEXT, _gloffset_AreTexturesResidentEXT, AreTexturesResidentEXT@12) + GL_STUB(GenTexturesEXT, _gloffset_GenTexturesEXT, GenTexturesEXT@8) + GL_STUB(IsTextureEXT, _gloffset_IsTextureEXT, IsTextureEXT@4) + GL_STUB(DetailTexFuncSGIS, _gloffset_DetailTexFuncSGIS, DetailTexFuncSGIS@12) + GL_STUB(GetDetailTexFuncSGIS, _gloffset_GetDetailTexFuncSGIS, GetDetailTexFuncSGIS@8) + GL_STUB(SharpenTexFuncSGIS, _gloffset_SharpenTexFuncSGIS, SharpenTexFuncSGIS@12) + GL_STUB(GetSharpenTexFuncSGIS, _gloffset_GetSharpenTexFuncSGIS, GetSharpenTexFuncSGIS@8) + GL_STUB(SampleMaskSGIS, _gloffset_SampleMaskSGIS, SampleMaskSGIS@8) + GL_STUB(SamplePatternSGIS, _gloffset_SamplePatternSGIS, SamplePatternSGIS@4) + GL_STUB(ColorPointerEXT, _gloffset_ColorPointerEXT, ColorPointerEXT@20) + GL_STUB(EdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT, EdgeFlagPointerEXT@12) + GL_STUB(IndexPointerEXT, _gloffset_IndexPointerEXT, IndexPointerEXT@16) + GL_STUB(NormalPointerEXT, _gloffset_NormalPointerEXT, NormalPointerEXT@16) + GL_STUB(TexCoordPointerEXT, _gloffset_TexCoordPointerEXT, TexCoordPointerEXT@20) + GL_STUB(VertexPointerEXT, _gloffset_VertexPointerEXT, VertexPointerEXT@20) + GL_STUB(SpriteParameterfSGIX, _gloffset_SpriteParameterfSGIX, SpriteParameterfSGIX@8) + GL_STUB(SpriteParameterfvSGIX, _gloffset_SpriteParameterfvSGIX, SpriteParameterfvSGIX@8) + GL_STUB(SpriteParameteriSGIX, _gloffset_SpriteParameteriSGIX, SpriteParameteriSGIX@8) + GL_STUB(SpriteParameterivSGIX, _gloffset_SpriteParameterivSGIX, SpriteParameterivSGIX@8) + GL_STUB(PointParameterfEXT, _gloffset_PointParameterfEXT, PointParameterfEXT@8) + GL_STUB(PointParameterfvEXT, _gloffset_PointParameterfvEXT, PointParameterfvEXT@8) + GL_STUB(GetInstrumentsSGIX, _gloffset_GetInstrumentsSGIX, GetInstrumentsSGIX@0) + GL_STUB(InstrumentsBufferSGIX, _gloffset_InstrumentsBufferSGIX, InstrumentsBufferSGIX@8) + GL_STUB(PollInstrumentsSGIX, _gloffset_PollInstrumentsSGIX, PollInstrumentsSGIX@4) + GL_STUB(ReadInstrumentsSGIX, _gloffset_ReadInstrumentsSGIX, ReadInstrumentsSGIX@4) + GL_STUB(StartInstrumentsSGIX, _gloffset_StartInstrumentsSGIX, StartInstrumentsSGIX@0) + GL_STUB(StopInstrumentsSGIX, _gloffset_StopInstrumentsSGIX, StopInstrumentsSGIX@4) + GL_STUB(FrameZoomSGIX, _gloffset_FrameZoomSGIX, FrameZoomSGIX@4) + GL_STUB(TagSampleBufferSGIX, _gloffset_TagSampleBufferSGIX, TagSampleBufferSGIX@0) + GL_STUB(ReferencePlaneSGIX, _gloffset_ReferencePlaneSGIX, ReferencePlaneSGIX@4) + GL_STUB(FlushRasterSGIX, _gloffset_FlushRasterSGIX, FlushRasterSGIX@0) + GL_STUB(GetListParameterfvSGIX, _gloffset_GetListParameterfvSGIX, GetListParameterfvSGIX@12) + GL_STUB(GetListParameterivSGIX, _gloffset_GetListParameterivSGIX, GetListParameterivSGIX@12) + GL_STUB(ListParameterfSGIX, _gloffset_ListParameterfSGIX, ListParameterfSGIX@12) + GL_STUB(ListParameterfvSGIX, _gloffset_ListParameterfvSGIX, ListParameterfvSGIX@12) + GL_STUB(ListParameteriSGIX, _gloffset_ListParameteriSGIX, ListParameteriSGIX@12) + GL_STUB(ListParameterivSGIX, _gloffset_ListParameterivSGIX, ListParameterivSGIX@12) + GL_STUB(FragmentColorMaterialSGIX, _gloffset_FragmentColorMaterialSGIX, FragmentColorMaterialSGIX@8) + GL_STUB(FragmentLightfSGIX, _gloffset_FragmentLightfSGIX, FragmentLightfSGIX@12) + GL_STUB(FragmentLightfvSGIX, _gloffset_FragmentLightfvSGIX, FragmentLightfvSGIX@12) + GL_STUB(FragmentLightiSGIX, _gloffset_FragmentLightiSGIX, FragmentLightiSGIX@12) + GL_STUB(FragmentLightivSGIX, _gloffset_FragmentLightivSGIX, FragmentLightivSGIX@12) + GL_STUB(FragmentLightModelfSGIX, _gloffset_FragmentLightModelfSGIX, FragmentLightModelfSGIX@8) + GL_STUB(FragmentLightModelfvSGIX, _gloffset_FragmentLightModelfvSGIX, FragmentLightModelfvSGIX@8) + GL_STUB(FragmentLightModeliSGIX, _gloffset_FragmentLightModeliSGIX, FragmentLightModeliSGIX@8) + GL_STUB(FragmentLightModelivSGIX, _gloffset_FragmentLightModelivSGIX, FragmentLightModelivSGIX@8) + GL_STUB(FragmentMaterialfSGIX, _gloffset_FragmentMaterialfSGIX, FragmentMaterialfSGIX@12) + GL_STUB(FragmentMaterialfvSGIX, _gloffset_FragmentMaterialfvSGIX, FragmentMaterialfvSGIX@12) + GL_STUB(FragmentMaterialiSGIX, _gloffset_FragmentMaterialiSGIX, FragmentMaterialiSGIX@12) + GL_STUB(FragmentMaterialivSGIX, _gloffset_FragmentMaterialivSGIX, FragmentMaterialivSGIX@12) + GL_STUB(GetFragmentLightfvSGIX, _gloffset_GetFragmentLightfvSGIX, GetFragmentLightfvSGIX@12) + GL_STUB(GetFragmentLightivSGIX, _gloffset_GetFragmentLightivSGIX, GetFragmentLightivSGIX@12) + GL_STUB(GetFragmentMaterialfvSGIX, _gloffset_GetFragmentMaterialfvSGIX, GetFragmentMaterialfvSGIX@12) + GL_STUB(GetFragmentMaterialivSGIX, _gloffset_GetFragmentMaterialivSGIX, GetFragmentMaterialivSGIX@12) + GL_STUB(LightEnviSGIX, _gloffset_LightEnviSGIX, LightEnviSGIX@8) + GL_STUB(VertexWeightfEXT, _gloffset_VertexWeightfEXT, VertexWeightfEXT@4) + GL_STUB(VertexWeightfvEXT, _gloffset_VertexWeightfvEXT, VertexWeightfvEXT@4) + GL_STUB(VertexWeightPointerEXT, _gloffset_VertexWeightPointerEXT, VertexWeightPointerEXT@16) + GL_STUB(FlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV, FlushVertexArrayRangeNV@0) + GL_STUB(VertexArrayRangeNV, _gloffset_VertexArrayRangeNV, VertexArrayRangeNV@8) + GL_STUB(CombinerParameterfvNV, _gloffset_CombinerParameterfvNV, CombinerParameterfvNV@8) + GL_STUB(CombinerParameterfNV, _gloffset_CombinerParameterfNV, CombinerParameterfNV@8) + GL_STUB(CombinerParameterivNV, _gloffset_CombinerParameterivNV, CombinerParameterivNV@8) + GL_STUB(CombinerParameteriNV, _gloffset_CombinerParameteriNV, CombinerParameteriNV@8) + GL_STUB(CombinerInputNV, _gloffset_CombinerInputNV, CombinerInputNV@24) + GL_STUB(CombinerOutputNV, _gloffset_CombinerOutputNV, CombinerOutputNV@40) + GL_STUB(FinalCombinerInputNV, _gloffset_FinalCombinerInputNV, FinalCombinerInputNV@16) + GL_STUB(GetCombinerInputParameterfvNV, _gloffset_GetCombinerInputParameterfvNV, GetCombinerInputParameterfvNV@20) + GL_STUB(GetCombinerInputParameterivNV, _gloffset_GetCombinerInputParameterivNV, GetCombinerInputParameterivNV@20) + GL_STUB(GetCombinerOutputParameterfvNV, _gloffset_GetCombinerOutputParameterfvNV, GetCombinerOutputParameterfvNV@16) + GL_STUB(GetCombinerOutputParameterivNV, _gloffset_GetCombinerOutputParameterivNV, GetCombinerOutputParameterivNV@16) + GL_STUB(GetFinalCombinerInputParameterfvNV, _gloffset_GetFinalCombinerInputParameterfvNV, GetFinalCombinerInputParameterfvNV@12) + GL_STUB(GetFinalCombinerInputParameterivNV, _gloffset_GetFinalCombinerInputParameterivNV, GetFinalCombinerInputParameterivNV@12) + GL_STUB(ResizeBuffersMESA, _gloffset_ResizeBuffersMESA, ResizeBuffersMESA@0) + GL_STUB(WindowPos2dMESA, _gloffset_WindowPos2dMESA, WindowPos2dMESA@16) + GL_STUB(WindowPos2dvMESA, _gloffset_WindowPos2dvMESA, WindowPos2dvMESA@4) + GL_STUB(WindowPos2fMESA, _gloffset_WindowPos2fMESA, WindowPos2fMESA@8) + GL_STUB(WindowPos2fvMESA, _gloffset_WindowPos2fvMESA, WindowPos2fvMESA@4) + GL_STUB(WindowPos2iMESA, _gloffset_WindowPos2iMESA, WindowPos2iMESA@8) + GL_STUB(WindowPos2ivMESA, _gloffset_WindowPos2ivMESA, WindowPos2ivMESA@4) + GL_STUB(WindowPos2sMESA, _gloffset_WindowPos2sMESA, WindowPos2sMESA@8) + GL_STUB(WindowPos2svMESA, _gloffset_WindowPos2svMESA, WindowPos2svMESA@4) + GL_STUB(WindowPos3dMESA, _gloffset_WindowPos3dMESA, WindowPos3dMESA@24) + GL_STUB(WindowPos3dvMESA, _gloffset_WindowPos3dvMESA, WindowPos3dvMESA@4) + GL_STUB(WindowPos3fMESA, _gloffset_WindowPos3fMESA, WindowPos3fMESA@12) + GL_STUB(WindowPos3fvMESA, _gloffset_WindowPos3fvMESA, WindowPos3fvMESA@4) + GL_STUB(WindowPos3iMESA, _gloffset_WindowPos3iMESA, WindowPos3iMESA@12) + GL_STUB(WindowPos3ivMESA, _gloffset_WindowPos3ivMESA, WindowPos3ivMESA@4) + GL_STUB(WindowPos3sMESA, _gloffset_WindowPos3sMESA, WindowPos3sMESA@12) + GL_STUB(WindowPos3svMESA, _gloffset_WindowPos3svMESA, WindowPos3svMESA@4) + GL_STUB(WindowPos4dMESA, _gloffset_WindowPos4dMESA, WindowPos4dMESA@32) + GL_STUB(WindowPos4dvMESA, _gloffset_WindowPos4dvMESA, WindowPos4dvMESA@4) + GL_STUB(WindowPos4fMESA, _gloffset_WindowPos4fMESA, WindowPos4fMESA@16) + GL_STUB(WindowPos4fvMESA, _gloffset_WindowPos4fvMESA, WindowPos4fvMESA@4) + GL_STUB(WindowPos4iMESA, _gloffset_WindowPos4iMESA, WindowPos4iMESA@16) + GL_STUB(WindowPos4ivMESA, _gloffset_WindowPos4ivMESA, WindowPos4ivMESA@4) + GL_STUB(WindowPos4sMESA, _gloffset_WindowPos4sMESA, WindowPos4sMESA@16) + GL_STUB(WindowPos4svMESA, _gloffset_WindowPos4svMESA, WindowPos4svMESA@4) + GL_STUB(BlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparateEXT@16) + GL_STUB(IndexMaterialEXT, _gloffset_IndexMaterialEXT, IndexMaterialEXT@8) + GL_STUB(IndexFuncEXT, _gloffset_IndexFuncEXT, IndexFuncEXT@8) + GL_STUB(LockArraysEXT, _gloffset_LockArraysEXT, LockArraysEXT@8) + GL_STUB(UnlockArraysEXT, _gloffset_UnlockArraysEXT, UnlockArraysEXT@0) + GL_STUB(CullParameterdvEXT, _gloffset_CullParameterdvEXT, CullParameterdvEXT@8) + GL_STUB(CullParameterfvEXT, _gloffset_CullParameterfvEXT, CullParameterfvEXT@8) + GL_STUB(HintPGI, _gloffset_HintPGI, HintPGI@8) + GL_STUB(FogCoordfEXT, _gloffset_FogCoordfEXT, FogCoordfEXT@4) + GL_STUB(FogCoordfvEXT, _gloffset_FogCoordfvEXT, FogCoordfvEXT@4) + GL_STUB(FogCoorddEXT, _gloffset_FogCoorddEXT, FogCoorddEXT@8) + GL_STUB(FogCoorddvEXT, _gloffset_FogCoorddvEXT, FogCoorddvEXT@4) + GL_STUB(FogCoordPointerEXT, _gloffset_FogCoordPointerEXT, FogCoordPointerEXT@12) + GL_STUB(GetColorTableEXT, _gloffset_GetColorTableEXT, GetColorTableEXT@16) + GL_STUB(GetColorTableParameterivEXT, _gloffset_GetColorTableParameterivEXT, GetColorTableParameterivEXT@12) + GL_STUB(GetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfvEXT, GetColorTableParameterfvEXT@12) + GL_STUB(TbufferMask3DFX, _gloffset_TbufferMask3DFX, TbufferMask3DFX@4) + GL_STUB(CompressedTexImage3DARB, _gloffset_CompressedTexImage3DARB, CompressedTexImage3DARB@36) + GL_STUB(CompressedTexImage2DARB, _gloffset_CompressedTexImage2DARB, CompressedTexImage2DARB@32) + GL_STUB(CompressedTexImage1DARB, _gloffset_CompressedTexImage1DARB, CompressedTexImage1DARB@28) + GL_STUB(CompressedTexSubImage3DARB, _gloffset_CompressedTexSubImage3DARB, CompressedTexSubImage3DARB@44) + GL_STUB(CompressedTexSubImage2DARB, _gloffset_CompressedTexSubImage2DARB, CompressedTexSubImage2DARB@36) + GL_STUB(CompressedTexSubImage1DARB, _gloffset_CompressedTexSubImage1DARB, CompressedTexSubImage1DARB@28) + GL_STUB(GetCompressedTexImageARB, _gloffset_GetCompressedTexImageARB, GetCompressedTexImageARB@12) + GL_STUB(SecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT, SecondaryColor3bEXT@12) + GL_STUB(SecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT, SecondaryColor3bvEXT@4) + GL_STUB(SecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT, SecondaryColor3dEXT@24) + GL_STUB(SecondaryColor3dvEXT, _gloffset_SecondaryColor3dvEXT, SecondaryColor3dvEXT@4) + GL_STUB(SecondaryColor3fEXT, _gloffset_SecondaryColor3fEXT, SecondaryColor3fEXT@12) + GL_STUB(SecondaryColor3fvEXT, _gloffset_SecondaryColor3fvEXT, SecondaryColor3fvEXT@4) + GL_STUB(SecondaryColor3iEXT, _gloffset_SecondaryColor3iEXT, SecondaryColor3iEXT@12) + GL_STUB(SecondaryColor3ivEXT, _gloffset_SecondaryColor3ivEXT, SecondaryColor3ivEXT@4) + GL_STUB(SecondaryColor3sEXT, _gloffset_SecondaryColor3sEXT, SecondaryColor3sEXT@12) + GL_STUB(SecondaryColor3svEXT, _gloffset_SecondaryColor3svEXT, SecondaryColor3svEXT@4) + GL_STUB(SecondaryColor3ubEXT, _gloffset_SecondaryColor3ubEXT, SecondaryColor3ubEXT@12) + GL_STUB(SecondaryColor3ubvEXT, _gloffset_SecondaryColor3ubvEXT, SecondaryColor3ubvEXT@4) + GL_STUB(SecondaryColor3uiEXT, _gloffset_SecondaryColor3uiEXT, SecondaryColor3uiEXT@12) + GL_STUB(SecondaryColor3uivEXT, _gloffset_SecondaryColor3uivEXT, SecondaryColor3uivEXT@4) + GL_STUB(SecondaryColor3usEXT, _gloffset_SecondaryColor3usEXT, SecondaryColor3usEXT@12) + GL_STUB(SecondaryColor3usvEXT, _gloffset_SecondaryColor3usvEXT, SecondaryColor3usvEXT@4) + GL_STUB(SecondaryColorPointerEXT, _gloffset_SecondaryColorPointerEXT, SecondaryColorPointerEXT@16) + GL_STUB(AreProgramsResidentNV, _gloffset_AreProgramsResidentNV, AreProgramsResidentNV@12) + GL_STUB(BindProgramNV, _gloffset_BindProgramNV, BindProgramNV@8) + GL_STUB(DeleteProgramsNV, _gloffset_DeleteProgramsNV, DeleteProgramsNV@8) + GL_STUB(ExecuteProgramNV, _gloffset_ExecuteProgramNV, ExecuteProgramNV@12) + GL_STUB(GenProgramsNV, _gloffset_GenProgramsNV, GenProgramsNV@8) + GL_STUB(GetProgramParameterdvNV, _gloffset_GetProgramParameterdvNV, GetProgramParameterdvNV@16) + GL_STUB(GetProgramParameterfvNV, _gloffset_GetProgramParameterfvNV, GetProgramParameterfvNV@16) + GL_STUB(GetProgramivNV, _gloffset_GetProgramivNV, GetProgramivNV@12) + GL_STUB(GetProgramStringNV, _gloffset_GetProgramStringNV, GetProgramStringNV@12) + GL_STUB(GetTrackMatrixivNV, _gloffset_GetTrackMatrixivNV, GetTrackMatrixivNV@16) + GL_STUB(GetVertexAttribdvNV, _gloffset_GetVertexAttribdvNV, GetVertexAttribdvNV@12) + GL_STUB(GetVertexAttribfvNV, _gloffset_GetVertexAttribfvNV, GetVertexAttribfvNV@12) + GL_STUB(GetVertexAttribivNV, _gloffset_GetVertexAttribivNV, GetVertexAttribivNV@12) + GL_STUB(GetVertexAttribPointervNV, _gloffset_GetVertexAttribPointervNV, GetVertexAttribPointervNV@12) + GL_STUB(IsProgramNV, _gloffset_IsProgramNV, IsProgramNV@4) + GL_STUB(LoadProgramNV, _gloffset_LoadProgramNV, LoadProgramNV@16) + GL_STUB(ProgramParameter4dNV, _gloffset_ProgramParameter4dNV, ProgramParameter4dNV@40) + GL_STUB(ProgramParameter4dvNV, _gloffset_ProgramParameter4dvNV, ProgramParameter4dvNV@12) + GL_STUB(ProgramParameter4fNV, _gloffset_ProgramParameter4fNV, ProgramParameter4fNV@24) + GL_STUB(ProgramParameter4fvNV, _gloffset_ProgramParameter4fvNV, ProgramParameter4fvNV@12) + GL_STUB(ProgramParameters4dvNV, _gloffset_ProgramParameters4dvNV, ProgramParameters4dvNV@16) + GL_STUB(ProgramParameters4fvNV, _gloffset_ProgramParameters4fvNV, ProgramParameters4fvNV@16) + GL_STUB(RequestResidentProgramsNV, _gloffset_RequestResidentProgramsNV, RequestResidentProgramsNV@8) + GL_STUB(TrackMatrixNV, _gloffset_TrackMatrixNV, TrackMatrixNV@16) + GL_STUB(VertexAttribPointerNV, _gloffset_VertexAttribPointerNV, VertexAttribPointerNV@20) + GL_STUB(VertexAttrib1dNV, _gloffset_VertexAttrib1dNV, VertexAttrib1dNV@12) + GL_STUB(VertexAttrib1dvNV, _gloffset_VertexAttrib1dvNV, VertexAttrib1dvNV@8) + GL_STUB(VertexAttrib1fNV, _gloffset_VertexAttrib1fNV, VertexAttrib1fNV@8) + GL_STUB(VertexAttrib1fvNV, _gloffset_VertexAttrib1fvNV, VertexAttrib1fvNV@8) + GL_STUB(VertexAttrib1sNV, _gloffset_VertexAttrib1sNV, VertexAttrib1sNV@8) + GL_STUB(VertexAttrib1svNV, _gloffset_VertexAttrib1svNV, VertexAttrib1svNV@8) + GL_STUB(VertexAttrib2dNV, _gloffset_VertexAttrib2dNV, VertexAttrib2dNV@20) + GL_STUB(VertexAttrib2dvNV, _gloffset_VertexAttrib2dvNV, VertexAttrib2dvNV@8) + GL_STUB(VertexAttrib2fNV, _gloffset_VertexAttrib2fNV, VertexAttrib2fNV@12) + GL_STUB(VertexAttrib2fvNV, _gloffset_VertexAttrib2fvNV, VertexAttrib2fvNV@8) + GL_STUB(VertexAttrib2sNV, _gloffset_VertexAttrib2sNV, VertexAttrib2sNV@12) + GL_STUB(VertexAttrib2svNV, _gloffset_VertexAttrib2svNV, VertexAttrib2svNV@8) + GL_STUB(VertexAttrib3dNV, _gloffset_VertexAttrib3dNV, VertexAttrib3dNV@28) + GL_STUB(VertexAttrib3dvNV, _gloffset_VertexAttrib3dvNV, VertexAttrib3dvNV@8) + GL_STUB(VertexAttrib3fNV, _gloffset_VertexAttrib3fNV, VertexAttrib3fNV@16) + GL_STUB(VertexAttrib3fvNV, _gloffset_VertexAttrib3fvNV, VertexAttrib3fvNV@8) + GL_STUB(VertexAttrib3sNV, _gloffset_VertexAttrib3sNV, VertexAttrib3sNV@16) + GL_STUB(VertexAttrib3svNV, _gloffset_VertexAttrib3svNV, VertexAttrib3svNV@8) + GL_STUB(VertexAttrib4dNV, _gloffset_VertexAttrib4dNV, VertexAttrib4dNV@36) + GL_STUB(VertexAttrib4dvNV, _gloffset_VertexAttrib4dvNV, VertexAttrib4dvNV@8) + GL_STUB(VertexAttrib4fNV, _gloffset_VertexAttrib4fNV, VertexAttrib4fNV@20) + GL_STUB(VertexAttrib4fvNV, _gloffset_VertexAttrib4fvNV, VertexAttrib4fvNV@8) + GL_STUB(VertexAttrib4sNV, _gloffset_VertexAttrib4sNV, VertexAttrib4sNV@20) + GL_STUB(VertexAttrib4svNV, _gloffset_VertexAttrib4svNV, VertexAttrib4svNV@8) + GL_STUB(VertexAttrib4ubNV, _gloffset_VertexAttrib4ubNV, VertexAttrib4ubNV@20) + GL_STUB(VertexAttrib4ubvNV, _gloffset_VertexAttrib4ubvNV, VertexAttrib4ubvNV@8) + GL_STUB(VertexAttribs1dvNV, _gloffset_VertexAttribs1dvNV, VertexAttribs1dvNV@12) + GL_STUB(VertexAttribs1fvNV, _gloffset_VertexAttribs1fvNV, VertexAttribs1fvNV@12) + GL_STUB(VertexAttribs1svNV, _gloffset_VertexAttribs1svNV, VertexAttribs1svNV@12) + GL_STUB(VertexAttribs2dvNV, _gloffset_VertexAttribs2dvNV, VertexAttribs2dvNV@12) + GL_STUB(VertexAttribs2fvNV, _gloffset_VertexAttribs2fvNV, VertexAttribs2fvNV@12) + GL_STUB(VertexAttribs2svNV, _gloffset_VertexAttribs2svNV, VertexAttribs2svNV@12) + GL_STUB(VertexAttribs3dvNV, _gloffset_VertexAttribs3dvNV, VertexAttribs3dvNV@12) + GL_STUB(VertexAttribs3fvNV, _gloffset_VertexAttribs3fvNV, VertexAttribs3fvNV@12) + GL_STUB(VertexAttribs3svNV, _gloffset_VertexAttribs3svNV, VertexAttribs3svNV@12) + GL_STUB(VertexAttribs4dvNV, _gloffset_VertexAttribs4dvNV, VertexAttribs4dvNV@12) + GL_STUB(VertexAttribs4fvNV, _gloffset_VertexAttribs4fvNV, VertexAttribs4fvNV@12) + GL_STUB(VertexAttribs4svNV, _gloffset_VertexAttribs4svNV, VertexAttribs4svNV@12) + GL_STUB(VertexAttribs4ubvNV, _gloffset_VertexAttribs4ubvNV, VertexAttribs4ubvNV@12) + GL_STUB(PointParameteriNV, _gloffset_PointParameteriNV, PointParameteriNV@8) + GL_STUB(PointParameterivNV, _gloffset_PointParameterivNV, PointParameterivNV@8) + GL_STUB(MultiDrawArraysEXT, _gloffset_MultiDrawArraysEXT, MultiDrawArraysEXT@16) + GL_STUB(MultiDrawElementsEXT, _gloffset_MultiDrawElementsEXT, MultiDrawElementsEXT@20) + GL_STUB(ActiveStencilFaceEXT, _gloffset_ActiveStencilFaceEXT, ActiveStencilFaceEXT@4) + GL_STUB(DeleteFencesNV, _gloffset_DeleteFencesNV, DeleteFencesNV@8) + GL_STUB(GenFencesNV, _gloffset_GenFencesNV, GenFencesNV@8) + GL_STUB(IsFenceNV, _gloffset_IsFenceNV, IsFenceNV@4) + GL_STUB(TestFenceNV, _gloffset_TestFenceNV, TestFenceNV@4) + GL_STUB(GetFenceivNV, _gloffset_GetFenceivNV, GetFenceivNV@12) + GL_STUB(FinishFenceNV, _gloffset_FinishFenceNV, FinishFenceNV@4) + GL_STUB(SetFenceNV, _gloffset_SetFenceNV, SetFenceNV@8) + GL_STUB(VertexAttrib4bvARB, _gloffset_VertexAttrib4bvARB, VertexAttrib4bvARB@8) + GL_STUB(VertexAttrib4ivARB, _gloffset_VertexAttrib4ivARB, VertexAttrib4ivARB@8) + GL_STUB(VertexAttrib4ubvARB, _gloffset_VertexAttrib4ubvARB, VertexAttrib4ubvARB@8) + GL_STUB(VertexAttrib4usvARB, _gloffset_VertexAttrib4usvARB, VertexAttrib4usvARB@8) + GL_STUB(VertexAttrib4uivARB, _gloffset_VertexAttrib4uivARB, VertexAttrib4uivARB@8) + GL_STUB(VertexAttrib4NbvARB, _gloffset_VertexAttrib4NbvARB, VertexAttrib4NbvARB@8) + GL_STUB(VertexAttrib4NsvARB, _gloffset_VertexAttrib4NsvARB, VertexAttrib4NsvARB@8) + GL_STUB(VertexAttrib4NivARB, _gloffset_VertexAttrib4NivARB, VertexAttrib4NivARB@8) + GL_STUB(VertexAttrib4NusvARB, _gloffset_VertexAttrib4NusvARB, VertexAttrib4NusvARB@8) + GL_STUB(VertexAttrib4NuivARB, _gloffset_VertexAttrib4NuivARB, VertexAttrib4NuivARB@8) + GL_STUB(VertexAttribPointerARB, _gloffset_VertexAttribPointerARB, VertexAttribPointerARB@24) + GL_STUB(EnableVertexAttribArrayARB, _gloffset_EnableVertexAttribArrayARB, EnableVertexAttribArrayARB@4) + GL_STUB(DisableVertexAttribArrayARB, _gloffset_DisableVertexAttribArrayARB, DisableVertexAttribArrayARB@4) + GL_STUB(ProgramStringARB, _gloffset_ProgramStringARB, ProgramStringARB@16) + GL_STUB(ProgramEnvParameter4dARB, _gloffset_ProgramEnvParameter4dARB, ProgramEnvParameter4dARB@40) + GL_STUB(ProgramEnvParameter4dvARB, _gloffset_ProgramEnvParameter4dvARB, ProgramEnvParameter4dvARB@12) + GL_STUB(ProgramEnvParameter4fARB, _gloffset_ProgramEnvParameter4fARB, ProgramEnvParameter4fARB@24) + GL_STUB(ProgramEnvParameter4fvARB, _gloffset_ProgramEnvParameter4fvARB, ProgramEnvParameter4fvARB@12) + GL_STUB(ProgramLocalParameter4dARB, _gloffset_ProgramLocalParameter4dARB, ProgramLocalParameter4dARB@40) + GL_STUB(ProgramLocalParameter4dvARB, _gloffset_ProgramLocalParameter4dvARB, ProgramLocalParameter4dvARB@12) + GL_STUB(ProgramLocalParameter4fARB, _gloffset_ProgramLocalParameter4fARB, ProgramLocalParameter4fARB@24) + GL_STUB(ProgramLocalParameter4fvARB, _gloffset_ProgramLocalParameter4fvARB, ProgramLocalParameter4fvARB@12) + GL_STUB(GetProgramEnvParameterdvARB, _gloffset_GetProgramEnvParameterdvARB, GetProgramEnvParameterdvARB@12) + GL_STUB(GetProgramEnvParameterfvARB, _gloffset_GetProgramEnvParameterfvARB, GetProgramEnvParameterfvARB@12) + GL_STUB(GetProgramLocalParameterdvARB, _gloffset_GetProgramLocalParameterdvARB, GetProgramLocalParameterdvARB@12) + GL_STUB(GetProgramLocalParameterfvARB, _gloffset_GetProgramLocalParameterfvARB, GetProgramLocalParameterfvARB@12) + GL_STUB(GetProgramivARB, _gloffset_GetProgramivARB, GetProgramivARB@12) + GL_STUB(GetProgramStringARB, _gloffset_GetProgramStringARB, GetProgramStringARB@12) + GL_STUB(ProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV, ProgramNamedParameter4fNV@28) + GL_STUB(ProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV, ProgramNamedParameter4dNV@44) + GL_STUB(ProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV, ProgramNamedParameter4fvNV@16) + GL_STUB(ProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV, ProgramNamedParameter4dvNV@16) + GL_STUB(GetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV, GetProgramNamedParameterfvNV@16) + GL_STUB(GetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV, GetProgramNamedParameterdvNV@16) + GL_STUB(BindBufferARB, _gloffset_BindBufferARB, BindBufferARB@8) + GL_STUB(BufferDataARB, _gloffset_BufferDataARB, BufferDataARB@16) + GL_STUB(BufferSubDataARB, _gloffset_BufferSubDataARB, BufferSubDataARB@16) + GL_STUB(DeleteBuffersARB, _gloffset_DeleteBuffersARB, DeleteBuffersARB@8) + GL_STUB(GenBuffersARB, _gloffset_GenBuffersARB, GenBuffersARB@8) + GL_STUB(GetBufferParameterivARB, _gloffset_GetBufferParameterivARB, GetBufferParameterivARB@12) + GL_STUB(GetBufferPointervARB, _gloffset_GetBufferPointervARB, GetBufferPointervARB@12) + GL_STUB(GetBufferSubDataARB, _gloffset_GetBufferSubDataARB, GetBufferSubDataARB@16) + GL_STUB(IsBufferARB, _gloffset_IsBufferARB, IsBufferARB@4) + GL_STUB(MapBufferARB, _gloffset_MapBufferARB, MapBufferARB@8) + GL_STUB(UnmapBufferARB, _gloffset_UnmapBufferARB, UnmapBufferARB@4) + GL_STUB(DepthBoundsEXT, _gloffset_DepthBoundsEXT, DepthBoundsEXT@16) + GL_STUB(GenQueriesARB, _gloffset_GenQueriesARB, GenQueriesARB@8) + GL_STUB(DeleteQueriesARB, _gloffset_DeleteQueriesARB, DeleteQueriesARB@8) + GL_STUB(IsQueryARB, _gloffset_IsQueryARB, IsQueryARB@4) + GL_STUB(BeginQueryARB, _gloffset_BeginQueryARB, BeginQueryARB@8) + GL_STUB(EndQueryARB, _gloffset_EndQueryARB, EndQueryARB@4) + GL_STUB(GetQueryivARB, _gloffset_GetQueryivARB, GetQueryivARB@12) + GL_STUB(GetQueryObjectivARB, _gloffset_GetQueryObjectivARB, GetQueryObjectivARB@12) + GL_STUB(GetQueryObjectuivARB, _gloffset_GetQueryObjectuivARB, GetQueryObjectuivARB@12) + GL_STUB(MultiModeDrawArraysIBM, _gloffset_MultiModeDrawArraysIBM, MultiModeDrawArraysIBM@20) + GL_STUB(MultiModeDrawElementsIBM, _gloffset_MultiModeDrawElementsIBM, MultiModeDrawElementsIBM@24) + GL_STUB(BlendEquationSeparateEXT, _gloffset_BlendEquationSeparateEXT, BlendEquationSeparateEXT@8) + GL_STUB(ActiveTexture, _gloffset_ActiveTextureARB, ActiveTexture@4) + GL_STUB(ClientActiveTexture, _gloffset_ClientActiveTextureARB, ClientActiveTexture@4) + GL_STUB(MultiTexCoord1d, _gloffset_MultiTexCoord1dARB, MultiTexCoord1d@12) + GL_STUB(MultiTexCoord1dv, _gloffset_MultiTexCoord1dvARB, MultiTexCoord1dv@8) + GL_STUB(MultiTexCoord1f, _gloffset_MultiTexCoord1fARB, MultiTexCoord1f@8) + GL_STUB(MultiTexCoord1fv, _gloffset_MultiTexCoord1fvARB, MultiTexCoord1fv@8) + GL_STUB(MultiTexCoord1i, _gloffset_MultiTexCoord1iARB, MultiTexCoord1i@8) + GL_STUB(MultiTexCoord1iv, _gloffset_MultiTexCoord1ivARB, MultiTexCoord1iv@8) + GL_STUB(MultiTexCoord1s, _gloffset_MultiTexCoord1sARB, MultiTexCoord1s@8) + GL_STUB(MultiTexCoord1sv, _gloffset_MultiTexCoord1svARB, MultiTexCoord1sv@8) + GL_STUB(MultiTexCoord2d, _gloffset_MultiTexCoord2dARB, MultiTexCoord2d@20) + GL_STUB(MultiTexCoord2dv, _gloffset_MultiTexCoord2dvARB, MultiTexCoord2dv@8) + GL_STUB(MultiTexCoord2f, _gloffset_MultiTexCoord2fARB, MultiTexCoord2f@12) + GL_STUB(MultiTexCoord2fv, _gloffset_MultiTexCoord2fvARB, MultiTexCoord2fv@8) + GL_STUB(MultiTexCoord2i, _gloffset_MultiTexCoord2iARB, MultiTexCoord2i@12) + GL_STUB(MultiTexCoord2iv, _gloffset_MultiTexCoord2ivARB, MultiTexCoord2iv@8) + GL_STUB(MultiTexCoord2s, _gloffset_MultiTexCoord2sARB, MultiTexCoord2s@12) + GL_STUB(MultiTexCoord2sv, _gloffset_MultiTexCoord2svARB, MultiTexCoord2sv@8) + GL_STUB(MultiTexCoord3d, _gloffset_MultiTexCoord3dARB, MultiTexCoord3d@28) + GL_STUB(MultiTexCoord3dv, _gloffset_MultiTexCoord3dvARB, MultiTexCoord3dv@8) + GL_STUB(MultiTexCoord3f, _gloffset_MultiTexCoord3fARB, MultiTexCoord3f@16) + GL_STUB(MultiTexCoord3fv, _gloffset_MultiTexCoord3fvARB, MultiTexCoord3fv@8) + GL_STUB(MultiTexCoord3i, _gloffset_MultiTexCoord3iARB, MultiTexCoord3i@16) + GL_STUB(MultiTexCoord3iv, _gloffset_MultiTexCoord3ivARB, MultiTexCoord3iv@8) + GL_STUB(MultiTexCoord3s, _gloffset_MultiTexCoord3sARB, MultiTexCoord3s@16) + GL_STUB(MultiTexCoord3sv, _gloffset_MultiTexCoord3svARB, MultiTexCoord3sv@8) + GL_STUB(MultiTexCoord4d, _gloffset_MultiTexCoord4dARB, MultiTexCoord4d@36) + GL_STUB(MultiTexCoord4dv, _gloffset_MultiTexCoord4dvARB, MultiTexCoord4dv@8) + GL_STUB(MultiTexCoord4f, _gloffset_MultiTexCoord4fARB, MultiTexCoord4f@20) + GL_STUB(MultiTexCoord4fv, _gloffset_MultiTexCoord4fvARB, MultiTexCoord4fv@8) + GL_STUB(MultiTexCoord4i, _gloffset_MultiTexCoord4iARB, MultiTexCoord4i@20) + GL_STUB(MultiTexCoord4iv, _gloffset_MultiTexCoord4ivARB, MultiTexCoord4iv@8) + GL_STUB(MultiTexCoord4s, _gloffset_MultiTexCoord4sARB, MultiTexCoord4s@20) + GL_STUB(MultiTexCoord4sv, _gloffset_MultiTexCoord4svARB, MultiTexCoord4sv@8) + GL_STUB(LoadTransposeMatrixf, _gloffset_LoadTransposeMatrixfARB, LoadTransposeMatrixf@4) + GL_STUB(LoadTransposeMatrixd, _gloffset_LoadTransposeMatrixdARB, LoadTransposeMatrixd@4) + GL_STUB(MultTransposeMatrixf, _gloffset_MultTransposeMatrixfARB, MultTransposeMatrixf@4) + GL_STUB(MultTransposeMatrixd, _gloffset_MultTransposeMatrixdARB, MultTransposeMatrixd@4) + GL_STUB(SampleCoverage, _gloffset_SampleCoverageARB, SampleCoverage@8) + GL_STUB(CompressedTexImage3D, _gloffset_CompressedTexImage3DARB, CompressedTexImage3D@36) + GL_STUB(CompressedTexImage2D, _gloffset_CompressedTexImage2DARB, CompressedTexImage2D@32) + GL_STUB(CompressedTexImage1D, _gloffset_CompressedTexImage1DARB, CompressedTexImage1D@28) + GL_STUB(CompressedTexSubImage3D, _gloffset_CompressedTexSubImage3DARB, CompressedTexSubImage3D@44) + GL_STUB(CompressedTexSubImage2D, _gloffset_CompressedTexSubImage2DARB, CompressedTexSubImage2D@36) + GL_STUB(CompressedTexSubImage1D, _gloffset_CompressedTexSubImage1DARB, CompressedTexSubImage1D@28) + GL_STUB(GetCompressedTexImage, _gloffset_GetCompressedTexImageARB, GetCompressedTexImage@12) + GL_STUB(BlendFuncSeparate, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparate@16) + GL_STUB(FogCoordf, _gloffset_FogCoordfEXT, FogCoordf@4) + GL_STUB(FogCoordfv, _gloffset_FogCoordfvEXT, FogCoordfv@4) + GL_STUB(FogCoordd, _gloffset_FogCoorddEXT, FogCoordd@8) + GL_STUB(FogCoorddv, _gloffset_FogCoorddvEXT, FogCoorddv@4) + GL_STUB(FogCoordPointer, _gloffset_FogCoordPointerEXT, FogCoordPointer@12) + GL_STUB(MultiDrawArrays, _gloffset_MultiDrawArraysEXT, MultiDrawArrays@16) + GL_STUB(MultiDrawElements, _gloffset_MultiDrawElementsEXT, MultiDrawElements@20) + GL_STUB(PointParameterf, _gloffset_PointParameterfEXT, PointParameterf@8) + GL_STUB(PointParameterfv, _gloffset_PointParameterfvEXT, PointParameterfv@8) + GL_STUB(PointParameteri, _gloffset_PointParameteriNV, PointParameteri@8) + GL_STUB(PointParameteriv, _gloffset_PointParameterivNV, PointParameteriv@8) + GL_STUB(SecondaryColor3b, _gloffset_SecondaryColor3bEXT, SecondaryColor3b@12) + GL_STUB(SecondaryColor3bv, _gloffset_SecondaryColor3bvEXT, SecondaryColor3bv@4) + GL_STUB(SecondaryColor3d, _gloffset_SecondaryColor3dEXT, SecondaryColor3d@24) + GL_STUB(SecondaryColor3dv, _gloffset_SecondaryColor3dvEXT, SecondaryColor3dv@4) + GL_STUB(SecondaryColor3f, _gloffset_SecondaryColor3fEXT, SecondaryColor3f@12) + GL_STUB(SecondaryColor3fv, _gloffset_SecondaryColor3fvEXT, SecondaryColor3fv@4) + GL_STUB(SecondaryColor3i, _gloffset_SecondaryColor3iEXT, SecondaryColor3i@12) + GL_STUB(SecondaryColor3iv, _gloffset_SecondaryColor3ivEXT, SecondaryColor3iv@4) + GL_STUB(SecondaryColor3s, _gloffset_SecondaryColor3sEXT, SecondaryColor3s@12) + GL_STUB(SecondaryColor3sv, _gloffset_SecondaryColor3svEXT, SecondaryColor3sv@4) + GL_STUB(SecondaryColor3ub, _gloffset_SecondaryColor3ubEXT, SecondaryColor3ub@12) + GL_STUB(SecondaryColor3ubv, _gloffset_SecondaryColor3ubvEXT, SecondaryColor3ubv@4) + GL_STUB(SecondaryColor3ui, _gloffset_SecondaryColor3uiEXT, SecondaryColor3ui@12) + GL_STUB(SecondaryColor3uiv, _gloffset_SecondaryColor3uivEXT, SecondaryColor3uiv@4) + GL_STUB(SecondaryColor3us, _gloffset_SecondaryColor3usEXT, SecondaryColor3us@12) + GL_STUB(SecondaryColor3usv, _gloffset_SecondaryColor3usvEXT, SecondaryColor3usv@4) + GL_STUB(SecondaryColorPointer, _gloffset_SecondaryColorPointerEXT, SecondaryColorPointer@16) + GL_STUB(WindowPos2d, _gloffset_WindowPos2dMESA, WindowPos2d@16) + GL_STUB(WindowPos2dv, _gloffset_WindowPos2dvMESA, WindowPos2dv@4) + GL_STUB(WindowPos2f, _gloffset_WindowPos2fMESA, WindowPos2f@8) + GL_STUB(WindowPos2fv, _gloffset_WindowPos2fvMESA, WindowPos2fv@4) + GL_STUB(WindowPos2i, _gloffset_WindowPos2iMESA, WindowPos2i@8) + GL_STUB(WindowPos2iv, _gloffset_WindowPos2ivMESA, WindowPos2iv@4) + GL_STUB(WindowPos2s, _gloffset_WindowPos2sMESA, WindowPos2s@8) + GL_STUB(WindowPos2sv, _gloffset_WindowPos2svMESA, WindowPos2sv@4) + GL_STUB(WindowPos3d, _gloffset_WindowPos3dMESA, WindowPos3d@24) + GL_STUB(WindowPos3dv, _gloffset_WindowPos3dvMESA, WindowPos3dv@4) + GL_STUB(WindowPos3f, _gloffset_WindowPos3fMESA, WindowPos3f@12) + GL_STUB(WindowPos3fv, _gloffset_WindowPos3fvMESA, WindowPos3fv@4) + GL_STUB(WindowPos3i, _gloffset_WindowPos3iMESA, WindowPos3i@12) + GL_STUB(WindowPos3iv, _gloffset_WindowPos3ivMESA, WindowPos3iv@4) + GL_STUB(WindowPos3s, _gloffset_WindowPos3sMESA, WindowPos3s@12) + GL_STUB(WindowPos3sv, _gloffset_WindowPos3svMESA, WindowPos3sv@4) + GL_STUB(BindBuffer, _gloffset_BindBufferARB, BindBuffer@8) + GL_STUB(BufferData, _gloffset_BufferDataARB, BufferData@16) + GL_STUB(BufferSubData, _gloffset_BufferSubDataARB, BufferSubData@16) + GL_STUB(DeleteBuffers, _gloffset_DeleteBuffersARB, DeleteBuffers@8) + GL_STUB(GenBuffers, _gloffset_GenBuffersARB, GenBuffers@8) + GL_STUB(GetBufferParameteriv, _gloffset_GetBufferParameterivARB, GetBufferParameteriv@12) + GL_STUB(GetBufferPointerv, _gloffset_GetBufferPointervARB, GetBufferPointerv@12) + GL_STUB(GetBufferSubData, _gloffset_GetBufferSubDataARB, GetBufferSubData@16) + GL_STUB(IsBuffer, _gloffset_IsBufferARB, IsBuffer@4) + GL_STUB(MapBuffer, _gloffset_MapBufferARB, MapBuffer@8) + GL_STUB(UnmapBuffer, _gloffset_UnmapBufferARB, UnmapBuffer@4) + GL_STUB(GenQueries, _gloffset_GenQueriesARB, GenQueries@8) + GL_STUB(DeleteQueries, _gloffset_DeleteQueriesARB, DeleteQueries@8) + GL_STUB(IsQuery, _gloffset_IsQueryARB, IsQuery@4) + GL_STUB(BeginQuery, _gloffset_BeginQueryARB, BeginQuery@8) + GL_STUB(EndQuery, _gloffset_EndQueryARB, EndQuery@4) + GL_STUB(GetQueryiv, _gloffset_GetQueryivARB, GetQueryiv@12) + GL_STUB(GetQueryObjectiv, _gloffset_GetQueryObjectivARB, GetQueryObjectiv@12) + GL_STUB(GetQueryObjectuiv, _gloffset_GetQueryObjectuivARB, GetQueryObjectuiv@12) + GL_STUB(PointParameterfARB, _gloffset_PointParameterfEXT, PointParameterfARB@8) + GL_STUB(PointParameterfvARB, _gloffset_PointParameterfvEXT, PointParameterfvARB@8) + GL_STUB(WindowPos2dARB, _gloffset_WindowPos2dMESA, WindowPos2dARB@16) + GL_STUB(WindowPos2fARB, _gloffset_WindowPos2fMESA, WindowPos2fARB@8) + GL_STUB(WindowPos2iARB, _gloffset_WindowPos2iMESA, WindowPos2iARB@8) + GL_STUB(WindowPos2sARB, _gloffset_WindowPos2sMESA, WindowPos2sARB@8) + GL_STUB(WindowPos2dvARB, _gloffset_WindowPos2dvMESA, WindowPos2dvARB@4) + GL_STUB(WindowPos2fvARB, _gloffset_WindowPos2fvMESA, WindowPos2fvARB@4) + GL_STUB(WindowPos2ivARB, _gloffset_WindowPos2ivMESA, WindowPos2ivARB@4) + GL_STUB(WindowPos2svARB, _gloffset_WindowPos2svMESA, WindowPos2svARB@4) + GL_STUB(WindowPos3dARB, _gloffset_WindowPos3dMESA, WindowPos3dARB@24) + GL_STUB(WindowPos3fARB, _gloffset_WindowPos3fMESA, WindowPos3fARB@12) + GL_STUB(WindowPos3iARB, _gloffset_WindowPos3iMESA, WindowPos3iARB@12) + GL_STUB(WindowPos3sARB, _gloffset_WindowPos3sMESA, WindowPos3sARB@12) + GL_STUB(WindowPos3dvARB, _gloffset_WindowPos3dvMESA, WindowPos3dvARB@4) + GL_STUB(WindowPos3fvARB, _gloffset_WindowPos3fvMESA, WindowPos3fvARB@4) + GL_STUB(WindowPos3ivARB, _gloffset_WindowPos3ivMESA, WindowPos3ivARB@4) + GL_STUB(WindowPos3svARB, _gloffset_WindowPos3svMESA, WindowPos3svARB@4) + GL_STUB(VertexAttrib1sARB, _gloffset_VertexAttrib1sNV, VertexAttrib1sARB@8) + GL_STUB(VertexAttrib1fARB, _gloffset_VertexAttrib1fNV, VertexAttrib1fARB@8) + GL_STUB(VertexAttrib1dARB, _gloffset_VertexAttrib1dNV, VertexAttrib1dARB@12) + GL_STUB(VertexAttrib2sARB, _gloffset_VertexAttrib2sNV, VertexAttrib2sARB@12) + GL_STUB(VertexAttrib2fARB, _gloffset_VertexAttrib2fNV, VertexAttrib2fARB@12) + GL_STUB(VertexAttrib2dARB, _gloffset_VertexAttrib2dNV, VertexAttrib2dARB@20) + GL_STUB(VertexAttrib3sARB, _gloffset_VertexAttrib3sNV, VertexAttrib3sARB@16) + GL_STUB(VertexAttrib3fARB, _gloffset_VertexAttrib3fNV, VertexAttrib3fARB@16) + GL_STUB(VertexAttrib3dARB, _gloffset_VertexAttrib3dNV, VertexAttrib3dARB@28) + GL_STUB(VertexAttrib4sARB, _gloffset_VertexAttrib4sNV, VertexAttrib4sARB@20) + GL_STUB(VertexAttrib4fARB, _gloffset_VertexAttrib4fNV, VertexAttrib4fARB@20) + GL_STUB(VertexAttrib4dARB, _gloffset_VertexAttrib4dNV, VertexAttrib4dARB@36) + GL_STUB(VertexAttrib4NubARB, _gloffset_VertexAttrib4ubNV, VertexAttrib4NubARB@20) + GL_STUB(VertexAttrib1svARB, _gloffset_VertexAttrib1svNV, VertexAttrib1svARB@8) + GL_STUB(VertexAttrib1fvARB, _gloffset_VertexAttrib1fvNV, VertexAttrib1fvARB@8) + GL_STUB(VertexAttrib1dvARB, _gloffset_VertexAttrib1dvNV, VertexAttrib1dvARB@8) + GL_STUB(VertexAttrib2svARB, _gloffset_VertexAttrib2svNV, VertexAttrib2svARB@8) + GL_STUB(VertexAttrib2fvARB, _gloffset_VertexAttrib2fvNV, VertexAttrib2fvARB@8) + GL_STUB(VertexAttrib2dvARB, _gloffset_VertexAttrib2dvNV, VertexAttrib2dvARB@8) + GL_STUB(VertexAttrib3svARB, _gloffset_VertexAttrib3svNV, VertexAttrib3svARB@8) + GL_STUB(VertexAttrib3fvARB, _gloffset_VertexAttrib3fvNV, VertexAttrib3fvARB@8) + GL_STUB(VertexAttrib3dvARB, _gloffset_VertexAttrib3dvNV, VertexAttrib3dvARB@8) + GL_STUB(VertexAttrib4svARB, _gloffset_VertexAttrib4svNV, VertexAttrib4svARB@8) + GL_STUB(VertexAttrib4fvARB, _gloffset_VertexAttrib4fvNV, VertexAttrib4fvARB@8) + GL_STUB(VertexAttrib4dvARB, _gloffset_VertexAttrib4dvNV, VertexAttrib4dvARB@8) + GL_STUB(VertexAttrib4NubvARB, _gloffset_VertexAttrib4ubvNV, VertexAttrib4NubvARB@8) + GL_STUB(BindProgramARB, _gloffset_BindProgramNV, BindProgramARB@8) + GL_STUB(DeleteProgramsARB, _gloffset_DeleteProgramsNV, DeleteProgramsARB@8) + GL_STUB(GenProgramsARB, _gloffset_GenProgramsNV, GenProgramsARB@8) + GL_STUB(IsProgramARB, _gloffset_IsProgramNV, IsProgramARB@4) + GL_STUB(GetVertexAttribdvARB, _gloffset_GetVertexAttribdvNV, GetVertexAttribdvARB@12) + GL_STUB(GetVertexAttribfvARB, _gloffset_GetVertexAttribfvNV, GetVertexAttribfvARB@12) + GL_STUB(GetVertexAttribivARB, _gloffset_GetVertexAttribivNV, GetVertexAttribivARB@12) + GL_STUB(GetVertexAttribPointervARB, _gloffset_GetVertexAttribPointervNV, GetVertexAttribPointervARB@12) + GL_STUB(BlendColorEXT, _gloffset_BlendColor, BlendColorEXT@16) + GL_STUB(TexImage3DEXT, _gloffset_TexImage3D, TexImage3DEXT@40) + GL_STUB(TexSubImage3DEXT, _gloffset_TexSubImage3D, TexSubImage3DEXT@44) + GL_STUB(TexSubImage1DEXT, _gloffset_TexSubImage1D, TexSubImage1DEXT@28) + GL_STUB(TexSubImage2DEXT, _gloffset_TexSubImage2D, TexSubImage2DEXT@36) + GL_STUB(CopyTexImage1DEXT, _gloffset_CopyTexImage1D, CopyTexImage1DEXT@28) + GL_STUB(CopyTexImage2DEXT, _gloffset_CopyTexImage2D, CopyTexImage2DEXT@32) + GL_STUB(CopyTexSubImage1DEXT, _gloffset_CopyTexSubImage1D, CopyTexSubImage1DEXT@24) + GL_STUB(CopyTexSubImage2DEXT, _gloffset_CopyTexSubImage2D, CopyTexSubImage2DEXT@32) + GL_STUB(CopyTexSubImage3DEXT, _gloffset_CopyTexSubImage3D, CopyTexSubImage3DEXT@36) + GL_STUB(HistogramEXT, _gloffset_Histogram, HistogramEXT@16) + GL_STUB(MinmaxEXT, _gloffset_Minmax, MinmaxEXT@12) + GL_STUB(ResetHistogramEXT, _gloffset_ResetHistogram, ResetHistogramEXT@4) + GL_STUB(ResetMinmaxEXT, _gloffset_ResetMinmax, ResetMinmaxEXT@4) + GL_STUB(ConvolutionFilter1DEXT, _gloffset_ConvolutionFilter1D, ConvolutionFilter1DEXT@24) + GL_STUB(ConvolutionFilter2DEXT, _gloffset_ConvolutionFilter2D, ConvolutionFilter2DEXT@28) + GL_STUB(ConvolutionParameterfEXT, _gloffset_ConvolutionParameterf, ConvolutionParameterfEXT@12) + GL_STUB(ConvolutionParameterfvEXT, _gloffset_ConvolutionParameterfv, ConvolutionParameterfvEXT@12) + GL_STUB(ConvolutionParameteriEXT, _gloffset_ConvolutionParameteri, ConvolutionParameteriEXT@12) + GL_STUB(ConvolutionParameterivEXT, _gloffset_ConvolutionParameteriv, ConvolutionParameterivEXT@12) + GL_STUB(CopyConvolutionFilter1DEXT, _gloffset_CopyConvolutionFilter1D, CopyConvolutionFilter1DEXT@20) + GL_STUB(CopyConvolutionFilter2DEXT, _gloffset_CopyConvolutionFilter2D, CopyConvolutionFilter2DEXT@24) + GL_STUB(SeparableFilter2DEXT, _gloffset_SeparableFilter2D, SeparableFilter2DEXT@32) + GL_STUB(ColorTableSGI, _gloffset_ColorTable, ColorTableSGI@24) + GL_STUB(ColorTableParameterfvSGI, _gloffset_ColorTableParameterfv, ColorTableParameterfvSGI@12) + GL_STUB(ColorTableParameterivSGI, _gloffset_ColorTableParameteriv, ColorTableParameterivSGI@12) + GL_STUB(CopyColorTableSGI, _gloffset_CopyColorTable, CopyColorTableSGI@20) + GL_STUB(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8) + GL_STUB(DeleteTexturesEXT, _gloffset_DeleteTextures, DeleteTexturesEXT@8) + GL_STUB(PrioritizeTexturesEXT, _gloffset_PrioritizeTextures, PrioritizeTexturesEXT@12) + GL_STUB(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4) + GL_STUB(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12) + GL_STUB(GetPointervEXT, _gloffset_GetPointerv, GetPointervEXT@8) + GL_STUB(BlendEquationEXT, _gloffset_BlendEquation, BlendEquationEXT@4) + GL_STUB(ColorSubTableEXT, _gloffset_ColorSubTable, ColorSubTableEXT@24) + GL_STUB(CopyColorSubTableEXT, _gloffset_CopyColorSubTable, CopyColorSubTableEXT@20) + GL_STUB(ColorTableEXT, _gloffset_ColorTable, ColorTableEXT@24) + GL_STUB(DrawRangeElementsEXT, _gloffset_DrawRangeElements, DrawRangeElementsEXT@24) + GL_STUB(SampleMaskEXT, _gloffset_SampleMaskSGIS, SampleMaskEXT@8) + GL_STUB(SamplePatternEXT, _gloffset_SamplePatternSGIS, SamplePatternEXT@4) + GL_STUB(BlendEquationSeparateATI, _gloffset_BlendEquationSeparateEXT, BlendEquationSeparateATI@8) + GL_STUB(BlendFuncSeparateINGR, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparateINGR@16) + GL_STUB(PointParameterfSGIS, _gloffset_PointParameterfEXT, PointParameterfSGIS@8) + GL_STUB(PointParameterfvSGIS, _gloffset_PointParameterfvEXT, PointParameterfvSGIS@8) + +#endif /* __WIN32__ */ diff --git a/src/mesa/x86/matypes.h b/src/mesa/x86/matypes.h new file mode 100644 index 0000000..06e0c34 --- /dev/null +++ b/src/mesa/x86/matypes.h @@ -0,0 +1,165 @@ +/* + * This file is automatically generated from the Mesa internal type + * definitions. Do not edit directly. + */ + +#ifndef __ASM_TYPES_H__ +#define __ASM_TYPES_H__ + +#include "assyntax.h" + + +/* ============================================================= + * Offsets for GLcontext + */ + +#define CTX_DRIVER_CTX 852 + +#define CTX_LIGHT_ENABLED 38212 +#define CTX_LIGHT_SHADE_MODEL 38216 +#define CTX_LIGHT_COLOR_MAT_FACE 38220 +#define CTX_LIGHT_COLOR_MAT_MODE 38224 +#define CTX_LIGHT_COLOR_MAT_MASK 38228 +#define CTX_LIGHT_COLOR_MAT_ENABLED 38232 +#define CTX_LIGHT_ENABLED_LIST 38236 +#define CTX_LIGHT_NEED_VERTS 42593 +#define CTX_LIGHT_FLAGS 42596 +#define CTX_LIGHT_BASE_COLOR 42600 + + +/* ============================================================= + * Offsets for struct vertex_buffer + */ + +#define VB_SIZE 0 +#define VB_COUNT 4 + +#define VB_ELTS 8 +#define VB_OBJ_PTR 12 +#define VB_EYE_PTR 16 +#define VB_CLIP_PTR 20 +#define VB_PROJ_CLIP_PTR 24 +#define VB_CLIP_OR_MASK 28 +#define VB_CLIP_MASK 32 +#define VB_NORMAL_PTR 36 +#define VB_EDGE_FLAG 44 +#define VB_TEX0_COORD_PTR 48 +#define VB_TEX1_COORD_PTR 52 +#define VB_TEX2_COORD_PTR 56 +#define VB_TEX3_COORD_PTR 60 +#define VB_INDEX_PTR 80 +#define VB_COLOR_PTR 88 +#define VB_SECONDARY_COLOR_PTR 96 +#define VB_FOG_COORD_PTR 108 +#define VB_POINT_SIZE_PTR 104 +#define VB_PRIMITIVE 112 + +#define VB_LAST_CLIPPED 244 + +/* + * Flags for struct vertex_buffer + */ + +#define VERT_BIT_OBJ 0x1 +#define VERT_BIT_NORM 0x4 +#define VERT_BIT_RGBA 0x8 +#define VERT_BIT_SPEC_RGB 0x10 +#define VERT_BIT_FOG_COORD 0x20 +#define VERT_BIT_TEX0 0x100 +#define VERT_BIT_TEX1 0x200 +#define VERT_BIT_TEX2 0x400 +#define VERT_BIT_TEX3 0x800 + + +/* ============================================================= + * Offsets for GLvector4f + */ + +#define V4F_DATA 0 +#define V4F_START 4 +#define V4F_COUNT 8 +#define V4F_STRIDE 12 +#define V4F_SIZE 16 +#define V4F_FLAGS 20 + +/* + * Flags for GLvector4f + */ + +#define VEC_MALLOC 0x10 +#define VEC_NOT_WRITEABLE 0x40 +#define VEC_BAD_STRIDE 0x100 + +#define VEC_SIZE_1 0x1 +#define VEC_SIZE_2 0x3 +#define VEC_SIZE_3 0x7 +#define VEC_SIZE_4 0xf + + +/* ============================================================= + * Offsets for GLmatrix + */ + +#define MATRIX_DATA 0 +#define MATRIX_INV 4 +#define MATRIX_FLAGS 8 +#define MATRIX_TYPE 12 + + +/* ============================================================= + * Offsets for struct gl_light + */ + +#define LIGHT_NEXT 0 +#define LIGHT_PREV 4 + +#define LIGHT_AMBIENT 8 +#define LIGHT_DIFFUSE 24 +#define LIGHT_SPECULAR 40 +#define LIGHT_EYE_POSITION 56 +#define LIGHT_EYE_DIRECTION 72 +#define LIGHT_SPOT_EXPONENT 88 +#define LIGHT_SPOT_CUTOFF 92 +#define LIGHT_COS_CUTOFF 96 +#define LIGHT_CONST_ATTEN 100 +#define LIGHT_LINEAR_ATTEN 104 +#define LIGHT_QUADRATIC_ATTEN 108 +#define LIGHT_ENABLED 112 + +#define LIGHT_FLAGS 116 + +#define LIGHT_POSITION 120 +#define LIGHT_VP_INF_NORM 136 +#define LIGHT_H_INF_NORM 148 +#define LIGHT_NORM_DIRECTION 160 +#define LIGHT_VP_INF_SPOT_ATTEN 176 + +#define LIGHT_SPOT_EXP_TABLE 180 +#define LIGHT_MAT_AMBIENT 4276 +#define LIGHT_MAT_DIFFUSE 4300 +#define LIGHT_MAT_SPECULAR 4324 + +#define SIZEOF_GL_LIGHT 4356 + +/* + * Flags for struct gl_light + */ + +#define LIGHT_SPOT 0x1 +#define LIGHT_LOCAL_VIEWER 0x2 +#define LIGHT_POSITIONAL 0x4 + +#define LIGHT_NEED_VERTICES 0x6 + + +/* ============================================================= + * Offsets for struct gl_lightmodel + */ + +#define LIGHT_MODEL_AMBIENT 0 +#define LIGHT_MODEL_LOCAL_VIEWER 16 +#define LIGHT_MODEL_TWO_SIDE 17 +#define LIGHT_MODEL_COLOR_CONTROL 20 + + +#endif /* __ASM_TYPES_H__ */ diff --git a/src/mesa/x86/mmx.h b/src/mesa/x86/mmx.h new file mode 100644 index 0000000..11f7863 --- /dev/null +++ b/src/mesa/x86/mmx.h @@ -0,0 +1,51 @@ +/* $Id: mmx.h,v 1.9 2002/04/19 20:12:30 jrfonseca Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef ASM_MMX_H +#define ASM_MMX_H + +extern void _ASMAPI +_mesa_mmx_blend_transparency( GLcontext *ctx, GLuint n, const GLubyte mask[], + GLubyte rgba[][4], const GLubyte dest[][4] ); + +extern void _ASMAPI +_mesa_mmx_blend_add( GLcontext *ctx, GLuint n, const GLubyte mask[], + GLubyte rgba[][4], const GLubyte dest[][4] ); + +extern void _ASMAPI +_mesa_mmx_blend_min( GLcontext *ctx, GLuint n, const GLubyte mask[], + GLubyte rgba[][4], const GLubyte dest[][4] ); + +extern void _ASMAPI +_mesa_mmx_blend_max( GLcontext *ctx, GLuint n, const GLubyte mask[], + GLubyte rgba[][4], const GLubyte dest[][4] ); + +extern void _ASMAPI +_mesa_mmx_blend_modulate( GLcontext *ctx, GLuint n, const GLubyte mask[], + GLubyte rgba[][4], const GLubyte dest[][4] ); + +#endif diff --git a/src/mesa/x86/mmx_blend.S b/src/mesa/x86/mmx_blend.S new file mode 100644 index 0000000..f7326cd --- /dev/null +++ b/src/mesa/x86/mmx_blend.S @@ -0,0 +1,366 @@ + +/* + * Written by José Fonseca <j_r_fonseca@yahoo.co.uk> + */ + + +#ifdef USE_MMX_ASM +#include "matypes.h" + +/* integer multiplication - alpha plus one + * + * makes the following approximation to the division (Sree) + * + * rgb*a/255 ~= (rgb*(a+1)) >> 256 + * + * which is the fastest method that satisfies the following OpenGL criteria + * + * 0*0 = 0 and 255*255 = 255 + * + * note that MX1 is a register with 0xffffffffffffffff constant which can be easily obtained making + * + * PCMPEQW ( MX1, MX1 ) + */ +#define GMB_MULT_AP1( MP1, MA1, MP2, MA2, MX1 ) \ + PSUBW ( MX1, MA1 ) /* a1 + 1 | a1 + 1 | a1 + 1 | a1 + 1 */ ;\ + PMULLW ( MP1, MA1 ) /* t1 = p1*a1 */ ;\ + ;\ +TWO(PSUBW ( MX1, MA2 )) /* a2 + 1 | a2 + 1 | a2 + 1 | a2 + 1 */ ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = p2*a2 */ ;\ + ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 ~= t1/255 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 ~= t2/255 */ + + +/* integer multiplication - geometric series + * + * takes the geometric series approximation to the division + * + * t/255 = (t >> 8) + (t >> 16) + (t >> 24) .. + * + * in this case just the first two terms to fit in 16bit arithmetic + * + * t/255 ~= (t + (t >> 8)) >> 8 + * + * note that just by itself it doesn't satisfies the OpenGL criteria, as 255*255 = 254, + * so the special case a = 255 must be accounted or roundoff must be used + */ +#define GMB_MULT_GS( MP1, MA1, MP2, MA2 ) \ + PMULLW ( MP1, MA1 ) /* t1 = p1*a1 */ ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = p2*a2 */ ;\ + ;\ + MOVQ ( MA1, MP1 ) ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 */ ;\ + ;\ +TWO(MOVQ ( MA2, MP2 )) ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 */ ;\ + ;\ + PADDW ( MP1, MA1 ) /* t1 + (t1 >> 8) ~= (t1/255) << 8 */ ;\ + PSRLW ( CONST(8), MA1 ) /* sa1 | sb1 | sg1 | sr1 */ ;\ + ;\ +TWO(PADDW ( MP2, MA2 )) /* t2 + (t2 >> 8) ~= (t2/255) << 8 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* sa2 | sb2 | sg2 | sr2 */ + + +/* integer multiplication - geometric series plus rounding + * + * when using a geometric series division instead of truncating the result + * use roundoff in the approximation (Jim Blinn) + * + * t = rgb*a + 0x80 + * + * achieving the exact results + * + * note that M80 is register with the 0x0080008000800080 constant + */ +#define GMB_MULT_GSR( MP1, MA1, MP2, MA2, M80 ) \ + PMULLW ( MP1, MA1 ) /* t1 = p1*a1 */ ;\ + PADDW ( M80, MA1 ) /* t1 += 0x80 */ ;\ + ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = p2*a2 */ ;\ +TWO(PADDW ( M80, MA2 )) /* t2 += 0x80 */ ;\ + ;\ + MOVQ ( MA1, MP1 ) ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 */ ;\ + ;\ +TWO(MOVQ ( MA2, MP2 )) ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 */ ;\ + ;\ + PADDW ( MP1, MA1 ) /* t1 + (t1 >> 8) ~= (t1/255) << 8 */ ;\ + PSRLW ( CONST(8), MA1 ) /* sa1 | sb1 | sg1 | sr1 */ ;\ + ;\ +TWO(PADDW ( MP2, MA2 )) /* t2 + (t2 >> 8) ~= (t2/255) << 8 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* sa2 | sb2 | sg2 | sr2 */ + + +/* linear interpolation - geometric series + */ +#define GMB_LERP_GS( MP1, MQ1, MA1, MP2, MQ2, MA2) \ + PSUBW ( MQ1, MP1 ) /* pa1 - qa1 | pb1 - qb1 | pg1 - qg1 | pr1 - qr1 */ ;\ + PSLLW ( CONST(8), MQ1 ) /* q1 << 8 */ ;\ + PMULLW ( MP1, MA1 ) /* t1 = (q1 - p1)*pa1 */ ;\ + ;\ +TWO(PSUBW ( MQ2, MP2 )) /* pa2 - qa2 | pb2 - qb2 | pg2 - qg2 | pr2 - qr2 */ ;\ +TWO(PSLLW ( CONST(8), MQ2 )) /* q2 << 8 */ ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = (q2 - p2)*pa2 */ ;\ + ;\ + MOVQ ( MA1, MP1 ) ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 */ ;\ + ;\ +TWO(MOVQ ( MA2, MP2 )) ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 */ ;\ + ;\ + PADDW ( MP1, MA1 ) /* t1 + (t1 >> 8) ~= (t1/255) << 8 */ ;\ +TWO(PADDW ( MP2, MA2 )) /* t2 + (t2 >> 8) ~= (t2/255) << 8 */ ;\ + ;\ + PADDW ( MQ1, MA1 ) /* (t1/255 + q1) << 8 */ ;\ +TWO(PADDW ( MQ2, MA2 )) /* (t2/255 + q2) << 8 */ ;\ + ;\ + PSRLW ( CONST(8), MA1 ) /* sa1 | sb1 | sg1 | sr1 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* sa2 | sb2 | sg2 | sr2 */ + + +/* linear interpolation - geometric series with roundoff + * + * this is a generalization of Blinn's formula to signed arithmetic + * + * note that M80 is a register with the 0x0080008000800080 constant + */ +#define GMB_LERP_GSR( MP1, MQ1, MA1, MP2, MQ2, MA2, M80) \ + PSUBW ( MQ1, MP1 ) /* pa1 - qa1 | pb1 - qb1 | pg1 - qg1 | pr1 - qr1 */ ;\ + PSLLW ( CONST(8), MQ1 ) /* q1 << 8 */ ;\ + PMULLW ( MP1, MA1 ) /* t1 = (q1 - p1)*pa1 */ ;\ + ;\ +TWO(PSUBW ( MQ2, MP2 )) /* pa2 - qa2 | pb2 - qb2 | pg2 - qg2 | pr2 - qr2 */ ;\ +TWO(PSLLW ( CONST(8), MQ2 )) /* q2 << 8 */ ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = (q2 - p2)*pa2 */ ;\ + ;\ + PSRLW ( CONST(15), MP1 ) /* q1 > p1 ? 1 : 0 */ ;\ +TWO(PSRLW ( CONST(15), MP2 )) /* q2 > q2 ? 1 : 0 */ ;\ + ;\ + PSLLW ( CONST(8), MP1 ) /* q1 > p1 ? 0x100 : 0 */ ;\ +TWO(PSLLW ( CONST(8), MP2 )) /* q2 > q2 ? 0x100 : 0 */ ;\ + ;\ + PSUBW ( MP1, MA1 ) /* t1 -=? 0x100 */ ;\ +TWO(PSUBW ( MP2, MA2 )) /* t2 -=? 0x100 */ ;\ + ;\ + PADDW ( M80, MA1 ) /* t1 += 0x80 */ ;\ +TWO(PADDW ( M80, MA2 )) /* t2 += 0x80 */ ;\ + ;\ + MOVQ ( MA1, MP1 ) ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 */ ;\ + ;\ +TWO(MOVQ ( MA2, MP2 )) ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 */ ;\ + ;\ + PADDW ( MP1, MA1 ) /* t1 + (t1 >> 8) ~= (t1/255) << 8 */ ;\ +TWO(PADDW ( MP2, MA2 )) /* t2 + (t2 >> 8) ~= (t2/255) << 8 */ ;\ + ;\ + PADDW ( MQ1, MA1 ) /* (t1/255 + q1) << 8 */ ;\ +TWO(PADDW ( MQ2, MA2 )) /* (t2/255 + q2) << 8 */ ;\ + ;\ + PSRLW ( CONST(8), MA1 ) /* sa1 | sb1 | sg1 | sr1 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* sa2 | sb2 | sg2 | sr2 */ + + +/* linear interpolation - geometric series with correction + * + * instead of the roundoff this adds a small correction to satisfy the OpenGL criteria + * + * t/255 ~= (t + (t >> 8) + (t >> 15)) >> 8 + * + * note that although is faster than rounding off it doesn't give always the exact results + */ +#define GMB_LERP_GSC( MP1, MQ1, MA1, MP2, MQ2, MA2) \ + PSUBW ( MQ1, MP1 ) /* pa1 - qa1 | pb1 - qb1 | pg1 - qg1 | pr1 - qr1 */ ;\ + PSLLW ( CONST(8), MQ1 ) /* q1 << 8 */ ;\ + PMULLW ( MP1, MA1 ) /* t1 = (q1 - p1)*pa1 */ ;\ + ;\ +TWO(PSUBW ( MQ2, MP2 )) /* pa2 - qa2 | pb2 - qb2 | pg2 - qg2 | pr2 - qr2 */ ;\ +TWO(PSLLW ( CONST(8), MQ2 )) /* q2 << 8 */ ;\ +TWO(PMULLW ( MP2, MA2 )) /* t2 = (q2 - p2)*pa2 */ ;\ + ;\ + MOVQ ( MA1, MP1 ) ;\ + PSRLW ( CONST(8), MA1 ) /* t1 >> 8 */ ;\ + ;\ +TWO(MOVQ ( MA2, MP2 )) ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* t2 >> 8 */ ;\ + ;\ + PADDW ( MA1, MP1 ) /* t1 + (t1 >> 8) ~= (t1/255) << 8 */ ;\ + PSRLW ( CONST(7), MA1 ) /* t1 >> 15 */ ;\ + ;\ +TWO(PADDW ( MA2, MP2 )) /* t2 + (t2 >> 8) ~= (t2/255) << 8 */ ;\ +TWO(PSRLW ( CONST(7), MA2 )) /* t2 >> 15 */ ;\ + ;\ + PADDW ( MP1, MA1 ) /* t1 + (t1 >> 8) + (t1 >>15) ~= (t1/255) << 8 */ ;\ +TWO(PADDW ( MP2, MA2 )) /* t2 + (t2 >> 8) + (t2 >>15) ~= (t2/255) << 8 */ ;\ + ;\ + PADDW ( MQ1, MA1 ) /* (t1/255 + q1) << 8 */ ;\ +TWO(PADDW ( MQ2, MA2 )) /* (t2/255 + q2) << 8 */ ;\ + ;\ + PSRLW ( CONST(8), MA1 ) /* sa1 | sb1 | sg1 | sr1 */ ;\ +TWO(PSRLW ( CONST(8), MA2 )) /* sa2 | sb2 | sg2 | sr2 */ + + +/* common blending setup code + * + * note that M00 is a register with 0x0000000000000000 constant which can be easily obtained making + * + * PXOR ( M00, M00 ) + */ +#define GMB_LOAD(rgba, dest, MPP, MQQ) \ +ONE(MOVD ( REGIND(rgba), MPP )) /* | | | | qa1 | qb1 | qg1 | qr1 */ ;\ +ONE(MOVD ( REGIND(dest), MQQ )) /* | | | | pa1 | pb1 | pg1 | pr1 */ ;\ + ;\ +TWO(MOVQ ( REGIND(rgba), MPP )) /* qa2 | qb2 | qg2 | qr2 | qa1 | qb1 | qg1 | qr1 */ ;\ +TWO(MOVQ ( REGIND(dest), MQQ )) /* pa2 | pb2 | pg2 | pr2 | pa1 | pb1 | pg1 | pr1 */ + +#define GMB_UNPACK(MP1, MQ1, MP2, MQ2, M00) \ +TWO(MOVQ ( MP1, MP2 )) ;\ +TWO(MOVQ ( MQ1, MQ2 )) ;\ + ;\ + PUNPCKLBW ( M00, MQ1 ) /* qa1 | qb1 | qg1 | qr1 */ ;\ +TWO(PUNPCKHBW ( M00, MQ2 )) /* qa2 | qb2 | qg2 | qr2 */ ;\ + PUNPCKLBW ( M00, MP1 ) /* pa1 | pb1 | pg1 | pr1 */ ;\ +TWO(PUNPCKHBW ( M00, MP2 )) /* pa2 | pb2 | pg2 | pr2 */ + +#define GMB_ALPHA(MP1, MA1, MP2, MA2) \ + MOVQ ( MP1, MA1 ) ;\ +TWO(MOVQ ( MP2, MA2 )) ;\ + ;\ + PUNPCKHWD ( MA1, MA1 ) /* pa1 | pa1 | | */ ;\ +TWO(PUNPCKHWD ( MA2, MA2 )) /* pa2 | pa2 | | */ ;\ + PUNPCKHDQ ( MA1, MA1 ) /* pa1 | pa1 | pa1 | pa1 */ ;\ +TWO(PUNPCKHDQ ( MA2, MA2 )) /* pa2 | pa2 | pa2 | pa2 */ + +#define GMB_PACK( MS1, MS2 ) \ + PACKUSWB ( MS2, MS1 ) /* sa2 | sb2 | sg2 | sr2 | sa1 | sb1 | sg1 | sr1 */ ;\ + +#define GMB_STORE(rgba, MSS ) \ +ONE(MOVD ( MSS, REGIND(rgba) )) /* | | | | sa1 | sb1 | sg1 | sr1 */ ;\ +TWO(MOVQ ( MSS, REGIND(rgba) )) /* sa2 | sb2 | sg2 | sr2 | sa1 | sb1 | sg1 | sr1 */ + + + SEG_DATA + +ALIGNDATA8 +const_0080: + D_LONG 0x00800080, 0x00800080 + +const_80: + D_LONG 0x80808080, 0x80808080 + + SEG_TEXT + + +/* Blend transparency function + */ + +#define TAG(x) CONCAT(x,_transparency) +#define LLTAG(x) LLBL2(x,_transparency) + +#define INIT \ + PXOR ( MM0, MM0 ) /* 0x0000 | 0x0000 | 0x0000 | 0x0000 */ + +#define MAIN( rgba, dest ) \ + GMB_LOAD( rgba, dest, MM1, MM2 ) ;\ + GMB_UNPACK( MM1, MM2, MM4, MM5, MM0 ) ;\ + GMB_ALPHA( MM1, MM3, MM4, MM6 ) ;\ + GMB_LERP_GSC( MM1, MM2, MM3, MM4, MM5, MM6 ) ;\ + GMB_PACK( MM3, MM6 ) ;\ + GMB_STORE( rgba, MM3 ) + +#include "mmx_blendtmp.h" + + +/* Blend add function + * + * FIXME: Add some loop unrolling here... + */ + +#define TAG(x) CONCAT(x,_add) +#define LLTAG(x) LLBL2(x,_add) + +#define INIT + +#define MAIN( rgba, dest ) \ +ONE(MOVD ( REGIND(rgba), MM1 )) /* | | | | qa1 | qb1 | qg1 | qr1 */ ;\ +ONE(MOVD ( REGIND(dest), MM2 )) /* | | | | pa1 | pb1 | pg1 | pr1 */ ;\ +ONE(PADDUSB ( MM2, MM1 )) ;\ +ONE(MOVD ( MM1, REGIND(rgba) )) /* | | | | sa1 | sb1 | sg1 | sr1 */ ;\ + ;\ +TWO(MOVQ ( REGIND(rgba), MM1 )) /* qa2 | qb2 | qg2 | qr2 | qa1 | qb1 | qg1 | qr1 */ ;\ +TWO(PADDUSB ( REGIND(dest), MM1 )) /* sa2 | sb2 | sg2 | sr2 | sa1 | sb1 | sg1 | sr1 */ ;\ +TWO(MOVQ ( MM1, REGIND(rgba) )) + +#include "mmx_blendtmp.h" + + +/* Blend min function + */ + +#define TAG(x) CONCAT(x,_min) +#define LLTAG(x) LLBL2(x,_min) + +#define INIT \ + MOVQ ( CONTENT(const_80), MM7 ) /* 0x80| 0x80| 0x80| 0x80| 0x80| 0x80| 0x80| 0x80*/ + +#define MAIN( rgba, dest ) \ + GMB_LOAD( rgba, dest, MM1, MM2 ) ;\ + MOVQ ( MM1, MM3 ) ;\ + MOVQ ( MM2, MM4 ) ;\ + PXOR ( MM7, MM3 ) /* unsigned -> signed */ ;\ + PXOR ( MM7, MM4 ) /* unsigned -> signed */ ;\ + PCMPGTB ( MM3, MM4 ) /* q > p ? 0xff : 0x00 */ ;\ + PAND ( MM4, MM1 ) /* q > p ? p : 0 */ ;\ + PANDN ( MM2, MM4 ) /* q > p ? 0 : q */ ;\ + POR ( MM1, MM4 ) /* q > p ? p : q */ ;\ + GMB_STORE( rgba, MM4 ) + +#include "mmx_blendtmp.h" + + +/* Blend max function + */ + +#define TAG(x) CONCAT(x,_max) +#define LLTAG(x) LLBL2(x,_max) + +#define INIT \ + MOVQ ( CONTENT(const_80), MM7 ) /* 0x80| 0x80| 0x80| 0x80| 0x80| 0x80| 0x80| 0x80*/ + +#define MAIN( rgba, dest ) \ + GMB_LOAD( rgba, dest, MM1, MM2 ) ;\ + MOVQ ( MM1, MM3 ) ;\ + MOVQ ( MM2, MM4 ) ;\ + PXOR ( MM7, MM3 ) /* unsigned -> signed */ ;\ + PXOR ( MM7, MM4 ) /* unsigned -> signed */ ;\ + PCMPGTB ( MM3, MM4 ) /* q > p ? 0xff : 0x00 */ ;\ + PAND ( MM4, MM2 ) /* q > p ? q : 0 */ ;\ + PANDN ( MM1, MM4 ) /* q > p ? 0 : p */ ;\ + POR ( MM2, MM4 ) /* q > p ? p : q */ ;\ + GMB_STORE( rgba, MM4 ) + +#include "mmx_blendtmp.h" + + +/* Blend modulate function + */ + +#define TAG(x) CONCAT(x,_modulate) +#define LLTAG(x) LLBL2(x,_modulate) + +#define INIT \ + PXOR ( MM0, MM0 ) /* 0x0000 | 0x0000 | 0x0000 | 0x0000 */ ;\ + MOVQ ( CONTENT(const_0080), MM7 ) /* 0x0080 | 0x0080 | 0x0080 | 0x0080 */ + +#define MAIN( rgba, dest ) \ + GMB_LOAD( rgba, dest, MM1, MM2 ) ;\ + GMB_UNPACK( MM1, MM2, MM4, MM5, MM0 ) ;\ + GMB_MULT_GSR( MM1, MM2, MM4, MM5, MM7 ) ;\ + GMB_PACK( MM2, MM5 ) ;\ + GMB_STORE( rgba, MM2 ) + +#include "mmx_blendtmp.h" + +#endif diff --git a/src/mesa/x86/mmx_blendtmp.h b/src/mesa/x86/mmx_blendtmp.h new file mode 100644 index 0000000..5d22f28 --- /dev/null +++ b/src/mesa/x86/mmx_blendtmp.h @@ -0,0 +1,114 @@ +/* + * Written by José Fonseca <j_r_fonseca@yahoo.co.uk> + */ + + +/* + * void _mesa_mmx_blend( GLcontext *ctx, + * GLuint n, + * const GLubyte mask[], + * GLchan rgba[][4], + * CONST GLchan dest[][4] ) + * + */ +ALIGNTEXT16 +GLOBL GLNAME( TAG(_mesa_mmx_blend) ) + +GLNAME( TAG(_mesa_mmx_blend) ): + + PUSH_L ( EBP ) + MOV_L ( ESP, EBP ) + PUSH_L ( ESI ) + PUSH_L ( EDI ) + PUSH_L ( EBX ) + + MOV_L ( REGOFF(12, EBP), ECX ) /* n */ + CMP_L ( CONST(0), ECX) + JE ( LLTAG(GMB_return) ) + + MOV_L ( REGOFF(16, EBP), EBX ) /* mask */ + MOV_L ( REGOFF(20, EBP), EDI ) /* rgba */ + MOV_L ( REGOFF(24, EBP), ESI ) /* dest */ + + INIT + + TEST_L ( CONST(4), EDI ) /* align rgba on an 8-byte boundary */ + JZ ( LLTAG(GMB_align_end) ) + + CMP_B ( CONST(0), REGIND(EBX) ) /* *mask == 0 */ + JE ( LLTAG(GMB_align_continue) ) + + /* runin */ +#define ONE(x) x +#define TWO(x) + MAIN ( EDI, ESI ) +#undef ONE +#undef TWO + +LLTAG(GMB_align_continue): + + DEC_L ( ECX ) /* n -= 1 */ + INC_L ( EBX ) /* mask += 1 */ + ADD_L ( CONST(4), EDI ) /* rgba += 1 */ + ADD_L ( CONST(4), ESI ) /* dest += 1 */ + +LLTAG(GMB_align_end): + + CMP_L ( CONST(2), ECX) + JB ( LLTAG(GMB_loop_end) ) + +ALIGNTEXT16 +LLTAG(GMB_loop_begin): + + CMP_W ( CONST(0), REGIND(EBX) ) /* *mask == 0 && *(mask + 1) == 0 */ + JE ( LLTAG(GMB_loop_continue) ) + + /* main loop */ +#define ONE(x) +#define TWO(x) x + MAIN ( EDI, ESI ) +#undef ONE +#undef TWO + +LLTAG(GMB_loop_continue): + + DEC_L ( ECX ) + DEC_L ( ECX ) /* n -= 2 */ + ADD_L ( CONST(2), EBX ) /* mask += 2 */ + ADD_L ( CONST(8), EDI ) /* rgba += 2 */ + ADD_L ( CONST(8), ESI ) /* dest += 2 */ + CMP_L ( CONST(2), ECX ) + JAE ( LLTAG(GMB_loop_begin) ) + +LLTAG(GMB_loop_end): + + CMP_L ( CONST(1), ECX ) + JB ( LLTAG(GMB_done) ) + + CMP_B ( CONST(0), REGIND(EBX) ) /* *mask == 0 */ + JE ( LLTAG(GMB_done) ) + + /* runout */ +#define ONE(x) x +#define TWO(x) + MAIN ( EDI, ESI ) +#undef ONE +#undef TWO + +LLTAG(GMB_done): + + EMMS + +LLTAG(GMB_return): + + POP_L ( EBX ) + POP_L ( EDI ) + POP_L ( ESI ) + MOV_L ( EBP, ESP ) + POP_L ( EBP ) + RET + +#undef TAG +#undef LLTAG +#undef INIT +#undef MAIN diff --git a/src/mesa/x86/norm_args.h b/src/mesa/x86/norm_args.h new file mode 100644 index 0000000..1b43d57 --- /dev/null +++ b/src/mesa/x86/norm_args.h @@ -0,0 +1,58 @@ +/* $Id: norm_args.h,v 1.4 2003/11/26 08:32:36 dborca Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Normal transform function interface for assembly code. Simply define + * FRAME_OFFSET to the number of bytes pushed onto the stack before + * using the ARG_* argument macros. + * + * Gareth Hughes + */ + +#ifndef __NORM_ARGS_H__ +#define __NORM_ARGS_H__ + +/* Offsets for normal_func arguments + * + * typedef void (*normal_func)( CONST GLmatrix *mat, + * GLfloat scale, + * CONST GLvector4f *in, + * CONST GLfloat lengths[], + * GLvector4f *dest ); + */ +#define OFFSET_MAT 4 +#define OFFSET_SCALE 8 +#define OFFSET_IN 12 +#define OFFSET_LENGTHS 16 +#define OFFSET_DEST 20 + +#define ARG_MAT REGOFF(FRAME_OFFSET+OFFSET_MAT, ESP) +#define ARG_SCALE REGOFF(FRAME_OFFSET+OFFSET_SCALE, ESP) +#define ARG_IN REGOFF(FRAME_OFFSET+OFFSET_IN, ESP) +#define ARG_LENGTHS REGOFF(FRAME_OFFSET+OFFSET_LENGTHS, ESP) +#define ARG_DEST REGOFF(FRAME_OFFSET+OFFSET_DEST, ESP) + +#endif diff --git a/src/mesa/x86/sse.c b/src/mesa/x86/sse.c new file mode 100644 index 0000000..45852c4 --- /dev/null +++ b/src/mesa/x86/sse.c @@ -0,0 +1,123 @@ +/* + * Mesa 3-D graphics library + * Version: 6.0 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * PentiumIII-SIMD (SSE) optimizations contributed by + * Andre Werthmann <wertmann@cs.uni-potsdam.de> + */ + +#include "glheader.h" +#include "context.h" +#include "math/m_xform.h" +#include "tnl/t_context.h" + +#include "sse.h" +#include "common_x86_macros.h" + +#ifdef DEBUG +#include "math/m_debug.h" +#endif + + +#ifdef USE_SSE_ASM +DECLARE_XFORM_GROUP( sse, 2 ) +DECLARE_XFORM_GROUP( sse, 3 ) + +#if 1 +/* Some functions are not written in SSE-assembly, because the fpu ones are faster */ +extern void _ASMAPI _mesa_sse_transform_normals_no_rot( NORM_ARGS ); +extern void _ASMAPI _mesa_sse_transform_rescale_normals( NORM_ARGS ); +extern void _ASMAPI _mesa_sse_transform_rescale_normals_no_rot( NORM_ARGS ); + +extern void _ASMAPI _mesa_sse_transform_points4_general( XFORM_ARGS ); +extern void _ASMAPI _mesa_sse_transform_points4_3d( XFORM_ARGS ); +/* XXX this function segfaults, see below */ +extern void _ASMAPI _mesa_sse_transform_points4_identity( XFORM_ARGS ); +/* XXX this one works, see below */ +extern void _ASMAPI _mesa_x86_transform_points4_identity( XFORM_ARGS ); +#else +DECLARE_NORM_GROUP( sse ) +#endif + + +extern void _ASMAPI +_mesa_v16_sse_general_xform( GLfloat *first_vert, + const GLfloat *m, + const GLfloat *src, + GLuint src_stride, + GLuint count ); + +extern void _ASMAPI +_mesa_sse_project_vertices( GLfloat *first, + GLfloat *last, + const GLfloat *m, + GLuint stride ); + +extern void _ASMAPI +_mesa_sse_project_clipped_vertices( GLfloat *first, + GLfloat *last, + const GLfloat *m, + GLuint stride, + const GLubyte *clipmask ); +#endif + + +void _mesa_init_sse_transform_asm( void ) +{ +#ifdef USE_SSE_ASM + ASSIGN_XFORM_GROUP( sse, 2 ); + ASSIGN_XFORM_GROUP( sse, 3 ); + +#if 1 + /* TODO: Finish these off. + */ + _mesa_transform_tab[4][MATRIX_GENERAL] = + _mesa_sse_transform_points4_general; + _mesa_transform_tab[4][MATRIX_3D] = + _mesa_sse_transform_points4_3d; + /* XXX NOTE: _mesa_sse_transform_points4_identity segfaults with the + conformance tests, so use the x86 version. + */ + _mesa_transform_tab[4][MATRIX_IDENTITY] = + _mesa_x86_transform_points4_identity;/*_mesa_sse_transform_points4_identity;*/ + + _mesa_normal_tab[NORM_TRANSFORM_NO_ROT] = + _mesa_sse_transform_normals_no_rot; + _mesa_normal_tab[NORM_TRANSFORM | NORM_RESCALE] = + _mesa_sse_transform_rescale_normals; + _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_RESCALE] = + _mesa_sse_transform_rescale_normals_no_rot; +#else + ASSIGN_XFORM_GROUP( sse, 4 ); + + ASSIGN_NORM_GROUP( sse ); +#endif + +#ifdef DEBUG + _math_test_all_transform_functions( "SSE" ); + _math_test_all_normal_transform_functions( "SSE" ); +#endif +#endif +} + diff --git a/src/mesa/x86/sse.h b/src/mesa/x86/sse.h new file mode 100644 index 0000000..98146a9 --- /dev/null +++ b/src/mesa/x86/sse.h @@ -0,0 +1,39 @@ +/* $Id: sse.h,v 1.2 2002/04/09 14:58:03 keithw Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * PentiumIII-SIMD (SSE) optimizations contributed by + * Andre Werthmann <wertmann@cs.uni-potsdam.de> + */ + +#ifndef __SSE_H__ +#define __SSE_H__ + +#include "math/m_xform.h" + +void _mesa_init_sse_transform_asm( void ); + +#endif diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S new file mode 100644 index 0000000..ca545f4 --- /dev/null +++ b/src/mesa/x86/sse_normal.S @@ -0,0 +1,254 @@ +/* $Id: sse_normal.S,v 1.4 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** TODO: + * - insert PREFETCH instructions to avoid cache-misses ! + * - some more optimizations are possible... + * - for 40-50% more performance in the SSE-functions, the + * data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned ! + */ + +#ifdef USE_SSE_ASM +#include "matypes.h" +#include "norm_args.h" + + SEG_TEXT + +#define M(i) REGOFF(i * 4, EDX) +#define S(i) REGOFF(i * 4, ESI) +#define D(i) REGOFF(i * 4, EDI) +#define STRIDE REGOFF(12, ESI) + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_sse_transform_rescale_normals_no_rot) +GLNAME(_mesa_sse_transform_rescale_normals_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L ( ARG_IN, ESI ) /* ptr to source GLvector3f */ + MOV_L ( ARG_DEST, EDI ) /* ptr to dest GLvector3f */ + + MOV_L ( ARG_MAT, EDX ) /* ptr to matrix */ + MOV_L ( REGOFF(MATRIX_INV, EDX), EDX) /* matrix->inv */ + + MOV_L ( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L ( ECX, ECX ) + JZ( LLBL(K_G3TRNNRR_finish) ) /* count was zero; go to finish */ + + MOV_L ( STRIDE, EAX ) /* stride */ + MOV_L ( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest-count */ + + IMUL_L( CONST(16), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* m0 */ + MOVSS ( M(5), XMM2 ) /* m5 */ + UNPCKLPS( XMM2, XMM1 ) /* m5 | m0 */ + MOVSS ( ARG_SCALE, XMM0 ) /* scale */ + SHUFPS ( CONST(0x0), XMM0, XMM0 ) /* scale | scale */ + MULPS ( XMM0, XMM1 ) /* m5*scale | m0*scale */ + MULSS ( M(10), XMM0 ) /* m10*scale */ + +ALIGNTEXT32 +LLBL(K_G3TRNNRR_top): + MOVLPS ( S(0), XMM2 ) /* uy | ux */ + MULPS ( XMM1, XMM2 ) /* uy*m5*scale | ux*m0*scale */ + MOVLPS ( XMM2, D(0) ) /* ->D(1) | D(0) */ + + MOVSS ( S(2), XMM2 ) /* uz */ + MULSS ( XMM0, XMM2 ) /* uz*m10*scale */ + MOVSS ( XMM2, D(2) ) /* ->D(2) */ + +LLBL(K_G3TRNNRR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_G3TRNNRR_top) ) + +LLBL(K_G3TRNNRR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_sse_transform_rescale_normals) +GLNAME(_mesa_sse_transform_rescale_normals): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L ( ARG_IN, ESI ) /* ptr to source GLvector3f */ + MOV_L ( ARG_DEST, EDI ) /* ptr to dest GLvector3f */ + + MOV_L ( ARG_MAT, EDX ) /* ptr to matrix */ + MOV_L ( REGOFF(MATRIX_INV, EDX), EDX) /* matrix->inv */ + + MOV_L ( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L ( ECX, ECX ) + JZ( LLBL(K_G3TRNR_finish) ) /* count was zero; go to finish */ + + MOV_L ( STRIDE, EAX ) /* stride */ + MOV_L ( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest-count */ + + IMUL_L( CONST(16), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM0 ) /* m0 */ + MOVSS ( M(4), XMM1 ) /* m4 */ + UNPCKLPS( XMM1, XMM0 ) /* m4 | m0 */ + + MOVSS ( ARG_SCALE, XMM4 ) /* scale */ + SHUFPS ( CONST(0x0), XMM4, XMM4 ) /* scale | scale */ + + MULPS ( XMM4, XMM0 ) /* m4*scale | m0*scale */ + MOVSS ( M(1), XMM1 ) /* m1 */ + MOVSS ( M(5), XMM2 ) /* m5 */ + UNPCKLPS( XMM2, XMM1 ) /* m5 | m1 */ + MULPS ( XMM4, XMM1 ) /* m5*scale | m1*scale */ + MOVSS ( M(2), XMM2 ) /* m2 */ + MOVSS ( M(6), XMM3 ) /* m6 */ + UNPCKLPS( XMM3, XMM2 ) /* m6 | m2 */ + MULPS ( XMM4, XMM2 ) /* m6*scale | m2*scale */ + + MOVSS ( M(8), XMM6 ) /* m8 */ + MULSS ( ARG_SCALE, XMM6 ) /* m8*scale */ + MOVSS ( M(9), XMM7 ) /* m9 */ + MULSS ( ARG_SCALE, XMM7 ) /* m9*scale */ + +ALIGNTEXT32 +LLBL(K_G3TRNR_top): + MOVSS ( S(0), XMM3 ) /* ux */ + SHUFPS ( CONST(0x0), XMM3, XMM3 ) /* ux | ux */ + MULPS ( XMM0, XMM3 ) /* ux*m4 | ux*m0 */ + MOVSS ( S(1), XMM4 ) /* uy */ + SHUFPS ( CONST(0x0), XMM4, XMM4 ) /* uy | uy */ + MULPS ( XMM1, XMM4 ) /* uy*m5 | uy*m1 */ + MOVSS ( S(2), XMM5 ) /* uz */ + SHUFPS ( CONST(0x0), XMM5, XMM5 ) /* uz | uz */ + MULPS ( XMM2, XMM5 ) /* uz*m6 | uz*m2 */ + + ADDPS ( XMM4, XMM3 ) + ADDPS ( XMM5, XMM3 ) + MOVLPS ( XMM3, D(0) ) + + MOVSS ( M(10), XMM3 ) /* m10 */ + MULSS ( ARG_SCALE, XMM3 ) /* m10*scale */ + MULSS ( S(2), XMM3 ) /* m10*scale*uz */ + MOVSS ( S(1), XMM4 ) /* uy */ + MULSS ( XMM7, XMM4 ) /* uy*m9*scale */ + MOVSS ( S(0), XMM5 ) /* ux */ + MULSS ( XMM6, XMM5 ) /* ux*m8*scale */ + + ADDSS ( XMM4, XMM3 ) + ADDSS ( XMM5, XMM3 ) + MOVSS ( XMM3, D(2) ) + +LLBL(K_G3TRNR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_G3TRNR_top) ) + +LLBL(K_G3TRNR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_sse_transform_normals_no_rot) +GLNAME(_mesa_sse_transform_normals_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L ( ARG_IN, ESI ) /* ptr to source GLvector3f */ + MOV_L ( ARG_DEST, EDI ) /* ptr to dest GLvector3f */ + + MOV_L ( ARG_MAT, EDX ) /* ptr to matrix */ + MOV_L ( REGOFF(MATRIX_INV, EDX), EDX) /* matrix->inv */ + + MOV_L ( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L ( ECX, ECX ) + JZ( LLBL(K_G3TNNRR_finish) ) /* count was zero; go to finish */ + + MOV_L ( STRIDE, EAX ) /* stride */ + MOV_L ( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest-count */ + + IMUL_L( CONST(16), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS( M(0), XMM0 ) /* m0 */ + MOVSS( M(5), XMM1 ) /* m5 */ + UNPCKLPS( XMM1, XMM0 ) /* m5 | m0 */ + MOVSS( M(10), XMM1 ) /* m10 */ + +ALIGNTEXT32 +LLBL(K_G3TNNRR_top): + MOVLPS( S(0), XMM2 ) /* uy | ux */ + MULPS( XMM0, XMM2 ) /* uy*m5 | ux*m0 */ + MOVLPS( XMM2, D(0) ) + + MOVSS( S(2), XMM2 ) /* uz */ + MULSS( XMM1, XMM2 ) /* uz*m10 */ + MOVSS( XMM2, D(2) ) + +LLBL(K_G3TNNRR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_G3TNNRR_top) ) + +LLBL(K_G3TNNRR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET +#endif diff --git a/src/mesa/x86/sse_xform1.S b/src/mesa/x86/sse_xform1.S new file mode 100644 index 0000000..90768ba --- /dev/null +++ b/src/mesa/x86/sse_xform1.S @@ -0,0 +1,435 @@ +/* $Id: sse_xform1.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** TODO: + * - insert PREFETCH instructions to avoid cache-misses ! + * - some more optimizations are possible... + * - for 40-50% more performance in the SSE-functions, the + * data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned ! + */ + +#ifdef USE_SSE_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define S(i) REGOFF(i * 4, ESI) +#define D(i) REGOFF(i * 4, EDI) +#define M(i) REGOFF(i * 4, EDX) + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_general) +GLNAME( _mesa_sse_transform_points1_general ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + CMP_L( CONST(0), ECX ) /* count == 0 ? */ + JE( LLBL(K_GTP1GR_finish) ) /* yes -> nothing to do. */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVAPS( M(0), XMM0 ) /* m3 | m2 | m1 | m0 */ + MOVAPS( M(12), XMM1 ) /* m15 | m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP1GR_top): + MOVSS( S(0), XMM2 ) /* ox */ + SHUFPS( CONST(0x0), XMM2, XMM2 ) /* ox | ox | ox | ox */ + MULPS( XMM0, XMM2 ) /* ox*m3 | ox*m2 | ox*m1 | ox*m0 */ + ADDPS( XMM1, XMM2 ) /* + | + | + | + */ + MOVUPS( XMM2, D(0) ) + +LLBL(K_GTP1GR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP1GR_top) ) + +LLBL(K_GTP1GR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_identity) +GLNAME( _mesa_sse_transform_points1_identity ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP1IR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_1), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(1), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + CMP_L( ESI, EDI ) + JE( LLBL(K_GTP1IR_finish) ) + + +ALIGNTEXT32 +LLBL(K_GTP1IR_top): + MOV_L( S(0), EDX ) + MOV_L( EDX, D(0) ) + +LLBL(K_GTP1IR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP1IR_top) ) + +LLBL(K_GTP1IR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_3d_no_rot) +GLNAME(_mesa_sse_transform_points1_3d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP13DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVSS( M(0), XMM0 ) /* m0 */ + MOVSS( M(12), XMM1 ) /* m12 */ + MOVSS( M(13), XMM2 ) /* m13 */ + MOVSS( M(14), XMM3 ) /* m14 */ + +ALIGNTEXT32 +LLBL(K_GTP13DNRR_top): + MOVSS( S(0), XMM4 ) /* ox */ + MULSS( XMM0, XMM4 ) /* ox*m0 */ + ADDSS( XMM1, XMM4 ) /* ox*m0+m12 */ + MOVSS( XMM4, D(0) ) + + MOVSS( XMM2, D(1) ) + MOVSS( XMM3, D(2) ) + +LLBL(K_GTP13DNRR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP13DNRR_top) ) + +LLBL(K_GTP13DNRR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_perspective) +GLNAME(_mesa_sse_transform_points1_perspective): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP13PR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + XORPS( XMM0, XMM0 ) /* 0 | 0 | 0 | 0 */ + MOVSS( M(0), XMM1 ) /* m0 */ + MOVSS( M(14), XMM2 ) /* m14 */ + +ALIGNTEXT32 +LLBL(K_GTP13PR_top): + MOVSS( S(0), XMM3 ) /* ox */ + MULSS( XMM1, XMM3 ) /* ox*m0 */ + MOVSS( XMM3, D(0) ) /* ox*m0->D(0) */ + MOVSS( XMM2, D(2) ) /* m14->D(2) */ + + MOVSS( XMM0, D(1) ) + MOVSS( XMM0, D(3) ) + +LLBL(K_GTP13PR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP13PR_top) ) + +LLBL(K_GTP13PR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_2d) +GLNAME(_mesa_sse_transform_points1_2d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP13P2DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVLPS( M(0), XMM0 ) /* m1 | m0 */ + MOVLPS( M(12), XMM1 ) /* m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP13P2DR_top): + MOVSS( S(0), XMM2 ) /* ox */ + SHUFPS( CONST(0x0), XMM2, XMM2 ) /* ox | ox | ox | ox */ + MULPS( XMM0, XMM2 ) /* - | - | ox*m1 | ox*m0 */ + ADDPS( XMM1, XMM2 ) /* - | - | ox*m1+m13 | ox*m0+m12 */ + MOVLPS( XMM2, D(0) ) + +LLBL(K_GTP13P2DR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP13P2DR_top) ) + +LLBL(K_GTP13P2DR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_2d_no_rot) +GLNAME(_mesa_sse_transform_points1_2d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP13P2DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS( M(0), XMM0 ) /* m0 */ + MOVSS( M(12), XMM1 ) /* m12 */ + MOVSS( M(13), XMM2 ) /* m13 */ + +ALIGNTEXT32 +LLBL(K_GTP13P2DNRR_top): + MOVSS( S(0), XMM3 ) /* ox */ + MULSS( XMM0, XMM3 ) /* ox*m0 */ + ADDSS( XMM1, XMM3 ) /* ox*m0+m12 */ + MOVSS( XMM3, D(0) ) + MOVSS( XMM2, D(1) ) + +LLBL(K_GTP13P2DNRR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP13P2DNRR_top) ) + +LLBL(K_GTP13P2DNRR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points1_3d) +GLNAME(_mesa_sse_transform_points1_3d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP13P3DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVAPS( M(0), XMM0 ) /* m3 | m2 | m1 | m0 */ + MOVAPS( M(12), XMM1 ) /* m15 | m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP13P3DR_top): + MOVSS( S(0), XMM2 ) /* ox */ + SHUFPS( CONST(0x0), XMM2, XMM2 ) /* ox | ox | ox | ox */ + MULPS( XMM0, XMM2 ) /* ox*m3 | ox*m2 | ox*m1 | ox*m0 */ + ADDPS( XMM1, XMM2 ) /* +m15 | +m14 | +m13 | +m12 */ + MOVLPS( XMM2, D(0) ) /* - | - | ->D(1)| ->D(0)*/ + UNPCKHPS( XMM2, XMM2 ) /* ox*m3+m15 | ox*m3+m15 | ox*m2+m14 | ox*m2+m14 */ + MOVSS( XMM2, D(2) ) + +LLBL(K_GTP13P3DR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP13P3DR_top) ) + +LLBL(K_GTP13P3DR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET +#endif diff --git a/src/mesa/x86/sse_xform2.S b/src/mesa/x86/sse_xform2.S new file mode 100644 index 0000000..49f9889 --- /dev/null +++ b/src/mesa/x86/sse_xform2.S @@ -0,0 +1,454 @@ +/* $Id: sse_xform2.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** TODO: + * - insert PREFETCH instructions to avoid cache-misses ! + * - some more optimizations are possible... + * - for 40-50% more performance in the SSE-functions, the + * data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned ! + */ + +#ifdef USE_SSE_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define S(i) REGOFF(i * 4, ESI) +#define D(i) REGOFF(i * 4, EDI) +#define M(i) REGOFF(i * 4, EDX) + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_general) +GLNAME( _mesa_sse_transform_points2_general ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX ) + JZ( LLBL(K_GTP2GR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVAPS( M(0), XMM0 ) /* m3 | m2 | m1 | m0 */ + MOVAPS( M(4), XMM1 ) /* m7 | m6 | m5 | m4 */ + MOVAPS( M(12), XMM2 ) /* m15 | m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP2GR_top): + MOVSS( S(0), XMM3 ) /* ox */ + SHUFPS( CONST(0x0), XMM3, XMM3 ) /* ox | ox | ox | ox */ + MULPS( XMM0, XMM3 ) /* ox*m3 | ox*m2 | ox*m1 | ox*m0 */ + MOVSS( S(1), XMM4 ) /* oy */ + SHUFPS( CONST(0x0), XMM4, XMM4 ) /* oy | oy | oy | oy */ + MULPS( XMM1, XMM4 ) /* oy*m7 | oy*m6 | oy*m5 | oy*m4 */ + + ADDPS( XMM4, XMM3 ) + ADDPS( XMM2, XMM3 ) + MOVAPS( XMM3, D(0) ) + +LLBL(K_GTP2GR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP2GR_top) ) + +LLBL(K_GTP2GR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_identity) +GLNAME( _mesa_sse_transform_points2_identity ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP2IR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + CMP_L( ESI, EDI ) + JE( LLBL(K_GTP2IR_finish) ) + + +ALIGNTEXT32 +LLBL(K_GTP2IR_top): + MOV_L ( S(0), EDX ) + MOV_L ( EDX, D(0) ) + MOV_L ( S(1), EDX ) + MOV_L ( EDX, D(1) ) + +LLBL(K_GTP2IR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP2IR_top) ) + +LLBL(K_GTP2IR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_3d_no_rot) +GLNAME(_mesa_sse_transform_points2_3d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP23DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* - | - | - | m0 */ + MOVSS ( M(5), XMM2 ) /* - | - | - | m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* - | - | m5 | m0 */ + MOVLPS ( M(12), XMM2 ) /* - | - | m13 | m12 */ + MOVSS ( M(14), XMM3 ) /* - | - | - | m14 */ + +ALIGNTEXT32 +LLBL(K_GTP23DNRR_top): + MOVLPS ( S(0), XMM0 ) /* - | - | oy | ox */ + MULPS ( XMM1, XMM0 ) /* - | - | oy*m5 | ox*m0 */ + ADDPS ( XMM2, XMM0 ) /* - | - | +m13 | +m12 */ + MOVLPS ( XMM0, D(0) ) /* -> D(1) | -> D(0) */ + + MOVSS ( XMM3, D(2) ) /* -> D(2) */ + +LLBL(K_GTP23DNRR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP23DNRR_top) ) + +LLBL(K_GTP23DNRR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_perspective) +GLNAME(_mesa_sse_transform_points2_perspective): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP23PR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* - | - | - | m0 */ + MOVSS ( M(5), XMM2 ) /* - | - | - | m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* - | - | m5 | m0 */ + MOVSS ( M(14), XMM3 ) /* m14 */ + XORPS ( XMM0, XMM0 ) /* 0 | 0 | 0 | 0 */ + +ALIGNTEXT32 +LLBL(K_GTP23PR_top): + MOVLPS( S(0), XMM4 ) /* oy | ox */ + MULPS( XMM1, XMM4 ) /* oy*m5 | ox*m0 */ + MOVLPS( XMM4, D(0) ) /* ->D(1) | ->D(0) */ + MOVSS( XMM3, D(2) ) /* ->D(2) */ + MOVSS( XMM0, D(3) ) /* ->D(3) */ + +LLBL(K_GTP23PR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP23PR_top) ) + +LLBL(K_GTP23PR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_2d) +GLNAME(_mesa_sse_transform_points2_2d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP23P2DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVLPS( M(0), XMM0 ) /* m1 | m0 */ + MOVLPS( M(4), XMM1 ) /* m5 | m4 */ + MOVLPS( M(12), XMM2 ) /* m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP23P2DR_top): + MOVSS( S(0), XMM3 ) /* ox */ + SHUFPS( CONST(0x0), XMM3, XMM3 ) /* ox | ox */ + MULPS( XMM0, XMM3 ) /* ox*m1 | ox*m0 */ + + MOVSS( S(1), XMM4 ) /* oy */ + SHUFPS( CONST(0x0), XMM4, XMM4 ) /* oy | oy */ + MULPS( XMM1, XMM4 ) /* oy*m5 | oy*m4 */ + + ADDPS( XMM4, XMM3 ) + ADDPS( XMM2, XMM3 ) + MOVLPS( XMM3, D(0) ) /* ->D(1) | ->D(0) */ + +LLBL(K_GTP23P2DR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP23P2DR_top) ) + +LLBL(K_GTP23P2DR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_2d_no_rot) +GLNAME(_mesa_sse_transform_points2_2d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP23P2DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* m0 */ + MOVSS ( M(5), XMM2 ) /* m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* m5 | m0 */ + MOVLPS ( M(12), XMM2 ) /* m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP23P2DNRR_top): + MOVLPS( S(0), XMM0 ) /* oy | ox */ + MULPS( XMM1, XMM0 ) /* oy*m5 | ox*m0 */ + ADDPS( XMM2, XMM0 ) /* +m13 | +m12 */ + MOVLPS( XMM0, D(0) ) /* ->D(1) | ->D(0) */ + +LLBL(K_GTP23P2DNRR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP23P2DNRR_top) ) + +LLBL(K_GTP23P2DNRR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points2_3d) +GLNAME(_mesa_sse_transform_points2_3d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP23P3DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVAPS( M(0), XMM0 ) /* m2 | m1 | m0 */ + MOVAPS( M(4), XMM1 ) /* m6 | m5 | m4 */ + MOVAPS( M(12), XMM2 ) /* m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP23P3DR_top): + MOVSS( S(0), XMM3 ) /* ox */ + SHUFPS( CONST(0x0), XMM3, XMM3 ) /* ox | ox | ox */ + MULPS( XMM0, XMM3 ) /* ox*m2 | ox*m1 | ox*m0 */ + + MOVSS( S(1), XMM4 ) /* oy */ + SHUFPS( CONST(0x0), XMM4, XMM4 ) /* oy | oy | oy */ + MULPS( XMM1, XMM4 ) /* oy*m6 | oy*m5 | oy*m4 */ + + ADDPS( XMM4, XMM3 ) + ADDPS( XMM2, XMM3 ) + + MOVLPS( XMM3, D(0) ) /* ->D(1) | ->D(0) */ + UNPCKHPS( XMM3, XMM3 ) + MOVSS( XMM3, D(2) ) /* ->D(2) */ + +LLBL(K_GTP23P3DR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP23P3DR_top) ) + +LLBL(K_GTP23P3DR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET +#endif diff --git a/src/mesa/x86/sse_xform3.S b/src/mesa/x86/sse_xform3.S new file mode 100644 index 0000000..f64baac --- /dev/null +++ b/src/mesa/x86/sse_xform3.S @@ -0,0 +1,500 @@ +/* $Id: sse_xform3.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** TODO: + * - insert PREFETCH instructions to avoid cache-misses ! + * - some more optimizations are possible... + * - for 40-50% more performance in the SSE-functions, the + * data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned ! + */ + +#ifdef USE_SSE_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define S(i) REGOFF(i * 4, ESI) +#define D(i) REGOFF(i * 4, EDI) +#define M(i) REGOFF(i * 4, EDX) + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_general) +GLNAME( _mesa_sse_transform_points3_general ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + CMP_L ( CONST(0), ECX ) /* count == 0 ? */ + JE ( LLBL(K_GTPGR_finish) ) /* yes -> nothing to do. */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVAPS ( REGOFF(0, EDX), XMM0 ) /* m0 | m1 | m2 | m3 */ + MOVAPS ( REGOFF(16, EDX), XMM1 ) /* m4 | m5 | m6 | m7 */ + MOVAPS ( REGOFF(32, EDX), XMM2 ) /* m8 | m9 | m10 | m11 */ + MOVAPS ( REGOFF(48, EDX), XMM3 ) /* m12 | m13 | m14 | m15 */ + + +ALIGNTEXT32 +LLBL(K_GTPGR_top): + MOVSS ( REGOFF(0, ESI), XMM4 ) /* | | | ox */ + SHUFPS ( CONST(0x0), XMM4, XMM4 ) /* ox | ox | ox | ox */ + MOVSS ( REGOFF(4, ESI), XMM5 ) /* | | | oy */ + SHUFPS ( CONST(0x0), XMM5, XMM5 ) /* oy | oy | oy | oy */ + MOVSS ( REGOFF(8, ESI), XMM6 ) /* | | | oz */ + SHUFPS ( CONST(0x0), XMM6, XMM6 ) /* oz | oz | oz | oz */ + + MULPS ( XMM0, XMM4 ) /* m3*ox | m2*ox | m1*ox | m0*ox */ + MULPS ( XMM1, XMM5 ) /* m7*oy | m6*oy | m5*oy | m4*oy */ + MULPS ( XMM2, XMM6 ) /* m11*oz | m10*oz | m9*oz | m8*oz */ + + ADDPS ( XMM5, XMM4 ) + ADDPS ( XMM6, XMM4 ) + ADDPS ( XMM3, XMM4 ) + + MOVAPS ( XMM4, REGOFF(0, EDI) ) + +LLBL(K_GTPGR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTPGR_top) ) + +LLBL(K_GTPGR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_identity) +GLNAME( _mesa_sse_transform_points3_identity ): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTPIR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + CMP_L( ESI, EDI ) + JE( LLBL(K_GTPIR_finish) ) + + +ALIGNTEXT32 +LLBL(K_GTPIR_top): + MOVLPS ( S(0), XMM0 ) + MOVLPS ( XMM0, D(0) ) + MOVSS ( S(2), XMM0 ) + MOVSS ( XMM0, D(2) ) + +LLBL(K_GTPIR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTPIR_top) ) + +LLBL(K_GTPIR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_3d_no_rot) +GLNAME(_mesa_sse_transform_points3_3d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP3DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* - | - | - | m0 */ + MOVSS ( M(5), XMM2 ) /* - | - | - | m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* - | - | m5 | m0 */ + MOVLPS ( M(12), XMM2 ) /* - | - | m13 | m12 */ + MOVSS ( M(10), XMM3 ) /* - | - | - | m10 */ + MOVSS ( M(14), XMM4 ) /* - | - | - | m14 */ + +ALIGNTEXT32 +LLBL(K_GTP3DNRR_top): + + MOVLPS ( S(0), XMM0 ) /* - | - | s1 | s0 */ + MULPS ( XMM1, XMM0 ) /* - | - | s1*m5 | s0*m0 */ + ADDPS ( XMM2, XMM0 ) /* - | - | +m13 | +m12 */ + MOVLPS ( XMM0, D(0) ) /* -> D(1) | -> D(0) */ + + MOVSS ( S(2), XMM0 ) /* sz */ + MULSS ( XMM3, XMM0 ) /* sz*m10 */ + ADDSS ( XMM4, XMM0 ) /* +m14 */ + MOVSS ( XMM0, D(2) ) /* -> D(2) */ + +LLBL(K_GTP3DNRR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP3DNRR_top) ) + +LLBL(K_GTP3DNRR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_perspective) +GLNAME(_mesa_sse_transform_points3_perspective): + +#define FRAME_OFFSET 8 + PUSH_L ( ESI ) + PUSH_L ( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP3PR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* - | - | - | m0 */ + MOVSS ( M(5), XMM2 ) /* - | - | - | m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* - | - | m5 | m0 */ + MOVLPS ( M(8), XMM2 ) /* - | - | m9 | m8 */ + MOVSS ( M(10), XMM3 ) /* m10 */ + MOVSS ( M(14), XMM4 ) /* m14 */ + XORPS ( XMM6, XMM6 ) /* 0 */ + +ALIGNTEXT32 +LLBL(K_GTP3PR_top): + MOVLPS ( S(0), XMM0 ) /* oy | ox */ + MULPS ( XMM1, XMM0 ) /* oy*m5 | ox*m0 */ + MOVSS ( S(2), XMM5 ) /* oz */ + SHUFPS ( CONST(0x0), XMM5, XMM5 ) /* oz | oz */ + MULPS ( XMM2, XMM5 ) /* oz*m9 | oz*m8 */ + ADDPS ( XMM5, XMM0 ) /* +oy*m5 | +ox*m0 */ + MOVLPS ( XMM0, D(0) ) /* ->D(1) | ->D(0) */ + + MOVSS ( S(2), XMM0 ) /* oz */ + MULSS ( XMM3, XMM0 ) /* oz*m10 */ + ADDSS ( XMM4, XMM0 ) /* +m14 */ + MOVSS ( XMM0, D(2) ) /* ->D(2) */ + + MOVSS ( S(2), XMM0 ) /* oz */ + MOVSS ( XMM6, XMM5 ) /* 0 */ + SUBPS ( XMM0, XMM5 ) /* -oz */ + MOVSS ( XMM5, D(3) ) /* ->D(3) */ + +LLBL(K_GTP3PR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP3PR_top) ) + +LLBL(K_GTP3PR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_2d) +GLNAME(_mesa_sse_transform_points3_2d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP3P2DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVLPS( M(0), XMM0 ) /* m1 | m0 */ + MOVLPS( M(4), XMM1 ) /* m5 | m4 */ + MOVLPS( M(12), XMM2 ) /* m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP3P2DR_top): + MOVSS ( S(0), XMM3 ) /* ox */ + SHUFPS ( CONST(0x0), XMM3, XMM3 ) /* ox | ox */ + MULPS ( XMM0, XMM3 ) /* ox*m1 | ox*m0 */ + MOVSS ( S(1), XMM4 ) /* oy */ + SHUFPS ( CONST(0x0), XMM4, XMM4 ) /* oy | oy */ + MULPS ( XMM1, XMM4 ) /* oy*m5 | oy*m4 */ + + ADDPS ( XMM4, XMM3 ) + ADDPS ( XMM2, XMM3 ) + MOVLPS ( XMM3, D(0) ) + + MOVSS ( S(2), XMM3 ) + MOVSS ( XMM3, D(2) ) + +LLBL(K_GTP3P2DR_skip): + ADD_L ( CONST(16), EDI ) + ADD_L ( EAX, ESI ) + CMP_L ( ECX, EDI ) + JNE ( LLBL(K_GTP3P2DR_top) ) + +LLBL(K_GTP3P2DR_finish): + POP_L ( EDI ) + POP_L ( ESI ) + RET +#undef FRAME_OFFSET + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_2d_no_rot) +GLNAME(_mesa_sse_transform_points3_2d_no_rot): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP3P2DNRR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + +ALIGNTEXT32 + MOVSS ( M(0), XMM1 ) /* m0 */ + MOVSS ( M(5), XMM2 ) /* m5 */ + UNPCKLPS ( XMM2, XMM1 ) /* m5 | m0 */ + MOVLPS ( M(12), XMM2 ) /* m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP3P2DNRR_top): + MOVLPS( S(0), XMM0 ) /* oy | ox */ + MULPS( XMM1, XMM0 ) /* oy*m5 | ox*m0 */ + ADDPS( XMM2, XMM0 ) /* +m13 | +m12 */ + MOVLPS( XMM0, D(0) ) /* ->D(1) | ->D(0) */ + + MOVSS( S(2), XMM0 ) + MOVSS( XMM0, D(2) ) + +LLBL(K_GTP3P2DNRR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP3P2DNRR_top) ) + +LLBL(K_GTP3P2DNRR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT4 +GLOBL GLNAME(_mesa_sse_transform_points3_3d) +GLNAME(_mesa_sse_transform_points3_3d): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( REGOFF(OFFSET_SOURCE+8, ESP), ESI ) /* ptr to source GLvector4f */ + MOV_L( REGOFF(OFFSET_DEST+8, ESP), EDI ) /* ptr to dest GLvector4f */ + + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP3P3DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) /* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + +ALIGNTEXT32 + MOVAPS( M(0), XMM0 ) /* m2 | m1 | m0 */ + MOVAPS( M(4), XMM1 ) /* m6 | m5 | m4 */ + MOVAPS( M(8), XMM2 ) /* m10 | m9 | m8 */ + MOVAPS( M(12), XMM3 ) /* m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL(K_GTP3P3DR_top): + MOVSS( S(0), XMM4 ) + SHUFPS( CONST(0x0), XMM4, XMM4 ) /* ox | ox | ox */ + MULPS( XMM0, XMM4 ) /* ox*m2 | ox*m1 | ox*m0 */ + + MOVSS( S(1), XMM5 ) + SHUFPS( CONST(0x0), XMM5, XMM5 ) /* oy | oy | oy */ + MULPS( XMM1, XMM5 ) /* oy*m6 | oy*m5 | oy*m4 */ + + MOVSS( S(2), XMM6 ) + SHUFPS( CONST(0x0), XMM6, XMM6 ) /* oz | oz | oz */ + MULPS( XMM2, XMM6 ) /* oz*m10 | oz*m9 | oz*m8 */ + + ADDPS( XMM5, XMM4 ) /* + | + | + */ + ADDPS( XMM6, XMM4 ) /* + | + | + */ + ADDPS( XMM3, XMM4 ) /* + | + | + */ + + MOVLPS( XMM4, D(0) ) /* => D(1) | => D(0) */ + UNPCKHPS( XMM4, XMM4 ) + MOVSS( XMM4, D(2) ) + +LLBL(K_GTP3P3DR_skip): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP3P3DR_top) ) + +LLBL(K_GTP3P3DR_finish): + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET +#endif diff --git a/src/mesa/x86/sse_xform4.S b/src/mesa/x86/sse_xform4.S new file mode 100644 index 0000000..5a64c9e --- /dev/null +++ b/src/mesa/x86/sse_xform4.S @@ -0,0 +1,228 @@ +/* $Id: sse_xform4.S,v 1.2 2004/04/26 10:10:25 alanh Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifdef USE_SSE_ASM +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FRAME_OFFSET 8 + +#define SRC(i) REGOFF(i * 4, ESI) +#define DST(i) REGOFF(i * 4, EDI) +#define MAT(i) REGOFF(i * 4, EDX) + +#define SELECT(r0, r1, r2, r3) CONST( r0 * 64 + r1 * 16 + r2 * 4 + r3 ) + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_sse_transform_points4_general ) +GLNAME( _mesa_sse_transform_points4_general ): + + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) /* verify non-zero count */ + JE( LLBL( sse_general_done ) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )/* set dest size */ + + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + + PREFETCHT0( REGIND(ESI) ) + + MOVAPS( MAT(0), XMM4 ) /* m3 | m2 | m1 | m0 */ + MOVAPS( MAT(4), XMM5 ) /* m7 | m6 | m5 | m4 */ + MOVAPS( MAT(8), XMM6 ) /* m11 | m10 | m9 | m8 */ + MOVAPS( MAT(12), XMM7 ) /* m15 | m14 | m13 | m12 */ + +ALIGNTEXT16 +LLBL( sse_general_loop ): + + MOVSS( SRC(0), XMM0 ) /* ox */ + SHUFPS( CONST(0x0), XMM0, XMM0 ) /* ox | ox | ox | ox */ + MULPS( XMM4, XMM0 ) /* ox*m3 | ox*m2 | ox*m1 | ox*m0 */ + + MOVSS( SRC(1), XMM1 ) /* oy */ + SHUFPS( CONST(0x0), XMM1, XMM1 ) /* oy | oy | oy | oy */ + MULPS( XMM5, XMM1 ) /* oy*m7 | oy*m6 | oy*m5 | oy*m4 */ + + MOVSS( SRC(2), XMM2 ) /* oz */ + SHUFPS( CONST(0x0), XMM2, XMM2 ) /* oz | oz | oz | oz */ + MULPS( XMM6, XMM2 ) /* oz*m11 | oz*m10 | oz*m9 | oz*m8 */ + + MOVSS( SRC(3), XMM3 ) /* ow */ + SHUFPS( CONST(0x0), XMM3, XMM3 ) /* ow | ow | ow | ow */ + MULPS( XMM7, XMM3 ) /* ow*m15 | ow*m14 | ow*m13 | ow*m12 */ + + ADDPS( XMM1, XMM0 ) /* ox*m3+oy*m7 | ... */ + ADDPS( XMM2, XMM0 ) /* ox*m3+oy*m7+oz*m11 | ... */ + ADDPS( XMM3, XMM0 ) /* ox*m3+oy*m7+oz*m11+ow*m15 | ... */ + MOVAPS( XMM0, DST(0) ) /* ->D(3) | ->D(2) | ->D(1) | ->D(0) */ + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + + DEC_L( ECX ) + JNZ( LLBL( sse_general_loop ) ) + +LLBL( sse_general_done ): + + POP_L( EDI ) + POP_L( ESI ) + RET + + + + +ALIGNTEXT4 +GLOBL GLNAME( _mesa_sse_transform_points4_3d ) +GLNAME( _mesa_sse_transform_points4_3d ): + + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) /* ptr to source GLvector4f */ + MOV_L( ARG_DEST, EDI ) /* ptr to dest GLvector4f */ + + MOV_L( ARG_MATRIX, EDX ) /* ptr to matrix */ + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) /* source count */ + + TEST_L( ECX, ECX) + JZ( LLBL(K_GTP43P3DR_finish) ) /* count was zero; go to finish */ + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) )/* set dest size */ + + SHL_L( CONST(4), ECX ) /* count *= 16 */ + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + ADD_L( EDI, ECX ) /* count += dest ptr */ + + MOVAPS( MAT(0), XMM0 ) /* m3 | m2 | m1 | m0 */ + MOVAPS( MAT(4), XMM1 ) /* m7 | m6 | m5 | m4 */ + MOVAPS( MAT(8), XMM2 ) /* m11 | m10 | m9 | m8 */ + MOVAPS( MAT(12), XMM3 ) /* m15 | m14 | m13 | m12 */ + +ALIGNTEXT32 +LLBL( K_GTP43P3DR_top ): + MOVSS( SRC(0), XMM4 ) /* ox */ + SHUFPS( CONST(0x0), XMM4, XMM4 ) /* ox | ox | ox | ox */ + MULPS( XMM0, XMM4 ) /* ox*m3 | ox*m2 | ox*m1 | ox*m0 */ + + MOVSS( SRC(1), XMM5 ) /* oy */ + SHUFPS( CONST(0x0), XMM5, XMM5 ) /* oy | oy | oy | oy */ + MULPS( XMM1, XMM5 ) /* oy*m7 | oy*m6 | oy*m5 | oy*m4 */ + + MOVSS( SRC(2), XMM6 ) /* oz */ + SHUFPS( CONST(0x0), XMM6, XMM6 ) /* oz | oz | oz | oz */ + MULPS( XMM2, XMM6 ) /* oz*m11 | oz*m10 | oz*m9 | oz*m8 */ + + MOVSS( SRC(3), XMM7 ) /* ow */ + SHUFPS( CONST(0x0), XMM7, XMM7 ) /* ow | ow | ow | ow */ + MULPS( XMM3, XMM7 ) /* ow*m15 | ow*m14 | ow*m13 | ow*m12 */ + + ADDPS( XMM5, XMM4 ) /* ox*m3+oy*m7 | ... */ + ADDPS( XMM6, XMM4 ) /* ox*m3+oy*m7+oz*m11 | ... */ + ADDPS( XMM7, XMM4 ) /* ox*m3+oy*m7+oz*m11+ow*m15 | ... */ + MOVAPS( XMM4, DST(0) ) /* ->D(3) | ->D(2) | ->D(1) | ->D(0) */ + + MOVSS( SRC(3), XMM4 ) /* ow */ + MOVSS( XMM4, DST(3) ) /* ->D(3) */ + +LLBL( K_GTP43P3DR_skip ): + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(K_GTP43P3DR_top) ) + +LLBL( K_GTP43P3DR_finish ): + POP_L( EDI ) + POP_L( ESI ) + RET + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_sse_transform_points4_identity ) +GLNAME( _mesa_sse_transform_points4_identity ): + + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) /* verify non-zero count */ + JE( LLBL( sse_identity_done ) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) /* stride */ + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) /* set dest flags */ + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) /* set dest count */ + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) )/* set dest size */ + + MOV_L( REGOFF(V4F_START, ESI), ESI ) /* ptr to first source vertex */ + MOV_L( REGOFF(V4F_START, EDI), EDI ) /* ptr to first dest vertex */ + +ALIGNTEXT16 +LLBL( sse_identity_loop ): + + PREFETCHNTA( REGOFF(32, ESI) ) + + MOVAPS( REGIND(ESI), XMM0 ) + ADD_L( EAX, ESI ) + + MOVAPS( XMM0, REGIND(EDI) ) + ADD_L( CONST(16), EDI ) + + DEC_L( ECX ) + JNZ( LLBL( sse_identity_loop ) ) + +LLBL( sse_identity_done ): + + POP_L( EDI ) + POP_L( ESI ) + RET +#endif diff --git a/src/mesa/x86/x86.c b/src/mesa/x86/x86.c new file mode 100644 index 0000000..16586b9 --- /dev/null +++ b/src/mesa/x86/x86.c @@ -0,0 +1,96 @@ +/* $Id: x86.c,v 1.25 2002/04/09 14:58:03 keithw Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Intel x86 assembly code by Josh Vanderhoof + */ + +#include "glheader.h" +#include "context.h" +#include "math/m_xform.h" +#include "tnl/t_context.h" + +#include "x86.h" +#include "common_x86_macros.h" + +#ifdef DEBUG +#include "math/m_debug.h" +#endif + + +#ifdef USE_X86_ASM +DECLARE_XFORM_GROUP( x86, 2 ) +DECLARE_XFORM_GROUP( x86, 3 ) +DECLARE_XFORM_GROUP( x86, 4 ) + + +extern GLvector4f * _ASMAPI +_mesa_x86_cliptest_points4( GLvector4f *clip_vec, + GLvector4f *proj_vec, + GLubyte clipMask[], + GLubyte *orMask, + GLubyte *andMask ); + +extern GLvector4f * _ASMAPI +_mesa_x86_cliptest_points4_np( GLvector4f *clip_vec, + GLvector4f *proj_vec, + GLubyte clipMask[], + GLubyte *orMask, + GLubyte *andMask ); + +extern void _ASMAPI +_mesa_v16_x86_cliptest_points4( GLfloat *first_vert, + GLfloat *last_vert, + GLubyte *or_mask, + GLubyte *and_mask, + GLubyte *clip_mask ); + +extern void _ASMAPI +_mesa_v16_x86_general_xform( GLfloat *dest, + const GLfloat *m, + const GLfloat *src, + GLuint src_stride, + GLuint count ); +#endif + + +void _mesa_init_x86_transform_asm( void ) +{ +#ifdef USE_X86_ASM + ASSIGN_XFORM_GROUP( x86, 2 ); + ASSIGN_XFORM_GROUP( x86, 3 ); + ASSIGN_XFORM_GROUP( x86, 4 ); + + _mesa_clip_tab[4] = _mesa_x86_cliptest_points4; + _mesa_clip_np_tab[4] = _mesa_x86_cliptest_points4_np; + +#ifdef DEBUG + _math_test_all_transform_functions( "x86" ); + _math_test_all_cliptest_functions( "x86" ); +#endif +#endif +} + diff --git a/src/mesa/x86/x86.h b/src/mesa/x86/x86.h new file mode 100644 index 0000000..a646aff --- /dev/null +++ b/src/mesa/x86/x86.h @@ -0,0 +1,36 @@ +/* $Id: x86.h,v 1.5 2002/04/09 14:58:03 keithw Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Intel x86 assembly code by Josh Vanderhoof + */ + +#ifndef __X86_H__ +#define __X86_H__ + +extern void _mesa_init_x86_transform_asm( void ); + +#endif diff --git a/src/mesa/x86/x86_cliptest.S b/src/mesa/x86/x86_cliptest.S new file mode 100644 index 0000000..745d1df --- /dev/null +++ b/src/mesa/x86/x86_cliptest.S @@ -0,0 +1,401 @@ +/* $Id: x86_cliptest.S,v 1.10 2002/06/11 01:26:58 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * NOTE: Avoid using spaces in between '(' ')' and arguments, especially + * with macros like CONST, LLBL that expand to CONCAT(...). Putting spaces + * in there will break the build on some platforms. + */ + +#include "matypes.h" +#include "clip_args.h" + +#define SRC0 REGOFF(0, ESI) +#define SRC1 REGOFF(4, ESI) +#define SRC2 REGOFF(8, ESI) +#define SRC3 REGOFF(12, ESI) +#define DST0 REGOFF(0, EDI) +#define DST1 REGOFF(4, EDI) +#define DST2 REGOFF(8, EDI) +#define DST3 REGOFF(12, EDI) +#define MAT0 REGOFF(0, EDX) +#define MAT1 REGOFF(4, EDX) +#define MAT2 REGOFF(8, EDX) +#define MAT3 REGOFF(12, EDX) + + +/* + * Table for clip test. + * + * bit6 = SRC3 < 0 + * bit5 = SRC2 < 0 + * bit4 = abs(S(2)) > abs(S(3)) + * bit3 = SRC1 < 0 + * bit2 = abs(S(1)) > abs(S(3)) + * bit1 = SRC0 < 0 + * bit0 = abs(S(0)) > abs(S(3)) + */ + + SEG_DATA + +clip_table: + D_BYTE 0x00, 0x01, 0x00, 0x02, 0x04, 0x05, 0x04, 0x06 + D_BYTE 0x00, 0x01, 0x00, 0x02, 0x08, 0x09, 0x08, 0x0a + D_BYTE 0x20, 0x21, 0x20, 0x22, 0x24, 0x25, 0x24, 0x26 + D_BYTE 0x20, 0x21, 0x20, 0x22, 0x28, 0x29, 0x28, 0x2a + D_BYTE 0x00, 0x01, 0x00, 0x02, 0x04, 0x05, 0x04, 0x06 + D_BYTE 0x00, 0x01, 0x00, 0x02, 0x08, 0x09, 0x08, 0x0a + D_BYTE 0x10, 0x11, 0x10, 0x12, 0x14, 0x15, 0x14, 0x16 + D_BYTE 0x10, 0x11, 0x10, 0x12, 0x18, 0x19, 0x18, 0x1a + D_BYTE 0x3f, 0x3d, 0x3f, 0x3e, 0x37, 0x35, 0x37, 0x36 + D_BYTE 0x3f, 0x3d, 0x3f, 0x3e, 0x3b, 0x39, 0x3b, 0x3a + D_BYTE 0x2f, 0x2d, 0x2f, 0x2e, 0x27, 0x25, 0x27, 0x26 + D_BYTE 0x2f, 0x2d, 0x2f, 0x2e, 0x2b, 0x29, 0x2b, 0x2a + D_BYTE 0x3f, 0x3d, 0x3f, 0x3e, 0x37, 0x35, 0x37, 0x36 + D_BYTE 0x3f, 0x3d, 0x3f, 0x3e, 0x3b, 0x39, 0x3b, 0x3a + D_BYTE 0x1f, 0x1d, 0x1f, 0x1e, 0x17, 0x15, 0x17, 0x16 + D_BYTE 0x1f, 0x1d, 0x1f, 0x1e, 0x1b, 0x19, 0x1b, 0x1a + + + SEG_TEXT + +/* + * _mesa_x86_cliptest_points4 + * + * AL: ormask + * AH: andmask + * EBX: temp0 + * ECX: temp1 + * EDX: clipmask[] + * ESI: clip[] + * EDI: proj[] + * EBP: temp2 + */ + +#if defined(__ELF__) && defined(__PIC__) && defined(GNU_ASSEMBLER) && !defined(ELFPIC) +#define ELFPIC +#endif + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_cliptest_points4 ) +GLNAME( _mesa_x86_cliptest_points4 ): + +#ifdef ELFPIC +#define FRAME_OFFSET 20 +#else +#define FRAME_OFFSET 16 +#endif + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBP ) + PUSH_L( EBX ) + +#ifdef ELFPIC + /* store pointer to clip_table on stack */ + CALL( LLBL(ctp4_get_eip) ) + ADD_L( CONST(_GLOBAL_OFFSET_TABLE_), EBX ) + MOV_L( REGOFF(clip_table@GOT, EBX), EBX ) + PUSH_L( EBX ) + JMP( LLBL(ctp4_clip_table_ready) ) + +LLBL(ctp4_get_eip): + /* store eip in ebx */ + MOV_L( REGIND(ESP), EBX ) + RET + +LLBL(ctp4_clip_table_ready): +#endif + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_CLIP, EDX ) + MOV_L( ARG_OR, EBX ) + + MOV_L( ARG_AND, EBP ) + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + MOV_L( EAX, ARG_SOURCE ) /* put stride in ARG_SOURCE */ + + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDX, ECX ) + + MOV_L( ECX, ARG_CLIP ) /* put clipmask + count in ARG_CLIP */ + CMP_L( ECX, EDX ) + + MOV_B( REGIND(EBX), AL ) + MOV_B( REGIND(EBP), AH ) + + JZ( LLBL(ctp4_finish) ) + +ALIGNTEXT16 +LLBL(ctp4_top): + + FLD1 /* F3 */ + FDIV_S( SRC3 ) /* GH: don't care about div-by-zero */ + + MOV_L( SRC3, EBP ) + MOV_L( SRC2, EBX ) + + XOR_L( ECX, ECX ) + ADD_L( EBP, EBP ) /* ebp = abs(S(3))*2 ; carry = sign of S(3) */ + + ADC_L( ECX, ECX ) + ADD_L( EBX, EBX ) /* ebx = abs(S(2))*2 ; carry = sign of S(2) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(2))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + MOV_L( SRC1, EBX ) + + ADD_L( EBX, EBX ) /* ebx = abs(S(1))*2 ; carry = sign of S(1) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(1))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + MOV_L( SRC0, EBX ) + + ADD_L( EBX, EBX ) /* ebx = abs(S(0))*2 ; carry = sign of S(0) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(0))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + +#ifdef ELFPIC + MOV_L( REGIND(ESP), EBP ) /* clip_table */ + + MOV_B( REGBI(EBP, ECX), CL ) +#else + MOV_B( REGOFF(clip_table,ECX), CL ) +#endif + + OR_B( CL, AL ) + AND_B( CL, AH ) + + TEST_B( CL, CL ) + MOV_B( CL, REGIND(EDX) ) + + JZ( LLBL(ctp4_proj) ) + +LLBL(ctp4_noproj): + + FSTP( ST(0) ) /* */ + + MOV_L( CONST(0), DST0 ) + MOV_L( CONST(0), DST1 ) + MOV_L( CONST(0), DST2 ) + MOV_L( CONST(0x3f800000), DST3 ) + + JMP( LLBL(ctp4_next) ) + +LLBL(ctp4_proj): + + FLD_S( SRC0 ) /* F0 F3 */ + FMUL2( ST(1), ST0 ) + + FLD_S( SRC1 ) /* F1 F0 F3 */ + FMUL2( ST(2), ST0 ) + + FLD_S( SRC2 ) /* F2 F1 F0 F3 */ + FMUL2( ST(3), ST0 ) + + FXCH( ST(2) ) /* F0 F1 F2 F3 */ + FSTP_S( DST0 ) /* F1 F2 F3 */ + FSTP_S( DST1 ) /* F2 F3 */ + FSTP_S( DST2 ) /* F3 */ + FSTP_S( DST3 ) /* */ + +LLBL(ctp4_next): + + INC_L( EDX ) + ADD_L( CONST(16), EDI ) + + ADD_L( ARG_SOURCE, ESI ) + CMP_L( EDX, ARG_CLIP ) + + JNZ( LLBL(ctp4_top) ) + + MOV_L( ARG_OR, ECX ) + MOV_L( ARG_AND, EDX ) + + MOV_B( AL, REGIND(ECX) ) + MOV_B( AH, REGIND(EDX) ) + +LLBL(ctp4_finish): + + MOV_L( ARG_DEST, EAX ) +#ifdef ELFPIC + POP_L( ESI ) /* discard ptr to clip_table */ +#endif + POP_L( EBX ) + POP_L( EBP ) + POP_L( EDI ) + POP_L( ESI ) + + RET + + + + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_cliptest_points4_np ) +GLNAME( _mesa_x86_cliptest_points4_np ): + +#ifdef ELFPIC +#define FRAME_OFFSET 20 +#else +#define FRAME_OFFSET 16 +#endif + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBP ) + PUSH_L( EBX ) + +#ifdef ELFPIC + /* store pointer to clip_table on stack */ + CALL( LLBL(ctp4_np_get_eip) ) + ADD_L( CONST(_GLOBAL_OFFSET_TABLE_), EBX ) + MOV_L( REGOFF(clip_table@GOT, EBX), EBX ) + PUSH_L( EBX ) + JMP( LLBL(ctp4_np_clip_table_ready) ) + +LLBL(ctp4_np_get_eip): + /* store eip in ebx */ + MOV_L( REGIND(ESP), EBX ) + RET + +LLBL(ctp4_np_clip_table_ready): +#endif + + MOV_L( ARG_SOURCE, ESI ) + /* slot */ + + MOV_L( ARG_CLIP, EDX ) + MOV_L( ARG_OR, EBX ) + + MOV_L( ARG_AND, EBP ) + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( EAX, ARG_DEST ) /* put stride in ARG_DEST */ + ADD_L( EDX, ECX ) + + MOV_L( ECX, EDI ) /* put clipmask + count in EDI */ + CMP_L( ECX, EDX ) + + MOV_B( REGIND(EBX), AL ) + MOV_B( REGIND(EBP), AH ) + + JZ( LLBL(ctp4_np_finish) ) + +ALIGNTEXT16 +LLBL(ctp4_np_top): + + MOV_L( SRC3, EBP ) + MOV_L( SRC2, EBX ) + + XOR_L( ECX, ECX ) + ADD_L( EBP, EBP ) /* ebp = abs(S(3))*2 ; carry = sign of S(3) */ + + ADC_L( ECX, ECX ) + ADD_L( EBX, EBX ) /* ebx = abs(S(2))*2 ; carry = sign of S(2) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(2))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + MOV_L( SRC1, EBX ) + + ADD_L( EBX, EBX ) /* ebx = abs(S(1))*2 ; carry = sign of S(1) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(1))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + MOV_L( SRC0, EBX ) + + ADD_L( EBX, EBX ) /* ebx = abs(S(0))*2 ; carry = sign of S(0) */ + + ADC_L( ECX, ECX ) + CMP_L( EBX, EBP ) /* carry = abs(S(0))*2 > abs(S(3))*2 */ + + ADC_L( ECX, ECX ) + +#ifdef ELFPIC + MOV_L( REGIND(ESP), EBP ) /* clip_table */ + + MOV_B( REGBI(EBP, ECX), CL ) +#else + MOV_B( REGOFF(clip_table,ECX), CL ) +#endif + + OR_B( CL, AL ) + AND_B( CL, AH ) + + TEST_B( CL, CL ) + MOV_B( CL, REGIND(EDX) ) + + INC_L( EDX ) + /* slot */ + + ADD_L( ARG_DEST, ESI ) + CMP_L( EDX, EDI ) + + JNZ( LLBL(ctp4_np_top) ) + + MOV_L( ARG_OR, ECX ) + MOV_L( ARG_AND, EDX ) + + MOV_B( AL, REGIND(ECX) ) + MOV_B( AH, REGIND(EDX) ) + +LLBL(ctp4_np_finish): + + MOV_L( ARG_SOURCE, EAX ) +#ifdef ELFPIC + POP_L( ESI ) /* discard ptr to clip_table */ +#endif + POP_L( EBX ) + POP_L( EBP ) + POP_L( EDI ) + POP_L( ESI ) + + RET diff --git a/src/mesa/x86/x86_xform2.S b/src/mesa/x86/x86_xform2.S new file mode 100644 index 0000000..b9528e1 --- /dev/null +++ b/src/mesa/x86/x86_xform2.S @@ -0,0 +1,563 @@ +/* $Id: x86_xform2.S,v 1.2 2002/03/07 21:40:08 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * NOTE: Avoid using spaces in between '(' ')' and arguments, especially + * with macros like CONST, LLBL that expand to CONCAT(...). Putting spaces + * in there will break the build on some platforms. + */ + +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FP_ONE 1065353216 +#define FP_ZERO 0 + +#define SRC0 REGOFF(0, ESI) +#define SRC1 REGOFF(4, ESI) +#define SRC2 REGOFF(8, ESI) +#define SRC3 REGOFF(12, ESI) +#define DST0 REGOFF(0, EDI) +#define DST1 REGOFF(4, EDI) +#define DST2 REGOFF(8, EDI) +#define DST3 REGOFF(12, EDI) +#define MAT0 REGOFF(0, EDX) +#define MAT1 REGOFF(4, EDX) +#define MAT2 REGOFF(8, EDX) +#define MAT3 REGOFF(12, EDX) +#define MAT4 REGOFF(16, EDX) +#define MAT5 REGOFF(20, EDX) +#define MAT6 REGOFF(24, EDX) +#define MAT7 REGOFF(28, EDX) +#define MAT8 REGOFF(32, EDX) +#define MAT9 REGOFF(36, EDX) +#define MAT10 REGOFF(40, EDX) +#define MAT11 REGOFF(44, EDX) +#define MAT12 REGOFF(48, EDX) +#define MAT13 REGOFF(52, EDX) +#define MAT14 REGOFF(56, EDX) +#define MAT15 REGOFF(60, EDX) + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_general ) +GLNAME( _mesa_x86_transform_points2_general ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_gr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p2_gr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + FLD_S( SRC0 ) /* F7 F6 F5 F4 */ + FMUL_S( MAT3 ) + + FLD_S( SRC1 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT6 ) + FLD_S( SRC1 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT7 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FXCH( ST(3) ) /* F4 F6 F5 F7 */ + FADD_S( MAT12 ) + FXCH( ST(2) ) /* F5 F6 F4 F7 */ + FADD_S( MAT13 ) + FXCH( ST(1) ) /* F6 F5 F4 F7 */ + FADD_S( MAT14 ) + FXCH( ST(3) ) /* F7 F5 F4 F6 */ + FADD_S( MAT15 ) + + FXCH( ST(2) ) /* F4 F5 F7 F6 */ + FSTP_S( DST0 ) /* F5 F7 F6 */ + FSTP_S( DST1 ) /* F7 F6 */ + FXCH( ST(1) ) /* F6 F7 */ + FSTP_S( DST2 ) /* F7 */ + FSTP_S( DST3 ) /* */ + +LLBL(x86_p2_gr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_gr_loop) ) + +LLBL(x86_p2_gr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_perspective ) +GLNAME( _mesa_x86_transform_points2_perspective ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_pr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + + MOV_L( MAT14, EBX ) + +ALIGNTEXT16 +LLBL(x86_p2_pr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F1 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F4 F1 */ + FSTP_S( DST0 ) /* F1 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + MOV_L( CONST(FP_ZERO), DST3 ) + +LLBL(x86_p2_pr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_pr_loop) ) + +LLBL(x86_p2_pr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_3d ) +GLNAME( _mesa_x86_transform_points2_3d ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_3dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p2_3dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + + FLD_S( SRC1 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT6 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + FXCH( ST(2) ) /* F4 F5 F6 */ + FADD_S( MAT12 ) + FXCH( ST(1) ) /* F5 F4 F6 */ + FADD_S( MAT13 ) + FXCH( ST(2) ) /* F6 F4 F5 */ + FADD_S( MAT14 ) + + FXCH( ST(1) ) /* F4 F6 F5 */ + FSTP_S( DST0 ) /* F6 F5 */ + FXCH( ST(1) ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + +LLBL(x86_p2_3dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_3dr_loop) ) + +LLBL(x86_p2_3dr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_3d_no_rot ) +GLNAME( _mesa_x86_transform_points2_3d_no_rot ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_3dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + + MOV_L( MAT14, EBX ) + +ALIGNTEXT16 +LLBL(x86_p2_3dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F1 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F4 F1 */ + FADD_S( MAT12 ) + FLD_S( MAT13 ) /* F5 F4 F1 */ + FXCH( ST(2) ) /* F1 F4 F5 */ + FADDP( ST0, ST(2) ) /* F4 F5 */ + + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + +LLBL(x86_p2_3dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_3dnrr_loop) ) + +LLBL(x86_p2_3dnrr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_2d ) +GLNAME( _mesa_x86_transform_points2_2d ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_2dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p2_2dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + + FLD_S( SRC1 ) /* F0 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F0 F1 F5 F4 */ + FADDP( ST0, ST(3) ) /* F1 F5 F4 */ + FADDP( ST0, ST(1) ) /* F5 F4 */ + + FXCH( ST(1) ) /* F4 F5 */ + FADD_S( MAT12 ) + FXCH( ST(1) ) /* F5 F4 */ + FADD_S( MAT13 ) + + FXCH( ST(1) ) /* F4 F5 */ + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + +LLBL(x86_p2_2dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_2dr_loop) ) + +LLBL(x86_p2_2dr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT4 +GLOBL GLNAME( _mesa_x86_transform_points2_2d_no_rot ) +GLNAME( _mesa_x86_transform_points2_2d_no_rot ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_2dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p2_2dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F1 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F4 F1 */ + FADD_S( MAT12 ) + FLD_S( MAT13 ) /* F5 F4 F1 */ + FXCH( ST(2) ) /* F1 F4 F5 */ + FADDP( ST0, ST(2) ) /* F4 F5 */ + + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + +LLBL(x86_p2_2dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_2dnrr_loop) ) + +LLBL(x86_p2_2dnrr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points2_identity ) +GLNAME( _mesa_x86_transform_points2_identity ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p2_ir_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_2), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(2), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + + CMP_L( ESI, EDI ) + JE( LLBL(x86_p2_ir_done) ) + +ALIGNTEXT16 +LLBL(x86_p2_ir_loop): + + MOV_L( SRC0, EBX ) + MOV_L( SRC1, EDX ) + + MOV_L( EBX, DST0 ) + MOV_L( EDX, DST1 ) + +LLBL(x86_p2_ir_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p2_ir_loop) ) + +LLBL(x86_p2_ir_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET diff --git a/src/mesa/x86/x86_xform3.S b/src/mesa/x86/x86_xform3.S new file mode 100644 index 0000000..d71a0fc --- /dev/null +++ b/src/mesa/x86/x86_xform3.S @@ -0,0 +1,633 @@ +/* $Id: x86_xform3.S,v 1.2 2002/03/07 21:40:08 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * NOTE: Avoid using spaces in between '(' ')' and arguments, especially + * with macros like CONST, LLBL that expand to CONCAT(...). Putting spaces + * in there will break the build on some platforms. + */ + +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FP_ONE 1065353216 +#define FP_ZERO 0 + +#define SRC0 REGOFF(0, ESI) +#define SRC1 REGOFF(4, ESI) +#define SRC2 REGOFF(8, ESI) +#define SRC3 REGOFF(12, ESI) +#define DST0 REGOFF(0, EDI) +#define DST1 REGOFF(4, EDI) +#define DST2 REGOFF(8, EDI) +#define DST3 REGOFF(12, EDI) +#define MAT0 REGOFF(0, EDX) +#define MAT1 REGOFF(4, EDX) +#define MAT2 REGOFF(8, EDX) +#define MAT3 REGOFF(12, EDX) +#define MAT4 REGOFF(16, EDX) +#define MAT5 REGOFF(20, EDX) +#define MAT6 REGOFF(24, EDX) +#define MAT7 REGOFF(28, EDX) +#define MAT8 REGOFF(32, EDX) +#define MAT9 REGOFF(36, EDX) +#define MAT10 REGOFF(40, EDX) +#define MAT11 REGOFF(44, EDX) +#define MAT12 REGOFF(48, EDX) +#define MAT13 REGOFF(52, EDX) +#define MAT14 REGOFF(56, EDX) +#define MAT15 REGOFF(60, EDX) + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_general ) +GLNAME( _mesa_x86_transform_points3_general ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_gr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_gr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + FLD_S( SRC0 ) /* F7 F6 F5 F4 */ + FMUL_S( MAT3 ) + + FLD_S( SRC1 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT6 ) + FLD_S( SRC1 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT7 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FLD_S( SRC2 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT10 ) + FLD_S( SRC2 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT11 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FXCH( ST(3) ) /* F4 F6 F5 F7 */ + FADD_S( MAT12 ) + FXCH( ST(2) ) /* F5 F6 F4 F7 */ + FADD_S( MAT13 ) + FXCH( ST(1) ) /* F6 F5 F4 F7 */ + FADD_S( MAT14 ) + FXCH( ST(3) ) /* F7 F5 F4 F6 */ + FADD_S( MAT15 ) + + FXCH( ST(2) ) /* F4 F5 F7 F6 */ + FSTP_S( DST0 ) /* F5 F7 F6 */ + FSTP_S( DST1 ) /* F7 F6 */ + FXCH( ST(1) ) /* F6 F7 */ + FSTP_S( DST2 ) /* F7 */ + FSTP_S( DST3 ) /* */ + +LLBL(x86_p3_gr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_gr_loop) ) + +LLBL(x86_p3_gr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_perspective ) +GLNAME( _mesa_x86_transform_points3_perspective ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_pr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_pr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F5 F4 */ + FMUL_S( MAT5 ) + + FLD_S( SRC2 ) /* F0 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F2 F1 F0 F5 F4 */ + FMUL_S( MAT10 ) + + FXCH( ST(2) ) /* F0 F1 F2 F5 F4 */ + FADDP( ST0, ST(4) ) /* F1 F2 F5 F4 */ + FADDP( ST0, ST(2) ) /* F2 F5 F4 */ + FLD_S( MAT14 ) /* F6 F2 F5 F4 */ + FXCH( ST(1) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + MOV_L( SRC2, EBX ) + XOR_L( CONST(-2147483648), EBX )/* change sign */ + + FXCH( ST(2) ) /* F4 F5 F6 */ + FSTP_S( DST0 ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + MOV_L( EBX, DST3 ) + +LLBL(x86_p3_pr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_pr_loop) ) + +LLBL(x86_p3_pr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_3d ) +GLNAME( _mesa_x86_transform_points3_3d ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_3dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_3dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + + FLD_S( SRC1 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT6 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + FLD_S( SRC2 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT10 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + FXCH( ST(2) ) /* F4 F5 F6 */ + FADD_S( MAT12 ) + FXCH( ST(1) ) /* F5 F4 F6 */ + FADD_S( MAT13 ) + FXCH( ST(2) ) /* F6 F4 F5 */ + FADD_S( MAT14 ) + + FXCH( ST(1) ) /* F4 F6 F5 */ + FSTP_S( DST0 ) /* F6 F5 */ + FXCH( ST(1) ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + +LLBL(x86_p3_3dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_3dr_loop) ) + +LLBL(x86_p3_3dr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_3d_no_rot ) +GLNAME( _mesa_x86_transform_points3_3d_no_rot ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_3dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_3dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F1 F4 */ + FMUL_S( MAT5 ) + + FLD_S( SRC2 ) /* F2 F1 F4 */ + FMUL_S( MAT10 ) + + FXCH( ST(2) ) /* F4 F1 F2 */ + FADD_S( MAT12 ) + FLD_S( MAT13 ) /* F5 F4 F1 F2 */ + FXCH( ST(2) ) /* F1 F4 F5 F2 */ + FADDP( ST0, ST(2) ) /* F4 F5 F2 */ + FLD_S( MAT14 ) /* F6 F4 F5 F2 */ + FXCH( ST(3) ) /* F2 F4 F5 F6 */ + FADDP( ST0, ST(3) ) /* F4 F5 F6 */ + + FSTP_S( DST0 ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + +LLBL(x86_p3_3dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_3dnrr_loop) ) + +LLBL(x86_p3_3dnrr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_2d ) +GLNAME( _mesa_x86_transform_points3_2d ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_2dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_2dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + + FLD_S( SRC1 ) /* F0 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F0 F1 F5 F4 */ + FADDP( ST0, ST(3) ) /* F1 F5 F4 */ + FADDP( ST0, ST(1) ) /* F5 F4 */ + + FXCH( ST(1) ) /* F4 F5 */ + FADD_S( MAT12 ) + FXCH( ST(1) ) /* F5 F4 */ + FADD_S( MAT13 ) + + MOV_L( SRC2, EBX ) + + FXCH( ST(1) ) /* F4 F5 */ + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + +LLBL(x86_p3_2dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_2dr_loop) ) + +LLBL(x86_p3_2dr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_2d_no_rot ) +GLNAME( _mesa_x86_transform_points3_2d_no_rot ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_2dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p3_2dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F1 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F4 F1 */ + FADD_S( MAT12 ) + FLD_S( MAT13 ) /* F5 F4 F1 */ + + FXCH( ST(2) ) /* F1 F4 F5 */ + FADDP( ST0, ST(2) ) /* F4 F5 */ + + MOV_L( SRC2, EBX ) + + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + +LLBL(x86_p3_2dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_2dnrr_loop) ) + +LLBL(x86_p3_2dnrr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points3_identity ) +GLNAME(_mesa_x86_transform_points3_identity ): + +#define FRAME_OFFSET 16 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + PUSH_L( EBP ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p3_ir_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_3), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(3), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + + CMP_L( ESI, EDI ) + JE( LLBL(x86_p3_ir_done) ) + +ALIGNTEXT16 +LLBL(x86_p3_ir_loop): + +#if 1 + MOV_L( SRC0, EBX ) + MOV_L( SRC1, EBP ) + MOV_L( SRC2, EDX ) + + MOV_L( EBX, DST0 ) + MOV_L( EBP, DST1 ) + MOV_L( EDX, DST2 ) +#else + FLD_S( SRC0 ) + FLD_S( SRC1 ) + FLD_S( SRC2 ) + + FSTP_S( DST2 ) + FSTP_S( DST1 ) + FSTP_S( DST0 ) +#endif + +LLBL(x86_p3_ir_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p3_ir_loop) ) + +LLBL(x86_p3_ir_done): + + POP_L( EBP ) + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET diff --git a/src/mesa/x86/x86_xform4.S b/src/mesa/x86/x86_xform4.S new file mode 100644 index 0000000..a5c1c1c --- /dev/null +++ b/src/mesa/x86/x86_xform4.S @@ -0,0 +1,666 @@ +/* $Id: x86_xform4.S,v 1.2 2002/03/07 21:40:08 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * NOTE: Avoid using spaces in between '(' ')' and arguments, especially + * with macros like CONST, LLBL that expand to CONCAT(...). Putting spaces + * in there will break the build on some platforms. + */ + +#include "matypes.h" +#include "xform_args.h" + + SEG_TEXT + +#define FP_ONE 1065353216 +#define FP_ZERO 0 + +#define SRC0 REGOFF(0, ESI) +#define SRC1 REGOFF(4, ESI) +#define SRC2 REGOFF(8, ESI) +#define SRC3 REGOFF(12, ESI) +#define DST0 REGOFF(0, EDI) +#define DST1 REGOFF(4, EDI) +#define DST2 REGOFF(8, EDI) +#define DST3 REGOFF(12, EDI) +#define MAT0 REGOFF(0, EDX) +#define MAT1 REGOFF(4, EDX) +#define MAT2 REGOFF(8, EDX) +#define MAT3 REGOFF(12, EDX) +#define MAT4 REGOFF(16, EDX) +#define MAT5 REGOFF(20, EDX) +#define MAT6 REGOFF(24, EDX) +#define MAT7 REGOFF(28, EDX) +#define MAT8 REGOFF(32, EDX) +#define MAT9 REGOFF(36, EDX) +#define MAT10 REGOFF(40, EDX) +#define MAT11 REGOFF(44, EDX) +#define MAT12 REGOFF(48, EDX) +#define MAT13 REGOFF(52, EDX) +#define MAT14 REGOFF(56, EDX) +#define MAT15 REGOFF(60, EDX) + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_general ) +GLNAME( _mesa_x86_transform_points4_general ): + +#define FRAME_OFFSET 8 + PUSH_L( ESI ) + PUSH_L( EDI ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_gr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_gr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + FLD_S( SRC0 ) /* F7 F6 F5 F4 */ + FMUL_S( MAT3 ) + + FLD_S( SRC1 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT6 ) + FLD_S( SRC1 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT7 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FLD_S( SRC2 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT10 ) + FLD_S( SRC2 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT11 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FLD_S( SRC3 ) /* F0 F7 F6 F5 F4 */ + FMUL_S( MAT12 ) + FLD_S( SRC3 ) /* F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT13 ) + FLD_S( SRC3 ) /* F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT14 ) + FLD_S( SRC3 ) /* F3 F2 F1 F0 F7 F6 F5 F4 */ + FMUL_S( MAT15 ) + + FXCH( ST(3) ) /* F0 F2 F1 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(7) ) /* F2 F1 F3 F7 F6 F5 F4 */ + FXCH( ST(1) ) /* F1 F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F2 F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F3 F7 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F7 F6 F5 F4 */ + + FXCH( ST(3) ) /* F4 F6 F5 F7 */ + FSTP_S( DST0 ) /* F6 F5 F7 */ + FXCH( ST(1) ) /* F5 F6 F7 */ + FSTP_S( DST1 ) /* F6 F7 */ + FSTP_S( DST2 ) /* F7 */ + FSTP_S( DST3 ) /* */ + +LLBL(x86_p4_gr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_gr_loop) ) + +LLBL(x86_p4_gr_done): + + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_perspective ) +GLNAME( _mesa_x86_transform_points4_perspective ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_pr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_pr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F5 F4 */ + FMUL_S( MAT5 ) + + FLD_S( SRC2 ) /* F0 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F6 F1 F0 F5 F4 */ + FMUL_S( MAT10 ) + + FXCH( ST(2) ) /* F0 F1 F6 F5 F4 */ + FADDP( ST0, ST(4) ) /* F1 F6 F5 F4 */ + FADDP( ST0, ST(2) ) /* F6 F5 F4 */ + + FLD_S( SRC3 ) /* F2 F6 F5 F4 */ + FMUL_S( MAT14 ) + + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + MOV_L( SRC2, EBX ) + XOR_L( CONST(-2147483648), EBX )/* change sign */ + + FXCH( ST(2) ) /* F4 F5 F6 */ + FSTP_S( DST0 ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + MOV_L( EBX, DST3 ) + +LLBL(x86_p4_pr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_pr_loop) ) + +LLBL(x86_p4_pr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_3d ) +GLNAME( _mesa_x86_transform_points4_3d ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_3dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_3dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + FLD_S( SRC0 ) /* F6 F5 F4 */ + FMUL_S( MAT2 ) + + FLD_S( SRC1 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT5 ) + FLD_S( SRC1 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT6 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + FLD_S( SRC2 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT8 ) + FLD_S( SRC2 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT9 ) + FLD_S( SRC2 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT10 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + FLD_S( SRC3 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT12 ) + FLD_S( SRC3 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT13 ) + FLD_S( SRC3 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT14 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + MOV_L( SRC3, EBX ) + + FXCH( ST(2) ) /* F4 F5 F6 */ + FSTP_S( DST0 ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + MOV_L( EBX, DST3 ) + +LLBL(x86_p4_3dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_3dr_loop) ) + +LLBL(x86_p4_3dr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME(_mesa_x86_transform_points4_3d_no_rot) +GLNAME(_mesa_x86_transform_points4_3d_no_rot): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_3dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_3dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F5 F4 */ + FMUL_S( MAT5 ) + + FLD_S( SRC2 ) /* F6 F5 F4 */ + FMUL_S( MAT10 ) + + FLD_S( SRC3 ) /* F0 F6 F5 F4 */ + FMUL_S( MAT12 ) + FLD_S( SRC3 ) /* F1 F0 F6 F5 F4 */ + FMUL_S( MAT13 ) + FLD_S( SRC3 ) /* F2 F1 F0 F6 F5 F4 */ + FMUL_S( MAT14 ) + + FXCH( ST(2) ) /* F0 F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(5) ) /* F1 F2 F6 F5 F4 */ + FADDP( ST0, ST(3) ) /* F2 F6 F5 F4 */ + FADDP( ST0, ST(1) ) /* F6 F5 F4 */ + + MOV_L( SRC3, EBX ) + + FXCH( ST(2) ) /* F4 F5 F6 */ + FSTP_S( DST0 ) /* F5 F6 */ + FSTP_S( DST1 ) /* F6 */ + FSTP_S( DST2 ) /* */ + MOV_L( EBX, DST3 ) + +LLBL(x86_p4_3dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_3dnrr_loop) ) + +LLBL(x86_p4_3dnrr_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_2d ) +GLNAME( _mesa_x86_transform_points4_2d ): + +#define FRAME_OFFSET 16 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + PUSH_L( EBP ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_2dr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_2dr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + FLD_S( SRC0 ) /* F5 F4 */ + FMUL_S( MAT1 ) + + FLD_S( SRC1 ) /* F0 F5 F4 */ + FMUL_S( MAT4 ) + FLD_S( SRC1 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT5 ) + + FXCH( ST(1) ) /* F0 F1 F5 F4 */ + FADDP( ST0, ST(3) ) /* F1 F5 F4 */ + FADDP( ST0, ST(1) ) /* F5 F4 */ + + FLD_S( SRC3 ) /* F0 F5 F4 */ + FMUL_S( MAT12 ) + FLD_S( SRC3 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT13 ) + + FXCH( ST(1) ) /* F0 F1 F5 F4 */ + FADDP( ST0, ST(3) ) /* F1 F5 F4 */ + FADDP( ST0, ST(1) ) /* F5 F4 */ + + MOV_L( SRC2, EBX ) + MOV_L( SRC3, EBP ) + + FXCH( ST(1) ) /* F4 F5 */ + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + MOV_L( EBP, DST3 ) + +LLBL(x86_p4_2dr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_2dr_loop) ) + +LLBL(x86_p4_2dr_done): + + POP_L( EBP ) + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_2d_no_rot ) +GLNAME( _mesa_x86_transform_points4_2d_no_rot ): + +#define FRAME_OFFSET 16 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + PUSH_L( EBP ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_2dnrr_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + +ALIGNTEXT16 +LLBL(x86_p4_2dnrr_loop): + + FLD_S( SRC0 ) /* F4 */ + FMUL_S( MAT0 ) + + FLD_S( SRC1 ) /* F5 F4 */ + FMUL_S( MAT5 ) + + FLD_S( SRC3 ) /* F0 F5 F4 */ + FMUL_S( MAT12 ) + FLD_S( SRC3 ) /* F1 F0 F5 F4 */ + FMUL_S( MAT13 ) + + FXCH( ST(1) ) /* F0 F1 F5 F4 */ + FADDP( ST0, ST(3) ) /* F1 F5 F4 */ + FADDP( ST0, ST(1) ) /* F5 F4 */ + + MOV_L( SRC2, EBX ) + MOV_L( SRC3, EBP ) + + FXCH( ST(1) ) /* F4 F5 */ + FSTP_S( DST0 ) /* F5 */ + FSTP_S( DST1 ) /* */ + MOV_L( EBX, DST2 ) + MOV_L( EBP, DST3 ) + +LLBL(x86_p4_2dnrr_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_2dnrr_loop) ) + +LLBL(x86_p4_2dnrr_done): + + POP_L( EBP ) + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET +#undef FRAME_OFFSET + + + + +ALIGNTEXT16 +GLOBL GLNAME( _mesa_x86_transform_points4_identity ) +GLNAME( _mesa_x86_transform_points4_identity ): + +#define FRAME_OFFSET 12 + PUSH_L( ESI ) + PUSH_L( EDI ) + PUSH_L( EBX ) + + MOV_L( ARG_SOURCE, ESI ) + MOV_L( ARG_DEST, EDI ) + + MOV_L( ARG_MATRIX, EDX ) + MOV_L( REGOFF(V4F_COUNT, ESI), ECX ) + + TEST_L( ECX, ECX ) + JZ( LLBL(x86_p4_ir_done) ) + + MOV_L( REGOFF(V4F_STRIDE, ESI), EAX ) + OR_L( CONST(VEC_SIZE_4), REGOFF(V4F_FLAGS, EDI) ) + + MOV_L( ECX, REGOFF(V4F_COUNT, EDI) ) + MOV_L( CONST(4), REGOFF(V4F_SIZE, EDI) ) + + SHL_L( CONST(4), ECX ) + MOV_L( REGOFF(V4F_START, ESI), ESI ) + + MOV_L( REGOFF(V4F_START, EDI), EDI ) + ADD_L( EDI, ECX ) + + CMP_L( ESI, EDI ) + JE( LLBL(x86_p4_ir_done) ) + +ALIGNTEXT16 +LLBL(x86_p4_ir_loop): + + MOV_L( SRC0, EBX ) + MOV_L( SRC1, EDX ) + + MOV_L( EBX, DST0 ) + MOV_L( EDX, DST1 ) + + MOV_L( SRC2, EBX ) + MOV_L( SRC3, EDX ) + + MOV_L( EBX, DST2 ) + MOV_L( EDX, DST3 ) + +LLBL(x86_p4_ir_skip): + + ADD_L( CONST(16), EDI ) + ADD_L( EAX, ESI ) + CMP_L( ECX, EDI ) + JNE( LLBL(x86_p4_ir_loop) ) + +LLBL(x86_p4_ir_done): + + POP_L( EBX ) + POP_L( EDI ) + POP_L( ESI ) + RET diff --git a/src/mesa/x86/xform_args.h b/src/mesa/x86/xform_args.h new file mode 100644 index 0000000..89a0420 --- /dev/null +++ b/src/mesa/x86/xform_args.h @@ -0,0 +1,52 @@ +/* $Id: xform_args.h,v 1.5 2002/10/29 20:28:58 brianp Exp $ */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Transform function interface for assembly code. Simply define + * FRAME_OFFSET to the number of bytes pushed onto the stack before + * using the ARG_* argument macros. + * + * Gareth Hughes + */ + +#ifndef __XFORM_ARGS_H__ +#define __XFORM_ARGS_H__ + +/* Offsets for transform_func arguments + * + * typedef void (*transform_func)( GLvector4f *to_vec, + * const GLfloat m[16], + * const GLvector4f *from_vec ); + */ +#define OFFSET_DEST 4 +#define OFFSET_MATRIX 8 +#define OFFSET_SOURCE 12 + +#define ARG_DEST REGOFF(FRAME_OFFSET+OFFSET_DEST, ESP) +#define ARG_MATRIX REGOFF(FRAME_OFFSET+OFFSET_MATRIX, ESP) +#define ARG_SOURCE REGOFF(FRAME_OFFSET+OFFSET_SOURCE, ESP) + +#endif |