Subversion Repositories shark

Compare Revisions

Ignore whitespace Rev 133 → Rev 134

/shark/trunk/ports/mesa/src/math/m_debug_xform.c
File deleted
/shark/trunk/ports/mesa/src/math/m_xform_tmp.h
File deleted
/shark/trunk/ports/mesa/src/math/m_debug_norm.c
File deleted
/shark/trunk/ports/mesa/src/math/m_debug_clip.c
File deleted
/shark/trunk/ports/mesa/src/math/m_xform.c
1,4 → 1,4
/* $Id: m_xform.c,v 1.2 2003-03-13 12:20:29 giacomo Exp $ */
/* $Id: m_xform.c,v 1.3 2003-04-24 13:37:47 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
81,7 → 81,7
#define STRIDE_LOOP for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) )
#define LOOP for ( i = 0 ; i < n ; i++ )
#define ARGS
#include "m_xform_tmp.h"
#include "m_tmp_xform.h"
#include "m_clip_tmp.h"
#include "m_norm_tmp.h"
#include "m_dotprod_tmp.h"
/shark/trunk/ports/mesa/src/math/m_clip_debug.c
0,0 → 1,367
/* $Id: m_clip_debug.c,v 1.1 2003-04-24 13:37:46 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
* Version: 5.0
*
* 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
*/
 
#include "glheader.h"
#include "context.h"
#include "macros.h"
#include "imports.h"
 
#include "m_matrix.h"
#include "m_xform.h"
 
#include "m_debug.h"
#include "m_debug_util.h"
 
#ifdef DEBUG /* This code only used for debugging */
 
static clip_func *clip_tab[2] = {
_mesa_clip_tab,
_mesa_clip_np_tab
};
static char *cnames[2] = {
"_mesa_clip_tab",
"_mesa_clip_np_tab"
};
#ifdef RUN_DEBUG_BENCHMARK
static char *cstrings[2] = {
"clip, perspective divide",
"clip, no divide"
};
#endif
 
 
/* =============================================================
* Reference cliptests
*/
 
static GLvector4f *ref_cliptest_points4( GLvector4f *clip_vec,
GLvector4f *proj_vec,
GLubyte clipMask[],
GLubyte *orMask,
GLubyte *andMask )
{
const GLuint stride = clip_vec->stride;
const GLuint count = clip_vec->count;
const GLfloat *from = (GLfloat *)clip_vec->start;
GLuint c = 0;
GLfloat (*vProj)[4] = (GLfloat (*)[4])proj_vec->start;
GLubyte tmpAndMask = *andMask;
GLubyte tmpOrMask = *orMask;
GLuint i;
for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
const GLfloat cx = from[0];
const GLfloat cy = from[1];
const GLfloat cz = from[2];
const GLfloat cw = from[3];
GLubyte mask = 0;
if ( -cx + cw < 0 ) mask |= CLIP_RIGHT_BIT;
if ( cx + cw < 0 ) mask |= CLIP_LEFT_BIT;
if ( -cy + cw < 0 ) mask |= CLIP_TOP_BIT;
if ( cy + cw < 0 ) mask |= CLIP_BOTTOM_BIT;
if ( -cz + cw < 0 ) mask |= CLIP_FAR_BIT;
if ( cz + cw < 0 ) mask |= CLIP_NEAR_BIT;
clipMask[i] = mask;
if ( mask ) {
c++;
tmpAndMask &= mask;
tmpOrMask |= mask;
vProj[i][0] = 0;
vProj[i][1] = 0;
vProj[i][2] = 0;
vProj[i][3] = 1;
} else {
GLfloat oow = 1.0F / cw;
vProj[i][0] = cx * oow;
vProj[i][1] = cy * oow;
vProj[i][2] = cz * oow;
vProj[i][3] = oow;
}
}
 
*orMask = tmpOrMask;
*andMask = (GLubyte) (c < count ? 0 : tmpAndMask);
 
proj_vec->flags |= VEC_SIZE_4;
proj_vec->size = 4;
proj_vec->count = clip_vec->count;
return proj_vec;
}
 
/* Keep these here for now, even though we don't use them...
*/
static GLvector4f *ref_cliptest_points3( GLvector4f *clip_vec,
GLvector4f *proj_vec,
GLubyte clipMask[],
GLubyte *orMask,
GLubyte *andMask )
{
const GLuint stride = clip_vec->stride;
const GLuint count = clip_vec->count;
const GLfloat *from = (GLfloat *)clip_vec->start;
 
GLubyte tmpOrMask = *orMask;
GLubyte tmpAndMask = *andMask;
GLuint i;
for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
const GLfloat cx = from[0], cy = from[1], cz = from[2];
GLubyte mask = 0;
if ( cx > 1.0 ) mask |= CLIP_RIGHT_BIT;
else if ( cx < -1.0 ) mask |= CLIP_LEFT_BIT;
if ( cy > 1.0 ) mask |= CLIP_TOP_BIT;
else if ( cy < -1.0 ) mask |= CLIP_BOTTOM_BIT;
if ( cz > 1.0 ) mask |= CLIP_FAR_BIT;
else if ( cz < -1.0 ) mask |= CLIP_NEAR_BIT;
clipMask[i] = mask;
tmpOrMask |= mask;
tmpAndMask &= mask;
}
 
*orMask = tmpOrMask;
*andMask = tmpAndMask;
return clip_vec;
}
 
static GLvector4f * ref_cliptest_points2( GLvector4f *clip_vec,
GLvector4f *proj_vec,
GLubyte clipMask[],
GLubyte *orMask,
GLubyte *andMask )
{
const GLuint stride = clip_vec->stride;
const GLuint count = clip_vec->count;
const GLfloat *from = (GLfloat *)clip_vec->start;
 
GLubyte tmpOrMask = *orMask;
GLubyte tmpAndMask = *andMask;
GLuint i;
for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
const GLfloat cx = from[0], cy = from[1];
GLubyte mask = 0;
if ( cx > 1.0 ) mask |= CLIP_RIGHT_BIT;
else if ( cx < -1.0 ) mask |= CLIP_LEFT_BIT;
if ( cy > 1.0 ) mask |= CLIP_TOP_BIT;
else if ( cy < -1.0 ) mask |= CLIP_BOTTOM_BIT;
clipMask[i] = mask;
tmpOrMask |= mask;
tmpAndMask &= mask;
}
 
*orMask = tmpOrMask;
*andMask = tmpAndMask;
return clip_vec;
}
 
static clip_func ref_cliptest[5] = {
0,
0,
ref_cliptest_points2,
ref_cliptest_points3,
ref_cliptest_points4
};
 
 
/* =============================================================
* Cliptest tests
*/
 
static GLfloat s[TEST_COUNT][4] ALIGN16;
static GLfloat d[TEST_COUNT][4] ALIGN16;
static GLfloat r[TEST_COUNT][4] ALIGN16;
 
static int test_cliptest_function( clip_func func, int np,
int psize, long *cycles )
{
GLvector4f source[1], dest[1], ref[1];
GLubyte dm[TEST_COUNT], dco, dca;
GLubyte rm[TEST_COUNT], rco, rca;
int i, j;
#ifdef RUN_DEBUG_BENCHMARK
int cycle_i; /* the counter for the benchmarks we run */
#endif
 
(void) cycles;
 
if ( psize > 4 ) {
_mesa_problem( NULL, "test_cliptest_function called with psize > 4\n" );
return 0;
}
 
for ( i = 0 ; i < TEST_COUNT ; i++) {
ASSIGN_4V( d[i], 0.0, 0.0, 0.0, 1.0 );
ASSIGN_4V( s[i], 0.0, 0.0, 0.0, 1.0 );
for ( j = 0 ; j < psize ; j++ )
s[i][j] = rnd();
}
 
source->data = (GLfloat(*)[4])s;
source->start = (GLfloat *)s;
source->count = TEST_COUNT;
source->stride = sizeof(s[0]);
source->size = 4;
source->flags = 0;
 
dest->data = (GLfloat(*)[4])d;
dest->start = (GLfloat *)d;
dest->count = TEST_COUNT;
dest->stride = sizeof(float[4]);
dest->size = 0;
dest->flags = 0;
 
ref->data = (GLfloat(*)[4])r;
ref->start = (GLfloat *)r;
ref->count = TEST_COUNT;
ref->stride = sizeof(float[4]);
ref->size = 0;
ref->flags = 0;
 
dco = rco = 0;
dca = rca = CLIP_ALL_BITS;
 
ref_cliptest[psize]( source, ref, rm, &rco, &rca );
 
if ( mesa_profile ) {
BEGIN_RACE( *cycles );
func( source, dest, dm, &dco, &dca );
END_RACE( *cycles );
}
else {
func( source, dest, dm, &dco, &dca );
}
 
if ( dco != rco ) {
_mesa_printf(NULL, "\n-----------------------------\n" );
_mesa_printf(NULL, "dco = 0x%02x rco = 0x%02x\n", dco, rco );
return 0;
}
if ( dca != rca ) {
_mesa_printf(NULL, "\n-----------------------------\n" );
_mesa_printf(NULL, "dca = 0x%02x rca = 0x%02x\n", dca, rca );
return 0;
}
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
if ( dm[i] != rm[i] ) {
_mesa_printf(NULL, "\n-----------------------------\n" );
_mesa_printf(NULL, "(i = %i)\n", i );
_mesa_printf(NULL, "dm = 0x%02x rm = 0x%02x\n", dm[i], rm[i] );
return 0;
}
}
 
/* Only verify output on projected points4 case. FIXME: Do we need
* to test other cases?
*/
if ( np || psize < 4 )
return 1;
 
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
_mesa_printf(NULL, "\n-----------------------------\n" );
_mesa_printf(NULL, "(i = %i, j = %i) dm = 0x%02x rm = 0x%02x\n",
i, j, dm[i], rm[i] );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][0], r[i][0], r[i][0]-d[i][0],
MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][1], r[i][1], r[i][1]-d[i][1],
MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][2], r[i][2], r[i][2]-d[i][2],
MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][3], r[i][3], r[i][3]-d[i][3],
MAX_PRECISION - significand_match( d[i][3], r[i][3] ) );
return 0;
}
}
}
 
return 1;
}
 
void _math_test_all_cliptest_functions( char *description )
{
int np, psize;
long benchmark_tab[2][4];
static int first_time = 1;
 
if ( first_time ) {
first_time = 0;
mesa_profile = _mesa_getenv( "MESA_PROFILE" );
}
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
if ( !counter_overhead ) {
INIT_COUNTER();
_mesa_printf(NULL, "counter overhead: %ld cycles\n\n", counter_overhead );
}
_mesa_printf(NULL, "cliptest results after hooking in %s functions:\n", description );
}
#endif
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
_mesa_printf(NULL, "\n\t" );
for ( psize = 2 ; psize <= 4 ; psize++ ) {
_mesa_printf(NULL, " p%d\t", psize );
}
_mesa_printf(NULL, "\n--------------------------------------------------------\n\t" );
}
#endif
 
for ( np = 0 ; np < 2 ; np++ ) {
for ( psize = 2 ; psize <= 4 ; psize++ ) {
clip_func func = clip_tab[np][psize];
long *cycles = &(benchmark_tab[np][psize-1]);
 
if ( test_cliptest_function( func, np, psize, cycles ) == 0 ) {
char buf[100];
_mesa_sprintf(NULL, buf, "%s[%d] failed test (%s)",
cnames[np], psize, description );
_mesa_problem( NULL, buf );
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, " %li\t", benchmark_tab[np][psize-1] );
#endif
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, " | [%s]\n\t", cstrings[np] );
#endif
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, "\n" );
#endif
}
 
 
#endif /* DEBUG */
/shark/trunk/ports/mesa/src/math/m_xform_debug.c
0,0 → 1,335
/* $Id: m_xform_debug.c,v 1.1 2003-04-24 13:37:47 giacomo 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.
*/
 
/*
* Updated for P6 architecture by Gareth Hughes.
*/
 
#include "glheader.h"
#include "context.h"
#include "macros.h"
#include "imports.h"
 
#include "m_matrix.h"
#include "m_xform.h"
 
#include "m_debug.h"
#include "m_debug_util.h"
 
#ifdef DEBUG /* This code only used for debugging */
 
 
/* Overhead of profiling counter in cycles. Automatically adjusted to
* your machine at run time - counter initialization should give very
* consistent results.
*/
long counter_overhead = 0;
 
/* This is the value of the environment variable MESA_PROFILE, and is
* used to determine if we should benchmark the functions as well as
* verify their correctness.
*/
char *mesa_profile = NULL;
 
 
static int m_general[16] = {
VAR, VAR, VAR, VAR,
VAR, VAR, VAR, VAR,
VAR, VAR, VAR, VAR,
VAR, VAR, VAR, VAR
};
static int m_identity[16] = {
ONE, NIL, NIL, NIL,
NIL, ONE, NIL, NIL,
NIL, NIL, ONE, NIL,
NIL, NIL, NIL, ONE
};
static int m_2d[16] = {
VAR, VAR, NIL, VAR,
VAR, VAR, NIL, VAR,
NIL, NIL, ONE, NIL,
NIL, NIL, NIL, ONE
};
static int m_2d_no_rot[16] = {
VAR, NIL, NIL, VAR,
NIL, VAR, NIL, VAR,
NIL, NIL, ONE, NIL,
NIL, NIL, NIL, ONE
};
static int m_3d[16] = {
VAR, VAR, VAR, VAR,
VAR, VAR, VAR, VAR,
VAR, VAR, VAR, VAR,
NIL, NIL, NIL, ONE
};
static int m_3d_no_rot[16] = {
VAR, NIL, NIL, VAR,
NIL, VAR, NIL, VAR,
NIL, NIL, VAR, VAR,
NIL, NIL, NIL, ONE
};
static int m_perspective[16] = {
VAR, NIL, VAR, NIL,
NIL, VAR, VAR, NIL,
NIL, NIL, VAR, VAR,
NIL, NIL, NEG, NIL
};
static int *templates[7] = {
m_general,
m_identity,
m_3d_no_rot,
m_perspective,
m_2d,
m_2d_no_rot,
m_3d
};
static int mtypes[7] = {
MATRIX_GENERAL,
MATRIX_IDENTITY,
MATRIX_3D_NO_ROT,
MATRIX_PERSPECTIVE,
MATRIX_2D,
MATRIX_2D_NO_ROT,
MATRIX_3D
};
static char *mstrings[7] = {
"MATRIX_GENERAL",
"MATRIX_IDENTITY",
"MATRIX_3D_NO_ROT",
"MATRIX_PERSPECTIVE",
"MATRIX_2D",
"MATRIX_2D_NO_ROT",
"MATRIX_3D"
};
 
 
/* =============================================================
* Reference transformations
*/
 
static void ref_transform( GLvector4f *dst,
const GLmatrix *mat,
const GLvector4f *src )
{
GLuint i;
GLfloat *s = (GLfloat *)src->start;
GLfloat (*d)[4] = (GLfloat (*)[4])dst->start;
const GLfloat *m = mat->m;
 
for ( i = 0 ; i < src->count ; i++ ) {
TRANSFORM_POINT( d[i], m, s );
s = (GLfloat *)((char *)s + src->stride);
}
}
 
 
/* =============================================================
* Vertex transformation tests
*/
 
static void init_matrix( GLfloat *m )
{
m[0] = 63.0; m[4] = 43.0; m[ 8] = 29.0; m[12] = 43.0;
m[1] = 55.0; m[5] = 17.0; m[ 9] = 31.0; m[13] = 7.0;
m[2] = 44.0; m[6] = 9.0; m[10] = 7.0; m[14] = 3.0;
m[3] = 11.0; m[7] = 23.0; m[11] = 91.0; m[15] = 9.0;
}
 
static GLfloat s[TEST_COUNT][4] ALIGN16;
static GLfloat d[TEST_COUNT][4] ALIGN16;
static GLfloat r[TEST_COUNT][4] ALIGN16;
 
static int test_transform_function( transform_func func, int psize,
int mtype, long *cycles )
{
GLvector4f source[1], dest[1], ref[1];
GLmatrix mat[1];
GLfloat *m;
int i, j;
#ifdef RUN_DEBUG_BENCHMARK
int cycle_i; /* the counter for the benchmarks we run */
#endif
 
(void) cycles;
 
if ( psize > 4 ) {
_mesa_problem( NULL, "test_transform_function called with psize > 4\n" );
return 0;
}
 
mat->m = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
mat->type = mtypes[mtype];
 
m = mat->m;
ASSERT( ((GLuint)m & 15) == 0 );
 
init_matrix( m );
 
for ( i = 0 ; i < 4 ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
switch ( templates[mtype][i * 4 + j] ) {
case NIL:
m[j * 4 + i] = 0.0;
break;
case ONE:
m[j * 4 + i] = 1.0;
break;
case NEG:
m[j * 4 + i] = -1.0;
break;
case VAR:
break;
default:
abort();
}
}
}
 
for ( i = 0 ; i < TEST_COUNT ; i++) {
ASSIGN_4V( d[i], 0.0, 0.0, 0.0, 1.0 );
ASSIGN_4V( s[i], 0.0, 0.0, 0.0, 1.0 );
for ( j = 0 ; j < psize ; j++ )
s[i][j] = rnd();
}
 
source->data = (GLfloat(*)[4])s;
source->start = (GLfloat *)s;
source->count = TEST_COUNT;
source->stride = sizeof(s[0]);
source->size = 4;
source->flags = 0;
 
dest->data = (GLfloat(*)[4])d;
dest->start = (GLfloat *)d;
dest->count = TEST_COUNT;
dest->stride = sizeof(float[4]);
dest->size = 0;
dest->flags = 0;
 
ref->data = (GLfloat(*)[4])r;
ref->start = (GLfloat *)r;
ref->count = TEST_COUNT;
ref->stride = sizeof(float[4]);
ref->size = 0;
ref->flags = 0;
 
ref_transform( ref, mat, source );
 
if ( mesa_profile ) {
BEGIN_RACE( *cycles );
func( dest, mat->m, source );
END_RACE( *cycles );
}
else {
func( dest, mat->m, source );
}
 
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
_mesa_printf(NULL, "-----------------------------\n" );
_mesa_printf(NULL, "(i = %i, j = %i)\n", i, j );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][0], r[i][0], r[i][0]-d[i][0],
MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][1], r[i][1], r[i][1]-d[i][1],
MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][2], r[i][2], r[i][2]-d[i][2],
MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
_mesa_printf(NULL, "%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][3], r[i][3], r[i][3]-d[i][3],
MAX_PRECISION - significand_match( d[i][3], r[i][3] ) );
return 0;
}
}
}
 
ALIGN_FREE( mat->m );
return 1;
}
 
void _math_test_all_transform_functions( char *description )
{
int psize, mtype;
long benchmark_tab[4][7];
static int first_time = 1;
 
if ( first_time ) {
first_time = 0;
mesa_profile = getenv( "MESA_PROFILE" );
}
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
if ( !counter_overhead ) {
INIT_COUNTER();
_mesa_printf(NULL, "counter overhead: %ld cycles\n\n", counter_overhead );
}
_mesa_printf(NULL, "transform results after hooking in %s functions:\n", description );
}
#endif
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
_mesa_printf(NULL, "\n" );
for ( psize = 1 ; psize <= 4 ; psize++ ) {
_mesa_printf(NULL, " p%d\t", psize );
}
_mesa_printf(NULL, "\n--------------------------------------------------------\n" );
}
#endif
 
for ( mtype = 0 ; mtype < 7 ; mtype++ ) {
for ( psize = 1 ; psize <= 4 ; psize++ ) {
transform_func func = _mesa_transform_tab[psize][mtypes[mtype]];
long *cycles = &(benchmark_tab[psize-1][mtype]);
 
if ( test_transform_function( func, psize, mtype, cycles ) == 0 ) {
char buf[100];
_mesa_sprintf(NULL, buf, "_mesa_transform_tab[0][%d][%s] failed test (%s)",
psize, mstrings[mtype], description );
_mesa_problem( NULL, buf );
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, " %li\t", benchmark_tab[psize-1][mtype] );
#endif
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, " | [%s]\n", mstrings[mtype] );
#endif
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile )
_mesa_printf(NULL, "\n" );
#endif
}
 
 
#endif /* DEBUG */
/shark/trunk/ports/mesa/src/math/m_tmp_xform.h
0,0 → 1,807
/* $Id: m_tmp_xform.h,v 1.1 2003-04-24 13:37:47 giacomo 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.
*/
 
/*
* New (3.1) transformation code written by Keith Whitwell.
*/
 
 
/*----------------------------------------------------------------------
* Begin Keith's new code
*
*----------------------------------------------------------------------
*/
 
/* KW: Fixed stride, now measured in bytes as is the OpenGL array stride.
*/
 
/* KW: These are now parameterized to produce two versions, one
* which transforms all incoming points, and a second which
* takes notice of a cullmask array, and only transforms
* unculled vertices.
*/
 
/* KW: 1-vectors can sneak into the texture pipeline via the array
* interface. These functions are here because I want consistant
* treatment of the vertex sizes and a lazy strategy for
* cleaning unused parts of the vector, and so as not to exclude
* them from the vertex array interface.
*
* Under our current analysis of matrices, there is no way that
* the product of a matrix and a 1-vector can remain a 1-vector,
* with the exception of the identity transform.
*/
 
/* KW: No longer zero-pad outgoing vectors. Now that external
* vectors can get into the pipeline we cannot ever assume
* that there is more to a vector than indicated by its
* size.
*/
 
/* KW: Now uses clipmask and a flag to allow us to skip both/either
* cliped and/or culled vertices.
*/
 
/* GH: Not any more -- it's easier (and faster) to just process the
* entire vector. Clipping and culling are handled further down
* the pipe, most often during or after the conversion to some
* driver-specific vertex format.
*/
 
static void _XFORMAPI
TAG(transform_points1_general)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m12 = m[12];
const GLfloat m1 = m[1], m13 = m[13];
const GLfloat m2 = m[2], m14 = m[14];
const GLfloat m3 = m[3], m15 = m[15];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox + m12;
to[i][1] = m1 * ox + m13;
to[i][2] = m2 * ox + m14;
to[i][3] = m3 * ox + m15;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points1_identity)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLuint count = from_vec->count;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint i;
if (to_vec == from_vec) return;
STRIDE_LOOP {
to[i][0] = from[0];
}
to_vec->size = 1;
to_vec->flags |= VEC_SIZE_1;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points1_2d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1];
const GLfloat m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox + m12;
to[i][1] = m1 * ox + m13;
}
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points1_2d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox + m12;
to[i][1] = m13;
}
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points1_3d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m2 = m[2];
const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox + m12;
to[i][1] = m1 * ox + m13;
to[i][2] = m2 * ox + m14;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
 
static void _XFORMAPI
TAG(transform_points1_3d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0];
const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox + m12;
to[i][1] = m13;
to[i][2] = m14;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points1_perspective)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0];
to[i][0] = m0 * ox ;
to[i][1] = 0 ;
to[i][2] = m14;
to[i][3] = 0;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
 
 
 
/* 2-vectors, which are a lot more relevant than 1-vectors, are
* present early in the geometry pipeline and throughout the
* texture pipeline.
*/
static void _XFORMAPI
TAG(transform_points2_general)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m4 = m[4], m12 = m[12];
const GLfloat m1 = m[1], m5 = m[5], m13 = m[13];
const GLfloat m2 = m[2], m6 = m[6], m14 = m[14];
const GLfloat m3 = m[3], m7 = m[7], m15 = m[15];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox + m4 * oy + m12;
to[i][1] = m1 * ox + m5 * oy + m13;
to[i][2] = m2 * ox + m6 * oy + m14;
to[i][3] = m3 * ox + m7 * oy + m15;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points2_identity)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
GLuint i;
if (to_vec == from_vec) return;
STRIDE_LOOP {
to[i][0] = from[0];
to[i][1] = from[1];
}
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points2_2d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
const GLfloat m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox + m4 * oy + m12;
to[i][1] = m1 * ox + m5 * oy + m13;
}
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points2_2d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox + m12;
to[i][1] = m5 * oy + m13;
}
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points2_3d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
const GLfloat m6 = m[6], m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox + m4 * oy + m12;
to[i][1] = m1 * ox + m5 * oy + m13;
to[i][2] = m2 * ox + m6 * oy + m14;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
 
/* I would actually say this was a fairly important function, from
* a texture transformation point of view.
*/
static void _XFORMAPI
TAG(transform_points2_3d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5];
const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox + m12;
to[i][1] = m5 * oy + m13;
to[i][2] = m14;
}
if (m14 == 0) {
to_vec->size = 2;
to_vec->flags |= VEC_SIZE_2;
} else {
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
}
to_vec->count = from_vec->count;
}
 
 
static void _XFORMAPI
TAG(transform_points2_perspective)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1];
to[i][0] = m0 * ox ;
to[i][1] = m5 * oy ;
to[i][2] = m14;
to[i][3] = 0;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
 
 
static void _XFORMAPI
TAG(transform_points3_general)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12];
const GLfloat m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13];
const GLfloat m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14];
const GLfloat m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m4 * oy + m8 * oz + m12;
to[i][1] = m1 * ox + m5 * oy + m9 * oz + m13;
to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14;
to[i][3] = m3 * ox + m7 * oy + m11 * oz + m15;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points3_identity)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
GLuint i;
if (to_vec == from_vec) return;
STRIDE_LOOP {
to[i][0] = from[0];
to[i][1] = from[1];
to[i][2] = from[2];
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points3_2d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
const GLfloat m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m4 * oy + m12 ;
to[i][1] = m1 * ox + m5 * oy + m13 ;
to[i][2] = + oz ;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points3_2d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m12 ;
to[i][1] = m5 * oy + m13 ;
to[i][2] = + oz ;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points3_3d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
const GLfloat m6 = m[6], m8 = m[8], m9 = m[9], m10 = m[10];
const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m4 * oy + m8 * oz + m12 ;
to[i][1] = m1 * ox + m5 * oy + m9 * oz + m13 ;
to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14 ;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
/* previously known as ortho...
*/
static void _XFORMAPI
TAG(transform_points3_3d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5];
const GLfloat m10 = m[10], m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m12 ;
to[i][1] = m5 * oy + m13 ;
to[i][2] = m10 * oz + m14 ;
}
to_vec->size = 3;
to_vec->flags |= VEC_SIZE_3;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points3_perspective)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m8 = m[8], m9 = m[9];
const GLfloat m10 = m[10], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2];
to[i][0] = m0 * ox + m8 * oz ;
to[i][1] = m5 * oy + m9 * oz ;
to[i][2] = m10 * oz + m14 ;
to[i][3] = -oz ;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
 
 
static void _XFORMAPI
TAG(transform_points4_general)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12];
const GLfloat m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13];
const GLfloat m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14];
const GLfloat m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m4 * oy + m8 * oz + m12 * ow;
to[i][1] = m1 * ox + m5 * oy + m9 * oz + m13 * ow;
to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14 * ow;
to[i][3] = m3 * ox + m7 * oy + m11 * oz + m15 * ow;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_identity)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
GLuint i;
if (to_vec == from_vec) return;
STRIDE_LOOP {
to[i][0] = from[0];
to[i][1] = from[1];
to[i][2] = from[2];
to[i][3] = from[3];
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_2d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m4 = m[4], m5 = m[5];
const GLfloat m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m4 * oy + m12 * ow;
to[i][1] = m1 * ox + m5 * oy + m13 * ow;
to[i][2] = + oz ;
to[i][3] = ow;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_2d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m12 = m[12], m13 = m[13];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m12 * ow;
to[i][1] = m5 * oy + m13 * ow;
to[i][2] = + oz ;
to[i][3] = ow;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_3d)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m1 = m[1], m2 = m[2], m4 = m[4], m5 = m[5];
const GLfloat m6 = m[6], m8 = m[8], m9 = m[9], m10 = m[10];
const GLfloat m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m4 * oy + m8 * oz + m12 * ow;
to[i][1] = m1 * ox + m5 * oy + m9 * oz + m13 * ow;
to[i][2] = m2 * ox + m6 * oy + m10 * oz + m14 * ow;
to[i][3] = ow;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_3d_no_rot)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5];
const GLfloat m10 = m[10], m12 = m[12], m13 = m[13], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m12 * ow;
to[i][1] = m5 * oy + m13 * ow;
to[i][2] = m10 * oz + m14 * ow;
to[i][3] = ow;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static void _XFORMAPI
TAG(transform_points4_perspective)( GLvector4f *to_vec,
const GLfloat m[16],
const GLvector4f *from_vec )
{
const GLuint stride = from_vec->stride;
GLfloat *from = from_vec->start;
GLfloat (*to)[4] = (GLfloat (*)[4])to_vec->start;
GLuint count = from_vec->count;
const GLfloat m0 = m[0], m5 = m[5], m8 = m[8], m9 = m[9];
const GLfloat m10 = m[10], m14 = m[14];
GLuint i;
STRIDE_LOOP {
const GLfloat ox = from[0], oy = from[1], oz = from[2], ow = from[3];
to[i][0] = m0 * ox + m8 * oz ;
to[i][1] = m5 * oy + m9 * oz ;
to[i][2] = m10 * oz + m14 * ow ;
to[i][3] = -oz ;
}
to_vec->size = 4;
to_vec->flags |= VEC_SIZE_4;
to_vec->count = from_vec->count;
}
 
static transform_func _XFORMAPI TAG(transform_tab_1)[7];
static transform_func _XFORMAPI TAG(transform_tab_2)[7];
static transform_func _XFORMAPI TAG(transform_tab_3)[7];
static transform_func _XFORMAPI TAG(transform_tab_4)[7];
 
/* Similar functions could be called several times, with more highly
* optimized routines overwriting the arrays. This only occurs during
* startup.
*/
static void _XFORMAPI TAG(init_c_transformations)( void )
{
#define TAG_TAB _mesa_transform_tab
#define TAG_TAB_1 TAG(transform_tab_1)
#define TAG_TAB_2 TAG(transform_tab_2)
#define TAG_TAB_3 TAG(transform_tab_3)
#define TAG_TAB_4 TAG(transform_tab_4)
 
TAG_TAB[1] = TAG_TAB_1;
TAG_TAB[2] = TAG_TAB_2;
TAG_TAB[3] = TAG_TAB_3;
TAG_TAB[4] = TAG_TAB_4;
 
/* 1-D points (ie texcoords) */
TAG_TAB_1[MATRIX_GENERAL] = TAG(transform_points1_general);
TAG_TAB_1[MATRIX_IDENTITY] = TAG(transform_points1_identity);
TAG_TAB_1[MATRIX_3D_NO_ROT] = TAG(transform_points1_3d_no_rot);
TAG_TAB_1[MATRIX_PERSPECTIVE] = TAG(transform_points1_perspective);
TAG_TAB_1[MATRIX_2D] = TAG(transform_points1_2d);
TAG_TAB_1[MATRIX_2D_NO_ROT] = TAG(transform_points1_2d_no_rot);
TAG_TAB_1[MATRIX_3D] = TAG(transform_points1_3d);
 
/* 2-D points */
TAG_TAB_2[MATRIX_GENERAL] = TAG(transform_points2_general);
TAG_TAB_2[MATRIX_IDENTITY] = TAG(transform_points2_identity);
TAG_TAB_2[MATRIX_3D_NO_ROT] = TAG(transform_points2_3d_no_rot);
TAG_TAB_2[MATRIX_PERSPECTIVE] = TAG(transform_points2_perspective);
TAG_TAB_2[MATRIX_2D] = TAG(transform_points2_2d);
TAG_TAB_2[MATRIX_2D_NO_ROT] = TAG(transform_points2_2d_no_rot);
TAG_TAB_2[MATRIX_3D] = TAG(transform_points2_3d);
 
/* 3-D points */
TAG_TAB_3[MATRIX_GENERAL] = TAG(transform_points3_general);
TAG_TAB_3[MATRIX_IDENTITY] = TAG(transform_points3_identity);
TAG_TAB_3[MATRIX_3D_NO_ROT] = TAG(transform_points3_3d_no_rot);
TAG_TAB_3[MATRIX_PERSPECTIVE] = TAG(transform_points3_perspective);
TAG_TAB_3[MATRIX_2D] = TAG(transform_points3_2d);
TAG_TAB_3[MATRIX_2D_NO_ROT] = TAG(transform_points3_2d_no_rot);
TAG_TAB_3[MATRIX_3D] = TAG(transform_points3_3d);
 
/* 4-D points */
TAG_TAB_4[MATRIX_GENERAL] = TAG(transform_points4_general);
TAG_TAB_4[MATRIX_IDENTITY] = TAG(transform_points4_identity);
TAG_TAB_4[MATRIX_3D_NO_ROT] = TAG(transform_points4_3d_no_rot);
TAG_TAB_4[MATRIX_PERSPECTIVE] = TAG(transform_points4_perspective);
TAG_TAB_4[MATRIX_2D] = TAG(transform_points4_2d);
TAG_TAB_4[MATRIX_2D_NO_ROT] = TAG(transform_points4_2d_no_rot);
TAG_TAB_4[MATRIX_3D] = TAG(transform_points4_3d);
 
#undef TAG_TAB
#undef TAG_TAB_1
#undef TAG_TAB_2
#undef TAG_TAB_3
#undef TAG_TAB_4
}
/shark/trunk/ports/mesa/src/math/m_norm_debug.c
0,0 → 1,381
/* $Id: m_norm_debug.c,v 1.1 2003-04-24 13:37:47 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
* Version: 4.1
*
* Copyright (C) 1999-2002 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
*/
 
#include "glheader.h"
#include "context.h"
#include "macros.h"
#include "imports.h"
#include "mmath.h"
 
#include "m_matrix.h"
#include "m_xform.h"
 
#include "m_debug.h"
#include "m_debug_util.h"
 
 
#ifdef DEBUG /* This code only used for debugging */
 
 
static int m_norm_identity[16] = {
ONE, NIL, NIL, NIL,
NIL, ONE, NIL, NIL,
NIL, NIL, ONE, NIL,
NIL, NIL, NIL, NIL
};
static int m_norm_general[16] = {
VAR, VAR, VAR, NIL,
VAR, VAR, VAR, NIL,
VAR, VAR, VAR, NIL,
NIL, NIL, NIL, NIL
};
static int m_norm_no_rot[16] = {
VAR, NIL, NIL, NIL,
NIL, VAR, NIL, NIL,
NIL, NIL, VAR, NIL,
NIL, NIL, NIL, NIL
};
static int *norm_templates[8] = {
m_norm_no_rot,
m_norm_no_rot,
m_norm_no_rot,
m_norm_general,
m_norm_general,
m_norm_general,
m_norm_identity,
m_norm_identity
};
static int norm_types[8] = {
NORM_TRANSFORM_NO_ROT,
NORM_TRANSFORM_NO_ROT | NORM_RESCALE,
NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE,
NORM_TRANSFORM,
NORM_TRANSFORM | NORM_RESCALE,
NORM_TRANSFORM | NORM_NORMALIZE,
NORM_RESCALE,
NORM_NORMALIZE
};
static int norm_scale_types[8] = { /* rescale factor */
NIL, /* NIL disables rescaling */
VAR,
NIL,
NIL,
VAR,
NIL,
VAR,
NIL
};
static int norm_normalize_types[8] = { /* normalizing ?? (no = 0) */
0,
0,
1,
0,
0,
1,
0,
1
};
static char *norm_strings[8] = {
"NORM_TRANSFORM_NO_ROT",
"NORM_TRANSFORM_NO_ROT | NORM_RESCALE",
"NORM_TRANSFORM_NO_ROT | NORM_NORMALIZE",
"NORM_TRANSFORM",
"NORM_TRANSFORM | NORM_RESCALE",
"NORM_TRANSFORM | NORM_NORMALIZE",
"NORM_RESCALE",
"NORM_NORMALIZE"
};
 
 
/* =============================================================
* Reference transformations
*/
 
static void ref_norm_transform_rescale( const GLmatrix *mat,
GLfloat scale,
const GLvector4f *in,
const GLfloat *lengths,
GLvector4f *dest )
{
GLuint i;
const GLfloat *s = in->start;
const GLfloat *m = mat->inv;
GLfloat (*out)[4] = (GLfloat (*)[4]) dest->start;
 
(void) lengths;
 
for ( i = 0 ; i < in->count ; i++ ) {
GLfloat t[3];
 
TRANSFORM_NORMAL( t, s, m );
SCALE_SCALAR_3V( out[i], scale, t );
 
s = (GLfloat *)((char *)s + in->stride);
}
}
 
static void ref_norm_transform_normalize( const GLmatrix *mat,
GLfloat scale,
const GLvector4f *in,
const GLfloat *lengths,
GLvector4f *dest )
{
GLuint i;
const GLfloat *s = in->start;
const GLfloat *m = mat->inv;
GLfloat (*out)[4] = (GLfloat (*)[4]) dest->start;
 
for ( i = 0 ; i < in->count ; i++ ) {
GLfloat t[3];
 
TRANSFORM_NORMAL( t, s, m );
 
if ( !lengths ) {
GLfloat len = LEN_SQUARED_3FV( t );
if ( len > 1e-20 ) {
/* Hmmm, don't know how we could test the precalculated
* length case...
*/
scale = 1.0 / sqrt( len );
SCALE_SCALAR_3V( out[i], scale, t );
} else {
out[i][0] = out[i][1] = out[i][2] = 0;
}
} else {
scale = lengths[i];;
SCALE_SCALAR_3V( out[i], scale, t );
}
 
s = (GLfloat *)((char *)s + in->stride);
}
}
 
 
/* =============================================================
* Normal transformation tests
*/
 
static void init_matrix( GLfloat *m )
{
m[0] = 63.0; m[4] = 43.0; m[ 8] = 29.0; m[12] = 43.0;
m[1] = 55.0; m[5] = 17.0; m[ 9] = 31.0; m[13] = 7.0;
m[2] = 44.0; m[6] = 9.0; m[10] = 7.0; m[14] = 3.0;
m[3] = 11.0; m[7] = 23.0; m[11] = 91.0; m[15] = 9.0;
}
 
 
static int test_norm_function( normal_func func, int mtype, long *cycles )
{
GLvector4f source[1], dest[1], dest2[1], ref[1], ref2[1];
GLmatrix mat[1];
GLfloat s[TEST_COUNT][5], d[TEST_COUNT][4], r[TEST_COUNT][4];
GLfloat d2[TEST_COUNT][4], r2[TEST_COUNT][4], length[TEST_COUNT];
GLfloat scale;
GLfloat *m;
int i, j;
#ifdef RUN_DEBUG_BENCHMARK
int cycle_i; /* the counter for the benchmarks we run */
#endif
 
(void) cycles;
 
mat->m = (GLfloat *) ALIGN_MALLOC( 16 * sizeof(GLfloat), 16 );
mat->inv = m = mat->m;
 
init_matrix( m );
 
scale = 1.0F + rnd () * norm_scale_types[mtype];
 
for ( i = 0 ; i < 4 ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
switch ( norm_templates[mtype][i * 4 + j] ) {
case NIL:
m[j * 4 + i] = 0.0;
break;
case ONE:
m[j * 4 + i] = 1.0;
break;
case NEG:
m[j * 4 + i] = -1.0;
break;
case VAR:
break;
default:
abort();
}
}
}
 
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
ASSIGN_3V( d[i], 0.0, 0.0, 0.0 );
ASSIGN_3V( s[i], 0.0, 0.0, 0.0 );
ASSIGN_3V( d2[i], 0.0, 0.0, 0.0 );
for ( j = 0 ; j < 3 ; j++ )
s[i][j] = rnd();
length[i] = 1 / sqrt( LEN_SQUARED_3FV( s[i] ) );
}
 
source->data = (GLfloat(*)[4]) s;
source->start = (GLfloat *) s;
source->count = TEST_COUNT;
source->stride = sizeof(s[0]);
source->flags = 0;
 
dest->data = d;
dest->start = (GLfloat *) d;
dest->count = TEST_COUNT;
dest->stride = sizeof(float[4]);
dest->flags = 0;
 
dest2->data = d2;
dest2->start = (GLfloat *) d2;
dest2->count = TEST_COUNT;
dest2->stride = sizeof(float[4]);
dest2->flags = 0;
 
ref->data = r;
ref->start = (GLfloat *) r;
ref->count = TEST_COUNT;
ref->stride = sizeof(float[4]);
ref->flags = 0;
 
ref2->data = r2;
ref2->start = (GLfloat *) r2;
ref2->count = TEST_COUNT;
ref2->stride = sizeof(float[4]);
ref2->flags = 0;
 
if ( norm_normalize_types[mtype] == 0 ) {
ref_norm_transform_rescale( mat, scale, source, NULL, ref );
} else {
ref_norm_transform_normalize( mat, scale, source, NULL, ref );
ref_norm_transform_normalize( mat, scale, source, length, ref2 );
}
 
if ( mesa_profile ) {
BEGIN_RACE( *cycles );
func( mat, scale, source, NULL, dest );
END_RACE( *cycles );
func( mat, scale, source, length, dest2 );
} else {
func( mat, scale, source, NULL, dest );
func( mat, scale, source, length, dest2 );
}
 
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
for ( j = 0 ; j < 3 ; j++ ) {
if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
_mesa_printf(NULL, "-----------------------------\n" );
_mesa_printf(NULL, "(i = %i, j = %i)\n", i, j );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d[i][0], r[i][0], r[i][0]/d[i][0],
MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d[i][1], r[i][1], r[i][1]/d[i][1],
MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d[i][2], r[i][2], r[i][2]/d[i][2],
MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
return 0;
}
 
if ( norm_normalize_types[mtype] != 0 ) {
if ( significand_match( d2[i][j], r2[i][j] ) < REQUIRED_PRECISION ) {
_mesa_printf(NULL, "------------------- precalculated length case ------\n" );
_mesa_printf(NULL, "(i = %i, j = %i)\n", i, j );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d2[i][0], r2[i][0], r2[i][0]/d2[i][0],
MAX_PRECISION - significand_match( d2[i][0], r2[i][0] ) );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d2[i][1], r2[i][1], r2[i][1]/d2[i][1],
MAX_PRECISION - significand_match( d2[i][1], r2[i][1] ) );
_mesa_printf(NULL, "%f \t %f \t [ratio = %e - %i bit missed]\n",
d2[i][2], r2[i][2], r2[i][2]/d2[i][2],
MAX_PRECISION - significand_match( d2[i][2], r2[i][2] ) );
return 0;
}
}
}
}
 
ALIGN_FREE( mat->m );
return 1;
}
 
void _math_test_all_normal_transform_functions( char *description )
{
int mtype;
long benchmark_tab[0xf];
static int first_time = 1;
 
if ( first_time ) {
first_time = 0;
mesa_profile = getenv( "MESA_PROFILE" );
}
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
if ( !counter_overhead ) {
INIT_COUNTER();
_mesa_printf(NULL, "counter overhead: %ld cycles\n\n", counter_overhead );
}
_mesa_printf(NULL, "normal transform results after hooking in %s functions:\n",
description );
_mesa_printf(NULL, "\n-------------------------------------------------------\n" );
}
#endif
 
for ( mtype = 0 ; mtype < 8 ; mtype++ ) {
normal_func func = _mesa_normal_tab[norm_types[mtype]];
long *cycles = &benchmark_tab[mtype];
 
if ( test_norm_function( func, mtype, cycles ) == 0 ) {
char buf[100];
_mesa_sprintf(NULL, buf, "_mesa_normal_tab[0][%s] failed test (%s)",
norm_strings[mtype], description );
_mesa_problem( NULL, buf );
}
 
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
_mesa_printf(NULL, " %li\t", benchmark_tab[mtype] );
_mesa_printf(NULL, " | [%s]\n", norm_strings[mtype] );
}
#endif
}
#ifdef RUN_DEBUG_BENCHMARK
if ( mesa_profile ) {
_mesa_printf(NULL, "\n" );
fflush( stdout );
}
#endif
}
 
 
#endif /* DEBUG */
/shark/trunk/ports/mesa/src/makefile
33,12 → 33,12
./tnl/t_vb_fog.o ./tnl/t_vb_light.o\
./tnl/t_vb_normals.o ./tnl/t_vb_points.o ./tnl/t_vb_program.o ./tnl/t_vb_render.o\
./tnl/t_vb_texgen.o ./tnl/t_vb_texmat.o ./tnl/t_vb_vertex.o\
./math/m_debug_clip.o ./math/m_debug_norm.o\
./math/m_debug_xform.o ./math/m_eval.o ./math/m_matrix.o ./math/m_translate.o\
./math/m_clip_debug.o ./math/m_norm_debug.o\
./math/m_xform_debug.o ./math/m_eval.o ./math/m_matrix.o ./math/m_translate.o\
./math/m_vector.o ./math/m_xform.o\
./x86/common_x86_asm.o ./x86/common_x86.o ./x86/glapi_x86.o\
./x86/mmx_blend.o ./x86/x86.o ./x86/x86_cliptest.o ./x86/x86_xform2.o\
./x86/x86_xform3.o ./x86/x86_xform4.o ./osmesa/osmesa.o imports.o \
./x86/asm_common_x86.o ./x86/common_x86.o ./x86/glapi_x86.o\
./x86/mmx_blend.o ./x86/x86.o ./x86/x86_cliptest.o ./x86/xform2_x86.o\
./x86/xform3_x86.o ./x86/xform4_x86.o ./osmesa/osmesa.o imports.o \
OBJS = $(OSMESA)
 
/shark/trunk/ports/mesa/src/x86/common_x86_macros.h
File deleted
/shark/trunk/ports/mesa/src/x86/x86_xform3.s
File deleted
/shark/trunk/ports/mesa/src/x86/x86_xform2.s
File deleted
/shark/trunk/ports/mesa/src/x86/x86_xform4.s
File deleted
/shark/trunk/ports/mesa/src/x86/common_x86_features.h
File deleted
/shark/trunk/ports/mesa/src/x86/common_x86_asm.s
File deleted
/shark/trunk/ports/mesa/src/x86/common_x86_asm.h
1,4 → 1,4
/* $Id: common_x86_asm.h,v 1.1 2003-03-13 12:11:48 giacomo Exp $ */
/* $Id: common_x86_asm.h,v 1.2 2003-04-24 13:36:03 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
41,7 → 41,7
 
/* Do not reference mtypes.h from this file.
*/
#include "common_x86_features.h"
#include "features_common_x86.h"
 
#ifdef HAVE_CONFIG_H
#include "conf.h"
/shark/trunk/ports/mesa/src/x86/macros_common_x86.h
0,0 → 1,107
/* $Id: macros_common_x86.h,v 1.1 2003-04-24 13:36:03 giacomo 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
/shark/trunk/ports/mesa/src/x86/asm_common_x86.s
0,0 → 1,239
/* $Id: asm_common_x86.s,v 1.1 2003-04-24 13:36:02 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
* Version: 4.0.3
*
* Copyright (C) 1999-2002 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 "features_common_x86.h"
 
 
/* Intel vendor string
*/
#define GENU 0x756e6547 /* "Genu" */
#define INEI 0x49656e69 /* "ineI" */
#define NTEL 0x6c65746e /* "ntel" */
 
/* AMD vendor string
*/
#define AUTH 0x68747541 /* "Auth" */
#define ENTI 0x69746e65 /* "enti" */
#define CAMD 0x444d4163 /* "cAMD" */
 
 
SEG_DATA
 
/* We might want to print out some useful messages.
*/
GLNAME( found_intel ): STRING( "Genuine Intel processor found\n\0" )
GLNAME( found_amd ): STRING( "Authentic AMD processor found\n\0" )
 
 
SEG_TEXT
 
ALIGNTEXT4
GLOBL GLNAME( _mesa_identify_x86_cpu_features )
GLNAME( _mesa_identify_x86_cpu_features ):
 
PUSH_L ( EBX )
PUSH_L ( ESI )
 
/* 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 )
JZ ( LLBL (cpuid_done) )
 
/* Get the CPU vendor info.
*/
XOR_L ( EAX, EAX )
CPUID
 
/* Test for Intel processors. We must look for the
* "GenuineIntel" string in EBX, ECX and EDX.
*/
CMP_L ( CONST(GENU), EBX )
JNE ( LLBL(cpuid_amd) )
CMP_L ( CONST(INEI), EDX )
JNE ( LLBL(cpuid_amd) )
CMP_L ( CONST(NTEL), ECX )
JNE ( LLBL(cpuid_amd) )
 
/* We have an Intel processor, so we can get the feature
* information with an CPUID input value of 1.
*/
MOV_L ( CONST(0x1), EAX )
CPUID
MOV_L ( EDX, EAX )
 
/* Mask out highest bit, which is used by AMD for 3dnow
* Newer Intel have this bit set, but do not support 3dnow
*/
AND_L ( CONST(0X7FFFFFFF), EAX)
JMP ( LLBL(cpuid_done) )
 
LLBL(cpuid_amd):
 
/* Test for AMD processors. We must look for the
* "AuthenticAMD" string in EBX, ECX and EDX.
*/
CMP_L ( CONST(AUTH), EBX )
JNE ( LLBL(cpuid_other) )
CMP_L ( CONST(ENTI), EDX )
JNE ( LLBL(cpuid_other) )
CMP_L ( CONST(CAMD), ECX )
JNE ( LLBL(cpuid_other) )
 
/* We have an AMD processor, so we can get the feature
* information after we verify that the extended functions are
* supported.
*/
/* The features we need are almost all in the extended set. The
* exception is SSE enable, which is in the standard set (0x1).
*/
MOV_L ( CONST(0x1), EAX )
CPUID
TEST_L ( EAX, EAX )
JZ ( LLBL (cpuid_failed) )
MOV_L ( EDX, ESI )
 
MOV_L ( CONST(0x80000000), EAX )
CPUID
TEST_L ( EAX, EAX )
JZ ( LLBL (cpuid_failed) )
 
MOV_L ( CONST(0x80000001), EAX )
CPUID
MOV_L ( EDX, EAX )
AND_L ( CONST(0x02000000), ESI ) /* OR in the SSE bit */
OR_L ( ESI, EAX )
JMP ( LLBL (cpuid_done) )
 
LLBL(cpuid_other):
 
/* Test for other processors here when required.
*/
 
LLBL(cpuid_failed):
 
/* If we can't determine the feature information, we must
* return zero to indicate that no platform-specific
* optimizations can be used.
*/
MOV_L ( CONST(0), EAX )
 
LLBL (cpuid_done):
 
POP_L ( ESI )
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 )
 
ADD_L ( CONST( 32 ), ESP )
 
DIVPS ( XMM0, XMM1 )
 
/* Restore the original MXCSR register value.
*/
LDMXCSR ( REGOFF( -4, EBP ) )
 
LEAVE
RET
 
#endif
/shark/trunk/ports/mesa/src/x86/features_common_x86.h
0,0 → 1,80
/* $Id: features_common_x86.h,v 1.1 2003-04-24 13:36:03 giacomo 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.
*/
 
/*
* 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__
 
/* Capabilities of CPUs
*/
#define X86_FEATURE_FPU 0x00000001
#define X86_FEATURE_VME 0x00000002
#define X86_FEATURE_DE 0x00000004
#define X86_FEATURE_PSE 0x00000008
#define X86_FEATURE_TSC 0x00000010
#define X86_FEATURE_MSR 0x00000020
#define X86_FEATURE_PAE 0x00000040
#define X86_FEATURE_MCE 0x00000080
#define X86_FEATURE_CX8 0x00000100
#define X86_FEATURE_APIC 0x00000200
#define X86_FEATURE_10 0x00000400
#define X86_FEATURE_SEP 0x00000800
#define X86_FEATURE_MTRR 0x00001000
#define X86_FEATURE_PGE 0x00002000
#define X86_FEATURE_MCA 0x00004000
#define X86_FEATURE_CMOV 0x00008000
#define X86_FEATURE_PAT 0x00010000
#define X86_FEATURE_PSE36 0x00020000
#define X86_FEATURE_18 0x00040000
#define X86_FEATURE_19 0x00080000
#define X86_FEATURE_20 0x00100000
#define X86_FEATURE_21 0x00200000
#define X86_FEATURE_MMXEXT 0x00400000
#define X86_FEATURE_MMX 0x00800000
#define X86_FEATURE_FXSR 0x01000000
#define X86_FEATURE_XMM 0x02000000
#define X86_FEATURE_XMM2 0x04000000
#define X86_FEATURE_27 0x08000000
#define X86_FEATURE_28 0x10000000
#define X86_FEATURE_29 0x20000000
#define X86_FEATURE_3DNOWEXT 0x40000000
#define X86_FEATURE_3DNOW 0x80000000
 
#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
/shark/trunk/ports/mesa/src/x86/xform2_x86.s
0,0 → 1,563
/* $Id: xform2_x86.s,v 1.1 2003-04-24 13:36:03 giacomo 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
/shark/trunk/ports/mesa/src/x86/xform3_x86.s
0,0 → 1,633
/* $Id: xform3_x86.s,v 1.1 2003-04-24 13:36:03 giacomo 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
/shark/trunk/ports/mesa/src/x86/xform4_x86.s
0,0 → 1,666
/* $Id: xform4_x86.s,v 1.1 2003-04-24 13:36:03 giacomo 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
/shark/trunk/ports/mesa/src/x86/x86.c
1,4 → 1,4
/* $Id: x86.c,v 1.1 2003-03-13 12:11:49 giacomo Exp $ */
/* $Id: x86.c,v 1.2 2003-04-24 13:36:03 giacomo Exp $ */
 
/*
* Mesa 3-D graphics library
34,7 → 34,7
#include "tnl/t_context.h"
 
#include "x86.h"
#include "common_x86_macros.h"
#include "macros_common_x86.h"
 
#ifdef DEBUG
#include "math/m_debug.h"