math: remove unused debug classes
Signed-off-by: António Monteiro <antonio.fmr.monteiro@gmail.com> Reviewed-by: Marek Olšák <marek.olsak@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19428>
This commit is contained in:
committed by
Marge Bot
parent
94e10ced83
commit
2d4d4df57a
@@ -135,8 +135,6 @@
|
||||
#include <mesa/main/varray.h>
|
||||
#include <mesa/main/version.h>
|
||||
#include <mesa/main/viewport.h>
|
||||
#include <mesa/math/m_debug.h>
|
||||
#include <mesa/math/m_debug_util.h>
|
||||
#include <mesa/math/m_eval.h>
|
||||
#include <mesa/math/m_matrix.h>
|
||||
#include <mesa/math/m_vector.h>
|
||||
|
||||
@@ -1,36 +0,0 @@
|
||||
|
||||
/*
|
||||
* Mesa 3-D graphics library
|
||||
*
|
||||
* 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
|
||||
* THE AUTHORS OR COPYRIGHT HOLDERS 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 __M_DEBUG_H__
|
||||
#define __M_DEBUG_H__
|
||||
|
||||
extern void _math_test_all_transform_functions( char *description );
|
||||
extern void _math_test_all_normal_transform_functions( char *description );
|
||||
extern void _math_test_all_cliptest_functions( char *description );
|
||||
|
||||
#endif
|
||||
@@ -1,409 +0,0 @@
|
||||
/*
|
||||
* Mesa 3-D graphics library
|
||||
*
|
||||
* Copyright (C) 1999-2005 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
|
||||
* THE AUTHORS OR COPYRIGHT HOLDERS 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 "util/glheader.h"
|
||||
#include "main/context.h"
|
||||
#include "main/macros.h"
|
||||
|
||||
|
||||
#include "m_matrix.h"
|
||||
#include "m_xform.h"
|
||||
|
||||
#include "m_debug.h"
|
||||
#include "m_debug_util.h"
|
||||
|
||||
#ifdef __UNIXOS2__
|
||||
/* The linker doesn't like empty files */
|
||||
static char dummy;
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_MATH /* 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,
|
||||
GLboolean viewport_z_clip )
|
||||
{
|
||||
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 (viewport_z_clip) {
|
||||
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,
|
||||
GLboolean viewport_z_clip )
|
||||
{
|
||||
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 (viewport_z_clip) {
|
||||
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,
|
||||
GLboolean viewport_z_clip )
|
||||
{
|
||||
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;
|
||||
|
||||
(void) viewport_z_clip;
|
||||
|
||||
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
|
||||
*/
|
||||
|
||||
ALIGN16 static GLfloat s[TEST_COUNT][4];
|
||||
ALIGN16 static GLfloat d[TEST_COUNT][4];
|
||||
ALIGN16 static GLfloat r[TEST_COUNT][4];
|
||||
|
||||
|
||||
/**
|
||||
* Check if X, Y or Z component of the coordinate is close to W, in terms
|
||||
* of the clip test.
|
||||
*/
|
||||
static GLboolean
|
||||
xyz_close_to_w(const GLfloat c[4])
|
||||
{
|
||||
float k = 0.0001;
|
||||
return (fabs(c[0] - c[3]) < k ||
|
||||
fabs(c[1] - c[3]) < k ||
|
||||
fabs(c[2] - c[3]) < k ||
|
||||
fabs(-c[0] - c[3]) < k ||
|
||||
fabs(-c[1] - c[3]) < k ||
|
||||
fabs(-c[2] - c[3]) < k);
|
||||
}
|
||||
|
||||
|
||||
|
||||
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
|
||||
GLboolean viewport_z_clip = GL_TRUE;
|
||||
|
||||
(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_FRUSTUM_BITS;
|
||||
|
||||
ref_cliptest[psize]( source, ref, rm, &rco, &rca, viewport_z_clip );
|
||||
|
||||
if ( mesa_profile ) {
|
||||
BEGIN_RACE( *cycles );
|
||||
func( source, dest, dm, &dco, &dca, viewport_z_clip );
|
||||
END_RACE( *cycles );
|
||||
}
|
||||
else {
|
||||
func( source, dest, dm, &dco, &dca, viewport_z_clip );
|
||||
}
|
||||
|
||||
if ( dco != rco ) {
|
||||
printf( "\n-----------------------------\n" );
|
||||
printf( "dco = 0x%02x rco = 0x%02x\n", dco, rco );
|
||||
return 0;
|
||||
}
|
||||
if ( dca != rca ) {
|
||||
printf( "\n-----------------------------\n" );
|
||||
printf( "dca = 0x%02x rca = 0x%02x\n", dca, rca );
|
||||
return 0;
|
||||
}
|
||||
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
|
||||
if ( dm[i] != rm[i] ) {
|
||||
GLfloat *c = source->start;
|
||||
STRIDE_F(c, source->stride * i);
|
||||
if (psize == 4 && xyz_close_to_w(c)) {
|
||||
/* The coordinate is very close to the clip plane. The clipmask
|
||||
* may vary depending on code path, but that's OK.
|
||||
*/
|
||||
continue;
|
||||
}
|
||||
printf( "\n-----------------------------\n" );
|
||||
printf( "mask[%d] = 0x%02x ref mask[%d] = 0x%02x\n", i, dm[i], i,rm[i] );
|
||||
printf(" coord = %f, %f, %f, %f\n",
|
||||
c[0], c[1], c[2], c[3]);
|
||||
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 ) {
|
||||
printf( "\n-----------------------------\n" );
|
||||
printf( "(i = %i, j = %i) dm = 0x%02x rm = 0x%02x\n",
|
||||
i, j, dm[i], rm[i] );
|
||||
printf( "%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] ) );
|
||||
printf( "%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] ) );
|
||||
printf( "%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] ) );
|
||||
printf( "%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 = getenv( "MESA_PROFILE" );
|
||||
}
|
||||
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile ) {
|
||||
if ( !counter_overhead ) {
|
||||
INIT_COUNTER();
|
||||
printf( "counter overhead: %ld cycles\n\n", counter_overhead );
|
||||
}
|
||||
printf( "cliptest results after hooking in %s functions:\n", description );
|
||||
}
|
||||
#endif
|
||||
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile ) {
|
||||
printf( "\n\t" );
|
||||
for ( psize = 2 ; psize <= 4 ; psize++ ) {
|
||||
printf( " p%d\t", psize );
|
||||
}
|
||||
printf( "\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];
|
||||
sprintf( buf, "%s[%d] failed test (%s)",
|
||||
cnames[np], psize, description );
|
||||
_mesa_problem( NULL, "%s", buf );
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf( " %li\t", benchmark_tab[np][psize-1] );
|
||||
#endif
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf( " | [%s]\n\t", cstrings[np] );
|
||||
#endif
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf( "\n" );
|
||||
#endif
|
||||
}
|
||||
|
||||
|
||||
#endif /* DEBUG_MATH */
|
||||
@@ -1,383 +0,0 @@
|
||||
|
||||
/*
|
||||
* Mesa 3-D graphics library
|
||||
*
|
||||
* Copyright (C) 1999-2003 Brian Paul All Rights Reserved.
|
||||
*
|
||||
* Permission is hereby granted, free of charge, to any person obtaining a
|
||||
* copy of this software and associated documentation files (the "Software"),
|
||||
* to deal in the Software without restriction, including without limitation
|
||||
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
||||
* and/or sell copies of the Software, and to permit persons to whom the
|
||||
* Software is furnished to do so, subject to the following conditions:
|
||||
*
|
||||
* The above copyright notice and this permission notice shall be included
|
||||
* in all copies or substantial portions of the Software.
|
||||
*
|
||||
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
||||
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
||||
* THE AUTHORS OR COPYRIGHT HOLDERS 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 "util/glheader.h"
|
||||
#include "main/context.h"
|
||||
#include "main/macros.h"
|
||||
|
||||
#include "m_matrix.h"
|
||||
#include "m_xform.h"
|
||||
|
||||
#include "m_debug.h"
|
||||
#include "m_debug_util.h"
|
||||
|
||||
#include <math.h>
|
||||
|
||||
#ifdef __UNIXOS2__
|
||||
/* The linker doesn't like empty files */
|
||||
static char dummy;
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_MATH /* 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 = 1.0f / sqrtf(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 = 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:
|
||||
exit(1);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
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] = 1.0f / sqrtf( 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 ) {
|
||||
printf( "-----------------------------\n" );
|
||||
printf( "(i = %i, j = %i)\n", i, j );
|
||||
printf( "%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] ) );
|
||||
printf( "%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] ) );
|
||||
printf( "%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 ) {
|
||||
printf( "------------------- precalculated length case ------\n" );
|
||||
printf( "(i = %i, j = %i)\n", i, j );
|
||||
printf( "%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] ) );
|
||||
printf( "%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] ) );
|
||||
printf( "%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();
|
||||
printf( "counter overhead: %ld cycles\n\n", counter_overhead );
|
||||
}
|
||||
printf( "normal transform results after hooking in %s functions:\n",
|
||||
description );
|
||||
printf( "\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];
|
||||
sprintf( buf, "_mesa_normal_tab[0][%s] failed test (%s)",
|
||||
norm_strings[mtype], description );
|
||||
_mesa_problem( NULL, "%s", buf );
|
||||
}
|
||||
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile ) {
|
||||
printf( " %li\t", benchmark_tab[mtype] );
|
||||
printf( " | [%s]\n", norm_strings[mtype] );
|
||||
}
|
||||
#endif
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile ) {
|
||||
printf( "\n" );
|
||||
}
|
||||
#endif
|
||||
}
|
||||
|
||||
|
||||
#endif /* DEBUG_MATH */
|
||||
@@ -1,307 +0,0 @@
|
||||
/*
|
||||
* Mesa 3-D graphics library
|
||||
*
|
||||
* Copyright (C) 1999-2004 Brian Paul All Rights Reserved.
|
||||
*
|
||||
* Permission is hereby granted, free of charge, to any person obtaining a
|
||||
* copy of this software and associated documentation files (the "Software"),
|
||||
* to deal in the Software without restriction, including without limitation
|
||||
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
||||
* and/or sell copies of the Software, and to permit persons to whom the
|
||||
* Software is furnished to do so, subject to the following conditions:
|
||||
*
|
||||
* The above copyright notice and this permission notice shall be included
|
||||
* in all copies or substantial portions of the Software.
|
||||
*
|
||||
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
||||
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
||||
* THE AUTHORS OR COPYRIGHT HOLDERS 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 __M_DEBUG_UTIL_H__
|
||||
#define __M_DEBUG_UTIL_H__
|
||||
|
||||
|
||||
#ifdef DEBUG_MATH /* This code only used for debugging */
|
||||
|
||||
|
||||
#include <math.h>
|
||||
|
||||
|
||||
/* Comment this out to deactivate the cycle counter.
|
||||
* NOTE: it works only on CPUs which know the 'rdtsc' command (586 or higher)
|
||||
* (hope, you don't try to debug Mesa on a 386 ;)
|
||||
*/
|
||||
#if defined(__GNUC__) && \
|
||||
((defined(__i386__) && defined(USE_X86_ASM)) || \
|
||||
(defined(__sparc__) && defined(USE_SPARC_ASM)))
|
||||
#define RUN_DEBUG_BENCHMARK
|
||||
#endif
|
||||
|
||||
#define TEST_COUNT 128 /* size of the tested vector array */
|
||||
|
||||
#define REQUIRED_PRECISION 10 /* allow 4 bits to miss */
|
||||
#define MAX_PRECISION 24 /* max. precision possible */
|
||||
|
||||
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
/* Overhead of profiling counter in cycles. Automatically adjusted to
|
||||
* your machine at run time - counter initialization should give very
|
||||
* consistent results.
|
||||
*/
|
||||
extern long counter_overhead;
|
||||
|
||||
/* 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.
|
||||
*/
|
||||
extern char *mesa_profile;
|
||||
|
||||
/* Modify the number of tests if you like.
|
||||
* We take the minimum of all results, because every error should be
|
||||
* positive (time used by other processes, task switches etc).
|
||||
* It is assumed that all calculations are done in the cache.
|
||||
*/
|
||||
|
||||
#if defined(__i386__)
|
||||
|
||||
#if 1 /* PPro, PII, PIII version */
|
||||
|
||||
/* Profiling on the P6 architecture requires a little more work, due to
|
||||
* the internal out-of-order execution. We must perform a serializing
|
||||
* 'cpuid' instruction before and after the 'rdtsc' instructions to make
|
||||
* sure no other uops are executed when we sample the timestamp counter.
|
||||
*/
|
||||
#define INIT_COUNTER() \
|
||||
do { \
|
||||
int cycle_i; \
|
||||
counter_overhead = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 8 ; cycle_i++ ) { \
|
||||
long cycle_tmp1 = 0, cycle_tmp2 = 0; \
|
||||
__asm__ __volatile__ ( "push %%ebx \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"rdtsc \n" \
|
||||
"mov %%eax, %0 \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"pop %%ebx \n" \
|
||||
"push %%ebx \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"rdtsc \n" \
|
||||
"mov %%eax, %1 \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"pop %%ebx \n" \
|
||||
: "=m" (cycle_tmp1), "=m" (cycle_tmp2) \
|
||||
: : "eax", "ecx", "edx" ); \
|
||||
if ( counter_overhead > (cycle_tmp2 - cycle_tmp1) ) { \
|
||||
counter_overhead = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
#define BEGIN_RACE(x) \
|
||||
x = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 10 ; cycle_i++ ) { \
|
||||
long cycle_tmp1 = 0, cycle_tmp2 = 0; \
|
||||
__asm__ __volatile__ ( "push %%ebx \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"rdtsc \n" \
|
||||
"mov %%eax, %0 \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"pop %%ebx \n" \
|
||||
: "=m" (cycle_tmp1) \
|
||||
: : "eax", "ecx", "edx" );
|
||||
|
||||
#define END_RACE(x) \
|
||||
__asm__ __volatile__ ( "push %%ebx \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"rdtsc \n" \
|
||||
"mov %%eax, %0 \n" \
|
||||
"xor %%eax, %%eax \n" \
|
||||
"cpuid \n" \
|
||||
"pop %%ebx \n" \
|
||||
: "=m" (cycle_tmp2) \
|
||||
: : "eax", "ecx", "edx" ); \
|
||||
if ( x > (cycle_tmp2 - cycle_tmp1) ) { \
|
||||
x = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
} \
|
||||
x -= counter_overhead;
|
||||
|
||||
#else /* PPlain, PMMX version */
|
||||
|
||||
/* To ensure accurate results, we stall the pipelines with the
|
||||
* non-pairable 'cdq' instruction. This ensures all the code being
|
||||
* profiled is complete when the 'rdtsc' instruction executes.
|
||||
*/
|
||||
#define INIT_COUNTER(x) \
|
||||
do { \
|
||||
int cycle_i; \
|
||||
x = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 32 ; cycle_i++ ) { \
|
||||
long cycle_tmp1, cycle_tmp2, dummy; \
|
||||
__asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp1) ); \
|
||||
__asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp2) ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "rdtsc" : "=a" (cycle_tmp1), "=d" (dummy) ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "rdtsc" : "=a" (cycle_tmp2), "=d" (dummy) ); \
|
||||
if ( x > (cycle_tmp2 - cycle_tmp1) ) \
|
||||
x = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
#define BEGIN_RACE(x) \
|
||||
x = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 16 ; cycle_i++ ) { \
|
||||
long cycle_tmp1, cycle_tmp2, dummy; \
|
||||
__asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp1) ); \
|
||||
__asm__ ( "mov %%eax, %0" : "=a" (cycle_tmp2) ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "rdtsc" : "=a" (cycle_tmp1), "=d" (dummy) );
|
||||
|
||||
|
||||
#define END_RACE(x) \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "cdq" ); \
|
||||
__asm__ ( "rdtsc" : "=a" (cycle_tmp2), "=d" (dummy) ); \
|
||||
if ( x > (cycle_tmp2 - cycle_tmp1) ) \
|
||||
x = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
x -= counter_overhead;
|
||||
|
||||
#endif
|
||||
|
||||
#elif defined(__x86_64__)
|
||||
|
||||
#define rdtscll(val) do { \
|
||||
unsigned int a,d; \
|
||||
__asm__ volatile("rdtsc" : "=a" (a), "=d" (d)); \
|
||||
(val) = ((unsigned long)a) | (((unsigned long)d)<<32); \
|
||||
} while(0)
|
||||
|
||||
/* Copied from i386 PIII version */
|
||||
#define INIT_COUNTER() \
|
||||
do { \
|
||||
int cycle_i; \
|
||||
counter_overhead = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 16 ; cycle_i++ ) { \
|
||||
unsigned long cycle_tmp1, cycle_tmp2; \
|
||||
rdtscll(cycle_tmp1); \
|
||||
rdtscll(cycle_tmp2); \
|
||||
if ( counter_overhead > (cycle_tmp2 - cycle_tmp1) ) { \
|
||||
counter_overhead = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
|
||||
#define BEGIN_RACE(x) \
|
||||
x = LONG_MAX; \
|
||||
for ( cycle_i = 0 ; cycle_i < 10 ; cycle_i++ ) { \
|
||||
unsigned long cycle_tmp1, cycle_tmp2; \
|
||||
rdtscll(cycle_tmp1);
|
||||
|
||||
#define END_RACE(x) \
|
||||
rdtscll(cycle_tmp2); \
|
||||
if ( x > (cycle_tmp2 - cycle_tmp1) ) { \
|
||||
x = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
} \
|
||||
x -= counter_overhead;
|
||||
|
||||
#elif defined(__sparc__)
|
||||
|
||||
#define INIT_COUNTER() \
|
||||
do { counter_overhead = 5; } while(0)
|
||||
|
||||
#define BEGIN_RACE(x) \
|
||||
x = LONG_MAX; \
|
||||
for (cycle_i = 0; cycle_i <10; cycle_i++) { \
|
||||
register long cycle_tmp1 __asm__("l0"); \
|
||||
register long cycle_tmp2 __asm__("l1"); \
|
||||
/* rd %tick, %l0 */ \
|
||||
__asm__ __volatile__ (".word 0xa1410000" : "=r" (cycle_tmp1)); /* save timestamp */
|
||||
|
||||
#define END_RACE(x) \
|
||||
/* rd %tick, %l1 */ \
|
||||
__asm__ __volatile__ (".word 0xa3410000" : "=r" (cycle_tmp2)); \
|
||||
if (x > (cycle_tmp2-cycle_tmp1)) x = cycle_tmp2 - cycle_tmp1; \
|
||||
} \
|
||||
x -= counter_overhead;
|
||||
|
||||
#else
|
||||
#error Your processor is not supported for RUN_XFORM_BENCHMARK
|
||||
#endif
|
||||
|
||||
#else
|
||||
|
||||
#define BEGIN_RACE(x)
|
||||
#define END_RACE(x)
|
||||
|
||||
#endif
|
||||
|
||||
|
||||
/* =============================================================
|
||||
* Helper functions
|
||||
*/
|
||||
|
||||
static GLfloat rnd( void )
|
||||
{
|
||||
GLfloat f = (GLfloat)rand() / (GLfloat)RAND_MAX;
|
||||
GLfloat gran = (GLfloat)(1 << 13);
|
||||
|
||||
f = (GLfloat)(GLint)(f * gran) / gran;
|
||||
|
||||
return f * 2.0 - 1.0;
|
||||
}
|
||||
|
||||
static int significand_match( GLfloat a, GLfloat b )
|
||||
{
|
||||
GLfloat d = a - b;
|
||||
int a_ex, b_ex, d_ex;
|
||||
|
||||
if ( d == 0.0F ) {
|
||||
return MAX_PRECISION; /* Exact match */
|
||||
}
|
||||
|
||||
if ( a == 0.0F || b == 0.0F ) {
|
||||
/* It would probably be better to check if the
|
||||
* non-zero number is denormalized and return
|
||||
* the index of the highest set bit here.
|
||||
*/
|
||||
return 0;
|
||||
}
|
||||
|
||||
frexpf( a, &a_ex );
|
||||
frexpf( b, &b_ex );
|
||||
frexpf( d, &d_ex );
|
||||
|
||||
if ( a_ex < b_ex ) {
|
||||
return a_ex - d_ex;
|
||||
} else {
|
||||
return b_ex - d_ex;
|
||||
}
|
||||
}
|
||||
|
||||
enum { NIL = 0, ONE = 1, NEG = -1, VAR = 2 };
|
||||
|
||||
#endif /* DEBUG_MATH */
|
||||
|
||||
#endif /* __M_DEBUG_UTIL_H__ */
|
||||
@@ -1,338 +0,0 @@
|
||||
/*
|
||||
* Mesa 3-D graphics library
|
||||
*
|
||||
* Copyright (C) 1999-2004 Brian Paul All Rights Reserved.
|
||||
*
|
||||
* Permission is hereby granted, free of charge, to any person obtaining a
|
||||
* copy of this software and associated documentation files (the "Software"),
|
||||
* to deal in the Software without restriction, including without limitation
|
||||
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
||||
* and/or sell copies of the Software, and to permit persons to whom the
|
||||
* Software is furnished to do so, subject to the following conditions:
|
||||
*
|
||||
* The above copyright notice and this permission notice shall be included
|
||||
* in all copies or substantial portions of the Software.
|
||||
*
|
||||
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
||||
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
||||
* THE AUTHORS OR COPYRIGHT HOLDERS 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 "util/glheader.h"
|
||||
#include "main/context.h"
|
||||
#include "main/macros.h"
|
||||
|
||||
#include "m_matrix.h"
|
||||
#include "m_xform.h"
|
||||
|
||||
#include "m_debug.h"
|
||||
#include "m_debug_util.h"
|
||||
|
||||
#ifdef __UNIXOS2__
|
||||
/* The linker doesn't like empty files */
|
||||
static char dummy;
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_MATH /* 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 enum GLmatrixtype 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;
|
||||
}
|
||||
|
||||
ALIGN16 static GLfloat s[TEST_COUNT][4];
|
||||
ALIGN16 static GLfloat d[TEST_COUNT][4];
|
||||
ALIGN16 static GLfloat r[TEST_COUNT][4];
|
||||
|
||||
static int test_transform_function( transform_func func, int psize,
|
||||
int mtype, unsigned 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 = align_malloc( 16 * sizeof(GLfloat), 16 );
|
||||
mat->type = mtypes[mtype];
|
||||
|
||||
m = mat->m;
|
||||
assert( ((long)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:
|
||||
assert(0);
|
||||
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;
|
||||
|
||||
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 ) {
|
||||
printf("-----------------------------\n" );
|
||||
printf("(i = %i, j = %i)\n", i, j );
|
||||
printf("%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] ) );
|
||||
printf("%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] ) );
|
||||
printf("%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] ) );
|
||||
printf("%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;
|
||||
unsigned 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();
|
||||
printf("counter overhead: %lu cycles\n\n", counter_overhead );
|
||||
}
|
||||
printf("transform results after hooking in %s functions:\n", description );
|
||||
}
|
||||
#endif
|
||||
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile ) {
|
||||
printf("\n" );
|
||||
for ( psize = 1 ; psize <= 4 ; psize++ ) {
|
||||
printf(" p%d\t", psize );
|
||||
}
|
||||
printf("\n--------------------------------------------------------\n" );
|
||||
}
|
||||
#endif
|
||||
|
||||
for ( mtype = 0 ; mtype < 7 ; mtype++ ) {
|
||||
for ( psize = 1 ; psize <= 4 ; psize++ ) {
|
||||
transform_func func = _mesa_transform_tab[psize][mtypes[mtype]];
|
||||
unsigned long *cycles = &(benchmark_tab[psize-1][mtype]);
|
||||
|
||||
if ( test_transform_function( func, psize, mtype, cycles ) == 0 ) {
|
||||
char buf[100];
|
||||
sprintf(buf, "_mesa_transform_tab[0][%d][%s] failed test (%s)",
|
||||
psize, mstrings[mtype], description );
|
||||
_mesa_problem( NULL, "%s", buf );
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf(" %li\t", benchmark_tab[psize-1][mtype] );
|
||||
#endif
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf(" | [%s]\n", mstrings[mtype] );
|
||||
#endif
|
||||
}
|
||||
#ifdef RUN_DEBUG_BENCHMARK
|
||||
if ( mesa_profile )
|
||||
printf( "\n" );
|
||||
#endif
|
||||
}
|
||||
|
||||
|
||||
#endif /* DEBUG_MATH */
|
||||
@@ -254,11 +254,6 @@ files_libmesa = files(
|
||||
'main/viewport.c',
|
||||
'main/viewport.h',
|
||||
'main/es1_conversion.c',
|
||||
'math/m_debug.h',
|
||||
'math/m_debug_clip.c',
|
||||
'math/m_debug_norm.c',
|
||||
'math/m_debug_util.h',
|
||||
'math/m_debug_xform.c',
|
||||
'math/m_eval.c',
|
||||
'math/m_eval.h',
|
||||
'math/m_matrix.c',
|
||||
|
||||
Reference in New Issue
Block a user