static int test_transform_function( transform_func func, int psize, int mtype,
int masked, long *cycles )
{
GLvector4f source[1], dest[1], ref[1];
GLmatrix mat[1];
GLfloat *m;
GLubyte mask[TEST_COUNT];
int i, j;
#ifdef RUN_XFORM_BENCHMARK
int cycle_i; /* the counter for the benchmarks we run */
#endif
(void) cycles;
if ( psize > 4 ) {
gl_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;
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;
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++) {
mask[i] = i % 2; /* mask every 2nd element */
d[i][0] = s[i][0] = 0.0;
d[i][1] = s[i][1] = 0.0;
d[i][2] = s[i][2] = 0.0;
d[i][3] = s[i][3] = 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, NULL, 0 );
if ( mesa_profile ) {
if ( masked ) {
BEGIN_RACE( *cycles );
func( dest, mat->m, source, mask, 1 );
END_RACE( *cycles );
} else {
BEGIN_RACE( *cycles );
func( dest, mat->m, source, NULL, 0 );
END_RACE( *cycles );
}
}
else {
if ( masked ) {
func( dest, mat->m, source, mask, 1 );
} else {
func( dest, mat->m, source, NULL, 0 );
}
}
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
if ( masked && (mask[i] & 1) )
continue;
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;
}
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 = (GLfloat *) 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 ) {
_mesa_printf("-----------------------------\n" );
_mesa_printf("(i = %i, j = %i)\n", i, j );
_mesa_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] ) );
_mesa_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] ) );
_mesa_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] ) );
_mesa_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;
}
