static void coldStartVars(Data * d, Work * w) {
idxint l = d->n + d->m + 1;
memset(w->u, 0, l * sizeof(pfloat));
memset(w->v, 0, l * sizeof(pfloat));
w->u[l - 1] = SQRTF((pfloat) l);
w->v[l - 1] = SQRTF((pfloat) l);
}
/**
* Compute mipmap LOD from partial derivatives.
* This the ideal solution, as given in the OpenGL spec.
*/
GLfloat
_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
{
GLfloat dudx = texW * ((s + dsdx) / (q + dqdx) - s * invQ);
GLfloat dvdx = texH * ((t + dtdx) / (q + dqdx) - t * invQ);
GLfloat dudy = texW * ((s + dsdy) / (q + dqdy) - s * invQ);
GLfloat dvdy = texH * ((t + dtdy) / (q + dqdy) - t * invQ);
GLfloat x = SQRTF(dudx * dudx + dvdx * dvdx);
GLfloat y = SQRTF(dudy * dudy + dvdy * dvdy);
GLfloat rho = MAX2(x, y);
GLfloat lambda = LOG2(rho);
return lambda;
}
static pfloat calcDualResid(Data * d, Work * w, pfloat * y, pfloat tau, pfloat *nmATy) {
idxint i;
pfloat dres = 0, scale, *dr = w->dr, *E = w->E;
*nmATy = 0;
memset(dr, 0, d->n * sizeof(pfloat));
accumByAtrans(d, w->p, y, dr); /* dr = A'y */
for (i = 0; i < d->n; ++i) {
scale = d->NORMALIZE ? E[i] / (w->sc_c * d->SCALE) : 1;
scale = scale * scale;
*nmATy += (dr[i] * dr[i]) * scale;
dres += (dr[i] + d->c[i] * tau) * (dr[i] + d->c[i] * tau) * scale;
}
*nmATy = SQRTF(*nmATy);
return SQRTF(dres); /* norm(A'y + c * tau) */
}
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 / 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);
}
}
static void setSolution(Data * d, Work * w, Sol * sol, Info * info) {
idxint l = d->n + d->m + 1;
setx(d, w, sol);
sety(d, w, sol);
sets(d, w, sol);
if (info->statusVal == 0 || info->statusVal == SOLVED) {
pfloat tau = w->u[l - 1];
pfloat kap = ABS(w->v[l - 1]);
if (tau > INDETERMINATE_TOL && tau > kap) {
info->statusVal = solved(d, sol, info, tau);
} else {
if (calcNorm(w->u, l) < INDETERMINATE_TOL * SQRTF((pfloat) l)) {
info->statusVal = indeterminate(d, sol, info);
} else {
pfloat bTy = innerProd(d->b, sol->y, d->m);
pfloat cTx = innerProd(d->c, sol->x, d->n);
if (bTy < cTx) {
info->statusVal = infeasible(d, sol, info);
} else {
info->statusVal = unbounded(d, sol, info);
}
}
}
} else if (info->statusVal == INFEASIBLE) {
info->statusVal = infeasible(d, sol, info);
} else {
info->statusVal = unbounded(d, sol, info);
}
}
static pfloat calcPrimalResid(Data * d, Work * w, pfloat * x, pfloat * s, pfloat tau, pfloat *nmAxs) {
idxint i;
pfloat pres = 0, scale, *pr = w->pr, *D = w->D;
*nmAxs = 0;
memset(pr, 0, d->m * sizeof(pfloat));
accumByA(d, w->p, x, pr);
addScaledArray(pr, s, d->m, 1.0); /* pr = Ax + s */
for (i = 0; i < d->m; ++i) {
scale = d->NORMALIZE ? D[i] / (w->sc_b * d->SCALE) : 1;
scale = scale * scale;
*nmAxs += (pr[i] * pr[i]) * scale;
pres += (pr[i] - d->b[i] * tau) * (pr[i] - d->b[i] * tau) * scale;
}
*nmAxs = SQRTF(*nmAxs);
return SQRTF(pres); /* norm(Ax + s - b * tau) */
}
/**
* Compute point size for each vertex from the vertex eye-space Z
* coordinate and the point size attenuation factors.
* Only done when point size attenuation is enabled and vertex program is
* disabled.
*/
static GLboolean
run_point_stage(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
{
if (ctx->Point._Attenuated) {
struct point_stage_data *store = POINT_STAGE_DATA(stage);
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
const GLfloat *eyeCoord = (GLfloat *) VB->EyePtr->data + 2;
const GLint eyeCoordStride = VB->EyePtr->stride / sizeof(GLfloat);
const GLfloat p0 = ctx->Point.Params[0];
const GLfloat p1 = ctx->Point.Params[1];
const GLfloat p2 = ctx->Point.Params[2];
const GLfloat pointSize = ctx->Point.Size;
GLfloat (*size)[4] = store->PointSize.data;
GLuint i;
for (i = 0; i < VB->Count; i++) {
const GLfloat dist = FABSF(*eyeCoord);
const GLfloat q = p0 + dist * (p1 + dist * p2);
const GLfloat atten = (q != 0.0F) ? SQRTF(1.0F / q) : 1.0F;
size[i][0] = pointSize * atten; /* clamping done in rasterization */
eyeCoord += eyeCoordStride;
}
VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->PointSize;
}
return GL_TRUE;
}
static pfloat fastCalcPrimalResid(Data * d, Work * w, pfloat * nmAxs) {
idxint i, n = d->n, m = d->m;
pfloat pres = 0, scale, *pr = w->pr, *D = w->D, tau = ABS(w->u[n + m]);
*nmAxs = 0;
memcpy(pr, &(w->u[n]), m * sizeof(pfloat)); /* overwrite pr */
addScaledArray(pr, &(w->u_prev[n]), m, d->ALPHA - 2);
addScaledArray(pr, &(w->u_t[n]), m, 1 - d->ALPHA);
addScaledArray(pr, d->b, m, w->u_t[n + m]); /* pr = Ax + s */
for (i = 0; i < m; ++i) {
scale = d->NORMALIZE ? D[i] / (w->sc_b * d->SCALE) : 1;
scale = scale * scale;
*nmAxs += (pr[i] * pr[i]) * scale;
pres += (pr[i] - d->b[i] * tau) * (pr[i] - d->b[i] * tau) * scale;
}
*nmAxs = SQRTF(*nmAxs);
return SQRTF(pres); /* norm(Ax + s - b * tau) */
}
scs_float calcNormDiff(const scs_float *a, const scs_float *b, scs_int l) {
scs_float nmDiff = 0.0, tmp;
scs_int i;
for (i = 0; i < l; ++i) {
tmp = (a[i] - b[i]);
nmDiff += tmp * tmp;
}
return SQRTF(nmDiff);
}
// input : current spectrum in the form of power *spec and phase *phase,
// the last two earlier spectrums are at position
// 512 and 1024 of the corresponding Input-Arrays.
// Array *vocal, which can mark an FFT_Linie as harmonic
// output: current amplitude *amp and unpredictability *cw
static void
CalcUnpred (PsyModel* m,
const int MaxLine,
const float* spec,
const float* phase,
const int* vocal,
float* amp0,
float* phs0,
float* cw )
{
int n;
float amp;
float tmp;
#define amp1 ((amp0) + 512) // amp[ 512...1023] contains data of frame-1
#define amp2 ((amp0) + 1024) // amp[1024...1535] contains data of frame-2
#define phs1 ((phs0) + 512) // phs[ 512...1023] contains data of frame-1
#define phs2 ((phs0) + 1024) // phs[1024...1535] contains data of frame-2
for ( n = 0; n < MaxLine; n++ ) {
tmp = COSF ((phs0[n] = phase[n]) - 2*phs1[n] + phs2[n]); // copy phase to output-array, predict phase and calculate predictive error
amp0[n] = SQRTF (spec[n]); // calculate and set amplitude
amp = 2*amp1[n] - amp2[n]; // predict amplitude
// calculate unpredictability
cw[n] = SQRTF (spec[n] + amp * (amp - 2*amp0[n] * tmp)) / (amp0[n] + FABS(amp));
}
// postprocessing of harmonic FFT-lines (*cw is set to CVD_UNPRED)
if ( m->CVD_used && vocal != NULL ) {
for ( n = 0; n < MAX_CVD_LINE; n++, cw++, vocal++ )
if ( *vocal != 0 && *cw > CVD_UNPRED * 0.01 * *vocal )
*cw = CVD_UNPRED * 0.01 * *vocal;
}
return;
}
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 *) _mesa_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 / 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;
}
}
}
}
_mesa_align_free( mat->m );
return 1;
}
/* ||v||_2 */
scs_float calcNorm(const scs_float * v, scs_int len) {
return SQRTF(calcNormSq(v, len));
}
/** * Generate a 4x4 transformation matrix from glRotate parameters, and * post-multiply the input matrix by it. * * \author * This function was contributed by Erich Boleyn ([email protected]). * Optimizations contributed by Rudolf Opalla ([email protected]). */ void _math_matrix_rotate( GLmatrix *mat, GLfloat angle, GLfloat x, GLfloat y, GLfloat z ) { GLfloat xx, yy, zz, xy, yz, zx, xs, ys, zs, one_c, s, c; GLfloat m[16]; GLboolean optimized; s = (GLfloat) sin( angle * DEG2RAD ); c = (GLfloat) cos( angle * DEG2RAD ); memcpy(m, Identity, sizeof(GLfloat)*16); optimized = GL_FALSE; #define M(row,col) m[col*4+row] if (x == 0.0F) { if (y == 0.0F) { if (z != 0.0F) { optimized = GL_TRUE; /* rotate only around z-axis */ M(0,0) = c; M(1,1) = c; if (z < 0.0F) { M(0,1) = s; M(1,0) = -s; } else { M(0,1) = -s; M(1,0) = s; } } } else if (z == 0.0F) { optimized = GL_TRUE; /* rotate only around y-axis */ M(0,0) = c; M(2,2) = c; if (y < 0.0F) { M(0,2) = -s; M(2,0) = s; } else { M(0,2) = s; M(2,0) = -s; } } } else if (y == 0.0F) { if (z == 0.0F) { optimized = GL_TRUE; /* rotate only around x-axis */ M(1,1) = c; M(2,2) = c; if (x < 0.0F) { M(1,2) = s; M(2,1) = -s; } else { M(1,2) = -s; M(2,1) = s; } } } if (!optimized) { const GLfloat mag = SQRTF(x * x + y * y + z * z); if (mag <= 1.0e-4) { /* no rotation, leave mat as-is */ return; } x /= mag; y /= mag; z /= mag; /* * Arbitrary axis rotation matrix. * * This is composed of 5 matrices, Rz, Ry, T, Ry', Rz', multiplied * like so: Rz * Ry * T * Ry' * Rz'. T is the final rotation * (which is about the X-axis), and the two composite transforms * Ry' * Rz' and Rz * Ry are (respectively) the rotations necessary * from the arbitrary axis to the X-axis then back. They are * all elementary rotations. * * Rz' is a rotation about the Z-axis, to bring the axis vector * into the x-z plane. Then Ry' is applied, rotating about the * Y-axis to bring the axis vector parallel with the X-axis. The * rotation about the X-axis is then performed. Ry and Rz are * simply the respective inverse transforms to bring the arbitrary * axis back to its original orientation. The first transforms * Rz' and Ry' are considered inverses, since the data from the * arbitrary axis gives you info on how to get to it, not how * to get away from it, and an inverse must be applied. * * The basic calculation used is to recognize that the arbitrary * axis vector (x, y, z), since it is of unit length, actually * represents the sines and cosines of the angles to rotate the * X-axis to the same orientation, with theta being the angle about * Z and phi the angle about Y (in the order described above) * as follows: * * cos ( theta ) = x / sqrt ( 1 - z^2 ) * sin ( theta ) = y / sqrt ( 1 - z^2 ) * * cos ( phi ) = sqrt ( 1 - z^2 ) * sin ( phi ) = z * * Note that cos ( phi ) can further be inserted to the above * formulas: * * cos ( theta ) = x / cos ( phi ) * sin ( theta ) = y / sin ( phi ) * * ...etc. Because of those relations and the standard trigonometric * relations, it is pssible to reduce the transforms down to what * is used below. It may be that any primary axis chosen will give the * same results (modulo a sign convention) using thie method. * * Particularly nice is to notice that all divisions that might * have caused trouble when parallel to certain planes or * axis go away with care paid to reducing the expressions. * After checking, it does perform correctly under all cases, since * in all the cases of division where the denominator would have * been zero, the numerator would have been zero as well, giving * the expected result. */ xx = x * x; yy = y * y; zz = z * z; xy = x * y; yz = y * z; zx = z * x; xs = x * s; ys = y * s; zs = z * s; one_c = 1.0F - c; /* We already hold the identity-matrix so we can skip some statements */ M(0,0) = (one_c * xx) + c; M(0,1) = (one_c * xy) - zs; M(0,2) = (one_c * zx) + ys; /* M(0,3) = 0.0F; */ M(1,0) = (one_c * xy) + zs; M(1,1) = (one_c * yy) + c; M(1,2) = (one_c * yz) - xs; /* M(1,3) = 0.0F; */ M(2,0) = (one_c * zx) - ys; M(2,1) = (one_c * yz) + xs; M(2,2) = (one_c * zz) + c; /* M(2,3) = 0.0F; */ /* M(3,0) = 0.0F; M(3,1) = 0.0F; M(3,2) = 0.0F; M(3,3) = 1.0F; */ } #undef M matrix_multf( mat, m, MAT_FLAG_ROTATION ); }
