void test_cbrt()
{
static_assert((std::is_same<decltype(cbrt((double)0)), double>::value), "");
static_assert((std::is_same<decltype(cbrtf(0)), float>::value), "");
static_assert((std::is_same<decltype(cbrtl(0)), long double>::value), "");
assert(cbrt(1) == 1);
}
/*
* Compute the target congestion window for the next RTT according to
* cubic equation when an ack is received.
*
* W(t) = C(t-K)^3 + W(last_max)
*/
static uint32_t
tcp_cubic_update(struct tcpcb *tp, u_int32_t rtt)
{
float K, var;
u_int32_t elapsed_time, win;
win = min(tp->snd_cwnd, tp->snd_wnd);
if (tp->t_ccstate->cub_last_max == 0)
tp->t_ccstate->cub_last_max = tp->snd_ssthresh;
if (tp->t_ccstate->cub_epoch_start == 0) {
/*
* This is the beginning of a new epoch, initialize some of
* the variables that we need to use for computing the
* congestion window later.
*/
tp->t_ccstate->cub_epoch_start = tcp_now;
if (tp->t_ccstate->cub_epoch_start == 0)
tp->t_ccstate->cub_epoch_start = 1;
if (win < tp->t_ccstate->cub_last_max) {
VERIFY(current_task() == kernel_task);
/*
* Compute cubic epoch period, this is the time
* period that the window will take to increase to
* last_max again after backoff due to loss.
*/
K = (tp->t_ccstate->cub_last_max - win)
/ tp->t_maxseg / tcp_cubic_coeff;
K = cbrtf(K);
tp->t_ccstate->cub_epoch_period = K * TCP_RETRANSHZ;
/* Origin point */
tp->t_ccstate->cub_origin_point =
tp->t_ccstate->cub_last_max;
} else {
tp->t_ccstate->cub_epoch_period = 0;
tp->t_ccstate->cub_origin_point = win;
}
tp->t_ccstate->cub_target_win = 0;
}
VERIFY(tp->t_ccstate->cub_origin_point > 0);
/*
* Compute the target window for the next RTT using smoothed RTT
* as an estimate for next RTT.
*/
elapsed_time = timer_diff(tcp_now, 0,
tp->t_ccstate->cub_epoch_start, 0);
if (tcp_cubic_use_minrtt)
elapsed_time += max(tcp_cubic_use_minrtt, rtt);
else
elapsed_time += rtt;
var = (elapsed_time - tp->t_ccstate->cub_epoch_period) / TCP_RETRANSHZ;
var = var * var * var * (tcp_cubic_coeff * tp->t_maxseg);
tp->t_ccstate->cub_target_win = tp->t_ccstate->cub_origin_point + var;
return (tp->t_ccstate->cub_target_win);
}
/**
* @brief Restores the state of the last simulation or places the particles in
*the shape of a cube if no state is found.
*
* @param[out] buffer Points to the particles buffer to be filled
* @param[in] parameters Contains the simulation parameters
*
*/
void sph_simulation::init_particles(particle* buffer,
const simulation_parameters& parameters) {
int particles_per_cube_side = ceil(cbrtf(parameters.particles_count));
float side_length = cbrtf(initial_volume);
float spacing = side_length / (float)particles_per_cube_side;
std::cout << "volume: " << initial_volume << " side_length: " << side_length
<< " spacing: " << spacing << std::endl;
// Last simualtion serialized its last frame
// Lets load that and pick up where it let off
std::filebuf fb;
if (fb.open("last_frame.bin", std::ios::in)) {
std::istream file_in(&fb);
cereal::BinaryInputArchive archive(file_in);
archive.loadBinary(buffer, sizeof(particle) * parameters.particles_count);
fb.close();
}
// No serialized particle array was found
// Initialize the particles in their default position
else {
for (unsigned int i = 0; i < parameters.particles_count; ++i) {
// Arrange the particles in the form of a cube
buffer[i].position.s[0] =
(float)(i % particles_per_cube_side) * spacing - side_length / 2.f;
buffer[i].position.s[1] =
((float)((i / particles_per_cube_side) % particles_per_cube_side) *
spacing);
buffer[i].position.s[2] =
(float)(i / (particles_per_cube_side * particles_per_cube_side)) *
spacing -
side_length / 2.f;
buffer[i].velocity.s[0] = 0.f;
buffer[i].velocity.s[1] = 0.f;
buffer[i].velocity.s[2] = 0.f;
buffer[i].intermediate_velocity.s[0] = 0.f;
buffer[i].intermediate_velocity.s[1] = 0.f;
buffer[i].intermediate_velocity.s[2] = 0.f;
buffer[i].density = 0.f;
buffer[i].pressure = 0.f;
}
}
}
inline float MathTrait<float>::cbrt( const float val )
{
#ifdef _MSC_VER
return cbrt( val ) ;
#else
return cbrtf( val ) ;
#endif
}
static float
husl_f(float t)
{
if(t > lab_e)
return cbrtf(t);
else
return 7.787f * t + 16.0f / 116.0f;
}
static float cubic_db_to_def(const float db)
{
if (db == 0.0f)
return 1.0f;
else if (db == -INFINITY)
return 0.0f;
return cbrtf(db_to_mul(db));
}
/**
* Calculate the "cubic deadband" system parameters.
* @param[in] The width of the deadband
* @param[in] Slope of deadband at in=0, sane values between 0 and 1.
* @param[out] m cubic weighting of function
* @param[out] integrated response at in=w
*/
static void vtol_deadband_setup(float w, float b, float *m, float *r)
{
/* So basically.. we want the function to be tangent to the
** linear sections-- have a slope of 1-- at -w and w. In the
** middle we want a slope of b. So the cube here does all the
** work b isn't doing. */
*m = cbrtf((1-b)/(3*powf(w,2)));
*r = powf(*m*w, 3)+b*w;
}
/* Assumes mvir in Msun and dt in Myr. */
inline float halo_tidal_range(float mvir, float av_a) {
/* TIDAL_FORCE_LIMIT = Gc*m/r^3 (km/s/Myr per comoving Mpc) */
/* r^3 = Gc*m/TIDAL_FORCE_LIMIT / a^2 */
/* Mutiply by 2 to be safe.*/
float tforce_limit = TIDAL_FORCE_LIMIT;
if (tidal_extra_range && TIDAL_FORCE_LIMIT > 0.001) {
tforce_limit = 0.01 * TIDAL_FORCE_LIMIT;
if (tforce_limit < 0.001) tforce_limit = 0.001;
}
return (2*cbrtf(fabs(mvir*Gc/(tforce_limit*av_a*av_a))));
}
// Possible cube root optimization.
// (borrowed from metamerist.com)
static inline float cbrtf2 ( float x )
{
#ifdef CONFIG_FLOAT32
// Avoid strict-aliasing optimization (gcc -O2).
union { float f; int i; } u;
u.f = x;
u.i = (u.i / 3) + 710235478;
return u.f;
#else
return cbrtf(x);
#endif
}
/* } */
void setup_system(atom a[]){
// initialize energy array
//for () {
//}
float width = ceil(cbrtf((float) NA));
first_L = cbrtf((float) NA/(DENSITY*100*100*100*6.022E23/39.948));
L = first_L;
float spacing = L/width;
float offset = spacing/2.0;
size_t count = 0;
for(float xpos = offset; xpos <= spacing*width; xpos +=spacing){
for(float ypos = offset; ypos <= spacing*width; ypos +=spacing){
for(float zpos = offset; zpos <= spacing*width; zpos +=spacing){
a[count].atomic_number = 18;
a[count].x = xpos;
a[count].y = ypos;
a[count].z = zpos;
count++;
}
}
}
}
static long
rgbaf_to_Labaf (float *src,
float *dst,
long samples)
{
long n = samples;
while (n--)
{
float r = src[0];
float g = src[1];
float b = src[2];
float a = src[3];
float xr = 0.43603516f / D50_WHITE_REF_X * r + 0.38511658f / D50_WHITE_REF_X * g + 0.14305115f / D50_WHITE_REF_X * b;
float yr = 0.22248840f / D50_WHITE_REF_Y * r + 0.71690369f / D50_WHITE_REF_Y * g + 0.06060791f / D50_WHITE_REF_Y * b;
float zr = 0.01391602f / D50_WHITE_REF_Z * r + 0.09706116f / D50_WHITE_REF_Z * g + 0.71392822f / D50_WHITE_REF_Z * b;
float fx = xr > LAB_EPSILON ? cbrtf (xr) : (LAB_KAPPA * xr + 16.0f) / 116.0f;
float fy = yr > LAB_EPSILON ? cbrtf (yr) : (LAB_KAPPA * yr + 16.0f) / 116.0f;
float fz = zr > LAB_EPSILON ? cbrtf (zr) : (LAB_KAPPA * zr + 16.0f) / 116.0f;
float L = 116.0f * fy - 16.0f;
float A = 500.0f * (fx - fy);
float B = 200.0f * (fy - fz);
dst[0] = L;
dst[1] = A;
dst[2] = B;
dst[3] = a;
src += 4;
dst += 4;
}
return samples;
}
void Player::HitBall()
{
for (size_t i = 0; i < _ballManager->GetNumberOfBalls(); i++)
{
Ball& ball = *_ballManager->GetBallAtIndex(i);
// don't test if ball belongs to other player
// not enough balls to justify it.
if (!ball.IsContained())
{
Vec2 ppos = this->convertToWorldSpace(Vec2());
Vec2 bpos = ball.getParent()->convertToWorldSpace(ball.getPosition());
Vec2 toBall = bpos - ppos;
float r = 100;
if (toBall.length() < r)
{
if (ball.getType() == BombBall::type)
{
Vec2 emptySpace; //Basically a placeholder because this will only be used in case of bomb balls, but hit requires a vec2 even if one is not used.
ball.Hit(emptySpace);
PlayerHitByBall(nullptr, &ball);
((Game_Scene*)(this->getParent()->getParent()))->SeeSaw(this, -4);
}
else if (ball.getType() == WalletBall::type)
{
Vec2 emptySpace; //Basically a placeholder because this will only be used in case of bomb balls, but hit requires a vec2 even if one is not used.
ball.Hit(emptySpace);
((Game_Scene*)(this->getParent()->getParent()))->SeeSaw(this, 4);
AudioHelper::PlayRandom("Moneyball", 2);
}
else
{
float difficulty = 0.1f; // 0=easy, 1=hard
float dy = toBall.y / r; // -1 -> 1
dy = (dy > 0 ? 1 : -1) * pow(abs(cbrtf(dy)), (1.0f - difficulty)); // cubic curve, harder to get y just right
dy = (dy + 1) / 2.0f; // 0 -> 1 for lerp
Vec2 hitDir = ccpLerp(Vec2(Settings::horizontalSpeed, -250), Vec2(Settings::horizontalSpeed, 550), dy); // lerp between mim/max hit strength
ball.Hit(hitDir);
}
AudioHelper::PlayRandom("bathitting", 3, 0.75 + rand_0_1() * 0.5);
}
}
}
}
void volume_move(){
total_vol_moves+=1.0;
float old_e = total_energy(old_argons);
float old_vol = powf(L,3);
float new_vol = expf(logf(old_vol) + (random() - 0.5)*VOL_MOVE);
float new_L = cbrtf(new_vol);
for(size_t i = NA; i>0;i--){
new_argons[i].x*=(new_L/L);
new_argons[i].y*=(new_L/L);
new_argons[i].z*=(new_L/L);
}
float new_e = total_energy(new_argons);
if (random()<expf(-(1.0/(KB *TEMPERATURE))*((new_e - old_e) + (PRESSURE*(new_vol- old_vol)) - ((NA+1)*(KB*TEMPERATURE)*logf(new_vol/old_vol))))){
accepted_vol_moves +=1.0;
memcpy(old_argons, new_argons, sizeof(new_argons));
L=new_L;
} else{
memcpy(new_argons,old_argons,sizeof(new_argons));
}
}
static long
Yaf_to_Laf (float *src,
float *dst,
long samples)
{
long n = samples;
while (n--)
{
float yr = src[0];
float a = src[1];
float L = yr > LAB_EPSILON ? 116.0 * cbrtf (yr) - 16 : LAB_KAPPA * yr;
dst[0] = L;
dst[1] = a;
src += 2;
dst += 2;
}
return samples;
}
static TACommandVerdict cbrtf_cmd(TAThread thread,TAInputStream stream)
{
float x, res;
// Prepare
x = readFloat(&stream);
START_TARGET_OPERATION(thread);
// Execute
res = cbrtf(x);
END_TARGET_OPERATION(thread);
// Response
writeFloat(thread, res);
sendResponse(thread);
return taDefaultVerdict;
}
int PolynomialSolver::solveCubic(float c[4], float s[3])
{
int i, num;
float sub,
A, B, C,
sq_A, p, q,
cb_p, D;
// normalize the equation:x ^ 3 + Ax ^ 2 + Bx + C = 0
A = c[2] / c[3];
B = c[1] / c[3];
C = c[0] / c[3];
// substitute x = y - A / 3 to eliminate the quadric term: x^3 + px + q = 0
sq_A = A * A;
p = 1.0f/3.0f * (-1.0f/3.0f * sq_A + B);
q = 1.0f/2.0f * (2.0f/27.0f * A *sq_A - 1.0f/3.0f * A * B + C);
// use Cardano's formula
cb_p = p * p * p;
D = q * q + cb_p;
if(isZero(D))
{
if(isZero(q))
{
// one triple solution
s[0] = 0.0f;
num = 1;
}
else
{
// one single and one float solution
float u = cbrtf(-q);
s[0] = 2.0f * u;
s[1] = - u;
num = 2;
}
}
else if(D < 0.0)
{
// casus irreductibilis: three real solutions
float phi = 1.0f/3.0f * std::acos(-q / std::sqrt(-cb_p));
float t = 2.0f * std::sqrt(-p);
s[0] = t * std::cos(phi);
s[1] = -t * std::cos(phi + pi / 3.0f);
s[2] = -t * std::cos(phi - pi / 3.0f);
num = 3;
}
else
{
// one real solution
float sqrt_D = std::sqrt(D);
float u = cbrtf(sqrt_D + std::abs(q));
if(q > 0.0f)
s[0] = - u + p / u;
else
s[0] = u - p / u;
num = 1;
}
// resubstitute
sub = 1.0f / 3.0f * A;
for(i = 0; i < num; i++)
s[i] -= sub;
return num;
}
inline float BB_TABLE_UNMAP(float T) {
// return powf ((T - BB_DRAPER) / BB_TABLE_SPACING, 1.0f/BB_TABLE_XPOWER);
float t = (T - BB_DRAPER) / BB_TABLE_SPACING;
float ic = cbrtf(t);
return ic * ic; // ^2/3
}
__device__ inline float occaCuda_fastCbrt(const float x){ return cbrtf(x); }
/**
* Calculate rate distortion cost for quantizing with given codebook
*
* @return quantization distortion
*/
static av_always_inline float quantize_and_encode_band_cost_template(
struct AACEncContext *s,
PutBitContext *pb, const float *in,
const float *scaled, int size, int scale_idx,
int cb, const float lambda, const float uplim,
int *bits, int BT_ZERO, int BT_UNSIGNED,
int BT_PAIR, int BT_ESC)
{
const float IQ = ff_aac_pow2sf_tab[POW_SF2_ZERO + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
const float Q = ff_aac_pow2sf_tab[POW_SF2_ZERO - scale_idx + SCALE_ONE_POS - SCALE_DIV_512];
const float CLIPPED_ESCAPE = 165140.0f*IQ;
int i, j;
float cost = 0;
const int dim = BT_PAIR ? 2 : 4;
int resbits = 0;
const float Q34 = sqrtf(Q * sqrtf(Q));
const int range = aac_cb_range[cb];
const int maxval = aac_cb_maxval[cb];
int off;
if (BT_ZERO) {
for (i = 0; i < size; i++)
cost += in[i]*in[i];
if (bits)
*bits = 0;
return cost * lambda;
}
if (!scaled) {
abs_pow34_v(s->scoefs, in, size);
scaled = s->scoefs;
}
quantize_bands(s->qcoefs, in, scaled, size, Q34, !BT_UNSIGNED, maxval);
if (BT_UNSIGNED) {
off = 0;
} else {
off = maxval;
}
for (i = 0; i < size; i += dim) {
const float *vec;
int *quants = s->qcoefs + i;
int curidx = 0;
int curbits;
float rd = 0.0f;
for (j = 0; j < dim; j++) {
curidx *= range;
curidx += quants[j] + off;
}
curbits = ff_aac_spectral_bits[cb-1][curidx];
vec = &ff_aac_codebook_vectors[cb-1][curidx*dim];
if (BT_UNSIGNED) {
for (j = 0; j < dim; j++) {
float t = fabsf(in[i+j]);
float di;
if (BT_ESC && vec[j] == 64.0f) { //FIXME: slow
if (t >= CLIPPED_ESCAPE) {
di = t - CLIPPED_ESCAPE;
curbits += 21;
} else {
int c = av_clip(quant(t, Q), 0, 8191);
di = t - c*cbrtf(c)*IQ;
curbits += av_log2(c)*2 - 4 + 1;
}
} else {
di = t - vec[j]*IQ;
}
if (vec[j] != 0.0f)
curbits++;
rd += di*di;
}
} else {
for (j = 0; j < dim; j++) {
float di = in[i+j] - vec[j]*IQ;
rd += di*di;
}
}
cost += rd * lambda + curbits;
resbits += curbits;
if (cost >= uplim)
return uplim;
if (pb) {
put_bits(pb, ff_aac_spectral_bits[cb-1][curidx], ff_aac_spectral_codes[cb-1][curidx]);
if (BT_UNSIGNED)
for (j = 0; j < dim; j++)
if (ff_aac_codebook_vectors[cb-1][curidx*dim+j] != 0.0f)
put_bits(pb, 1, in[i+j] < 0.0f);
if (BT_ESC) {
for (j = 0; j < 2; j++) {
if (ff_aac_codebook_vectors[cb-1][curidx*2+j] == 64.0f) {
int coef = av_clip(quant(fabsf(in[i+j]), Q), 0, 8191);
int len = av_log2(coef);
put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
put_bits(pb, len, coef & ((1 << len) - 1));
}
}
}
}
}
if (bits)
*bits = resbits;
return cost;
}
/**
* Calculate rate distortion cost for quantizing with given codebook
*
* @return quantization distortion
*/
static float quantize_band_cost(struct AACEncContext *s, const float *in,
const float *scaled, int size, int scale_idx,
int cb, const float lambda, const float uplim,
int *bits)
{
const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512];
const float CLIPPED_ESCAPE = 165140.0f*IQ;
int i, j, k;
float cost = 0;
const int dim = cb < FIRST_PAIR_BT ? 4 : 2;
int resbits = 0;
#ifndef USE_REALLY_FULL_SEARCH
const float Q34 = sqrtf(Q * sqrtf(Q));
const int range = aac_cb_range[cb];
const int maxval = aac_cb_maxval[cb];
int offs[4];
#endif /* USE_REALLY_FULL_SEARCH */
if (!cb) {
for (i = 0; i < size; i++)
cost += in[i]*in[i];
if (bits)
*bits = 0;
return cost * lambda;
}
#ifndef USE_REALLY_FULL_SEARCH
offs[0] = 1;
for (i = 1; i < dim; i++)
offs[i] = offs[i-1]*range;
quantize_bands(s->qcoefs, in, scaled, size, Q34, !IS_CODEBOOK_UNSIGNED(cb), maxval);
#endif /* USE_REALLY_FULL_SEARCH */
for (i = 0; i < size; i += dim) {
float mincost;
int minidx = 0;
int minbits = 0;
const float *vec;
#ifndef USE_REALLY_FULL_SEARCH
int (*quants)[2] = &s->qcoefs[i];
mincost = 0.0f;
for (j = 0; j < dim; j++)
mincost += in[i+j]*in[i+j];
minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
minbits = ff_aac_spectral_bits[cb-1][minidx];
mincost = mincost * lambda + minbits;
for (j = 0; j < (1<<dim); j++) {
float rd = 0.0f;
int curbits;
int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
int same = 0;
for (k = 0; k < dim; k++) {
if ((j & (1 << k)) && quants[k][0] == quants[k][1]) {
same = 1;
break;
}
}
if (same)
continue;
for (k = 0; k < dim; k++)
curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k];
curbits = ff_aac_spectral_bits[cb-1][curidx];
vec = &ff_aac_codebook_vectors[cb-1][curidx*dim];
#else
mincost = INFINITY;
vec = ff_aac_codebook_vectors[cb-1];
for (j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim) {
float rd = 0.0f;
int curbits = ff_aac_spectral_bits[cb-1][j];
#endif /* USE_REALLY_FULL_SEARCH */
if (IS_CODEBOOK_UNSIGNED(cb)) {
for (k = 0; k < dim; k++) {
float t = fabsf(in[i+k]);
float di;
if (vec[k] == 64.0f) { //FIXME: slow
//do not code with escape sequence small values
if (t < 39.0f*IQ) {
rd = INFINITY;
break;
}
if (t >= CLIPPED_ESCAPE) {
di = t - CLIPPED_ESCAPE;
curbits += 21;
} else {
int c = av_clip(quant(t, Q), 0, 8191);
di = t - c*cbrtf(c)*IQ;
curbits += av_log2(c)*2 - 4 + 1;
}
} else {
di = t - vec[k]*IQ;
}
if (vec[k] != 0.0f)
curbits++;
rd += di*di;
}
} else {
for (k = 0; k < dim; k++) {
float di = in[i+k] - vec[k]*IQ;
rd += di*di;
}
}
rd = rd * lambda + curbits;
if (rd < mincost) {
mincost = rd;
minidx = j;
minbits = curbits;
}
}
cost += mincost;
resbits += minbits;
if (cost >= uplim)
return uplim;
}
if (bits)
*bits = resbits;
return cost;
}
static void quantize_and_encode_band(struct AACEncContext *s, PutBitContext *pb,
const float *in, int size, int scale_idx,
int cb, const float lambda)
{
const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512];
const float CLIPPED_ESCAPE = 165140.0f*IQ;
const int dim = (cb < FIRST_PAIR_BT) ? 4 : 2;
int i, j, k;
#ifndef USE_REALLY_FULL_SEARCH
const float Q34 = sqrtf(Q * sqrtf(Q));
const int range = aac_cb_range[cb];
const int maxval = aac_cb_maxval[cb];
int offs[4];
float *scaled = s->scoefs;
#endif /* USE_REALLY_FULL_SEARCH */
//START_TIMER
if (!cb)
return;
#ifndef USE_REALLY_FULL_SEARCH
offs[0] = 1;
for (i = 1; i < dim; i++)
offs[i] = offs[i-1]*range;
abs_pow34_v(scaled, in, size);
quantize_bands(s->qcoefs, in, scaled, size, Q34, !IS_CODEBOOK_UNSIGNED(cb), maxval);
#endif /* USE_REALLY_FULL_SEARCH */
for (i = 0; i < size; i += dim) {
float mincost;
int minidx = 0;
int minbits = 0;
const float *vec;
#ifndef USE_REALLY_FULL_SEARCH
int (*quants)[2] = &s->qcoefs[i];
mincost = 0.0f;
for (j = 0; j < dim; j++)
mincost += in[i+j]*in[i+j];
minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
minbits = ff_aac_spectral_bits[cb-1][minidx];
mincost = mincost * lambda + minbits;
for (j = 0; j < (1<<dim); j++) {
float rd = 0.0f;
int curbits;
int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
int same = 0;
for (k = 0; k < dim; k++) {
if ((j & (1 << k)) && quants[k][0] == quants[k][1]) {
same = 1;
break;
}
}
if (same)
continue;
for (k = 0; k < dim; k++)
curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k];
curbits = ff_aac_spectral_bits[cb-1][curidx];
vec = &ff_aac_codebook_vectors[cb-1][curidx*dim];
#else
vec = ff_aac_codebook_vectors[cb-1];
mincost = INFINITY;
for (j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim) {
float rd = 0.0f;
int curbits = ff_aac_spectral_bits[cb-1][j];
int curidx = j;
#endif /* USE_REALLY_FULL_SEARCH */
if (IS_CODEBOOK_UNSIGNED(cb)) {
for (k = 0; k < dim; k++) {
float t = fabsf(in[i+k]);
float di;
if (vec[k] == 64.0f) { //FIXME: slow
//do not code with escape sequence small values
if (t < 39.0f*IQ) {
rd = INFINITY;
break;
}
if (t >= CLIPPED_ESCAPE) {
di = t - CLIPPED_ESCAPE;
curbits += 21;
} else {
int c = av_clip(quant(t, Q), 0, 8191);
di = t - c*cbrtf(c)*IQ;
curbits += av_log2(c)*2 - 4 + 1;
}
} else {
di = t - vec[k]*IQ;
}
if (vec[k] != 0.0f)
curbits++;
rd += di*di;
}
} else {
for (k = 0; k < dim; k++) {
float di = in[i+k] - vec[k]*IQ;
rd += di*di;
}
}
rd = rd * lambda + curbits;
if (rd < mincost) {
mincost = rd;
minidx = curidx;
minbits = curbits;
}
}
put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]);
if (IS_CODEBOOK_UNSIGNED(cb))
for (j = 0; j < dim; j++)
if (ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f)
put_bits(pb, 1, in[i+j] < 0.0f);
if (cb == ESC_BT) {
for (j = 0; j < 2; j++) {
if (ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f) {
int coef = av_clip(quant(fabsf(in[i+j]), Q), 0, 8191);
int len = av_log2(coef);
put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
put_bits(pb, len, coef & ((1 << len) - 1));
}
}
}
}
//STOP_TIMER("quantize_and_encode")
}
/**
* structure used in optimal codebook search
*/
typedef struct BandCodingPath {
int prev_idx; ///< pointer to the previous path point
float cost; ///< path cost
int run;
} BandCodingPath;
/**
* Encode band info for single window group bands.
*/
static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce,
int win, int group_len, const float lambda)
{
BandCodingPath path[120][12];
int w, swb, cb, start, start2, size;
int i, j;
const int max_sfb = sce->ics.max_sfb;
const int run_bits = sce->ics.num_windows == 1 ? 5 : 3;
const int run_esc = (1 << run_bits) - 1;
int idx, ppos, count;
int stackrun[120], stackcb[120], stack_len;
float next_minrd = INFINITY;
int next_mincb = 0;
abs_pow34_v(s->scoefs, sce->coeffs, 1024);
start = win*128;
for (cb = 0; cb < 12; cb++) {
path[0][cb].cost = 0.0f;
path[0][cb].prev_idx = -1;
path[0][cb].run = 0;
}
for (swb = 0; swb < max_sfb; swb++) {
start2 = start;
size = sce->ics.swb_sizes[swb];
if (sce->zeroes[win*16 + swb]) {
for (cb = 0; cb < 12; cb++) {
path[swb+1][cb].prev_idx = cb;
path[swb+1][cb].cost = path[swb][cb].cost;
path[swb+1][cb].run = path[swb][cb].run + 1;
}
} else {
float minrd = next_minrd;
int mincb = next_mincb;
next_minrd = INFINITY;
next_mincb = 0;
for (cb = 0; cb < 12; cb++) {
float cost_stay_here, cost_get_here;
float rd = 0.0f;
for (w = 0; w < group_len; w++) {
FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(win+w)*16+swb];
rd += quantize_band_cost(s, sce->coeffs + start + w*128,
s->scoefs + start + w*128, size,
sce->sf_idx[(win+w)*16+swb], cb,
lambda / band->threshold, INFINITY, NULL);
}
cost_stay_here = path[swb][cb].cost + rd;
cost_get_here = minrd + rd + run_bits + 4;
if ( run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run]
!= run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run+1])
cost_stay_here += run_bits;
if (cost_get_here < cost_stay_here) {
path[swb+1][cb].prev_idx = mincb;
path[swb+1][cb].cost = cost_get_here;
path[swb+1][cb].run = 1;
} else {
path[swb+1][cb].prev_idx = cb;
path[swb+1][cb].cost = cost_stay_here;
path[swb+1][cb].run = path[swb][cb].run + 1;
}
if (path[swb+1][cb].cost < next_minrd) {
next_minrd = path[swb+1][cb].cost;
next_mincb = cb;
}
}
}
start += sce->ics.swb_sizes[swb];
}
//convert resulting path from backward-linked list
stack_len = 0;
idx = 0;
for (cb = 1; cb < 12; cb++)
if (path[max_sfb][cb].cost < path[max_sfb][idx].cost)
idx = cb;
ppos = max_sfb;
while (ppos > 0) {
cb = idx;
stackrun[stack_len] = path[ppos][cb].run;
stackcb [stack_len] = cb;
idx = path[ppos-path[ppos][cb].run+1][cb].prev_idx;
ppos -= path[ppos][cb].run;
stack_len++;
}
//perform actual band info encoding
start = 0;
for (i = stack_len - 1; i >= 0; i--) {
put_bits(&s->pb, 4, stackcb[i]);
count = stackrun[i];
memset(sce->zeroes + win*16 + start, !stackcb[i], count);
//XXX: memset when band_type is also uint8_t
for (j = 0; j < count; j++) {
sce->band_type[win*16 + start] = stackcb[i];
start++;
}
while (count >= run_esc) {
put_bits(&s->pb, run_bits, run_esc);
count -= run_esc;
}
put_bits(&s->pb, run_bits, count);
}
}
typedef struct TrellisPath {
float cost;
int prev;
int min_val;
int max_val;
} TrellisPath;
#define TRELLIS_STAGES 121
#define TRELLIS_STATES 256
static void search_for_quantizers_anmr(AVCodecContext *avctx, AACEncContext *s,
SingleChannelElement *sce,
const float lambda)
{
int q, w, w2, g, start = 0;
int i, j;
int idx;
TrellisPath paths[TRELLIS_STAGES][TRELLIS_STATES];
int bandaddr[TRELLIS_STAGES];
int minq;
float mincost;
for (i = 0; i < TRELLIS_STATES; i++) {
paths[0][i].cost = 0.0f;
paths[0][i].prev = -1;
paths[0][i].min_val = i;
paths[0][i].max_val = i;
}
for (j = 1; j < TRELLIS_STAGES; j++) {
for (i = 0; i < TRELLIS_STATES; i++) {
paths[j][i].cost = INFINITY;
paths[j][i].prev = -2;
paths[j][i].min_val = INT_MAX;
paths[j][i].max_val = 0;
}
}
idx = 1;
abs_pow34_v(s->scoefs, sce->coeffs, 1024);
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = w*128;
for (g = 0; g < sce->ics.num_swb; g++) {
const float *coefs = sce->coeffs + start;
float qmin, qmax;
int nz = 0;
bandaddr[idx] = w * 16 + g;
qmin = INT_MAX;
qmax = 0.0f;
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g];
if (band->energy <= band->threshold || band->threshold == 0.0f) {
sce->zeroes[(w+w2)*16+g] = 1;
continue;
}
sce->zeroes[(w+w2)*16+g] = 0;
nz = 1;
for (i = 0; i < sce->ics.swb_sizes[g]; i++) {
float t = fabsf(coefs[w2*128+i]);
if (t > 0.0f)
qmin = FFMIN(qmin, t);
qmax = FFMAX(qmax, t);
}
}
if (nz) {
int minscale, maxscale;
float minrd = INFINITY;
//minimum scalefactor index is when minimum nonzero coefficient after quantizing is not clipped
minscale = av_clip_uint8(log2(qmin)*4 - 69 + SCALE_ONE_POS - SCALE_DIV_512);
//maximum scalefactor index is when maximum coefficient after quantizing is still not zero
maxscale = av_clip_uint8(log2(qmax)*4 + 6 + SCALE_ONE_POS - SCALE_DIV_512);
for (q = minscale; q < maxscale; q++) {
float dists[12], dist;
memset(dists, 0, sizeof(dists));
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g];
int cb;
for (cb = 0; cb <= ESC_BT; cb++)
dists[cb] += quantize_band_cost(s, coefs + w2*128, s->scoefs + start + w2*128, sce->ics.swb_sizes[g],
q, cb, lambda / band->threshold, INFINITY, NULL);
}
dist = dists[0];
for (i = 1; i <= ESC_BT; i++)
dist = FFMIN(dist, dists[i]);
minrd = FFMIN(minrd, dist);
for (i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, TRELLIS_STATES); i++) {
float cost;
int minv, maxv;
if (isinf(paths[idx - 1][i].cost))
continue;
cost = paths[idx - 1][i].cost + dist
+ ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO];
minv = FFMIN(paths[idx - 1][i].min_val, q);
maxv = FFMAX(paths[idx - 1][i].max_val, q);
if (cost < paths[idx][q].cost && maxv-minv < SCALE_MAX_DIFF) {
paths[idx][q].cost = cost;
paths[idx][q].prev = i;
paths[idx][q].min_val = minv;
paths[idx][q].max_val = maxv;
}
}
}
} else {
for (q = 0; q < TRELLIS_STATES; q++) {
if (!isinf(paths[idx - 1][q].cost)) {
paths[idx][q].cost = paths[idx - 1][q].cost + 1;
paths[idx][q].prev = q;
paths[idx][q].min_val = FFMIN(paths[idx - 1][q].min_val, q);
paths[idx][q].max_val = FFMAX(paths[idx - 1][q].max_val, q);
continue;
}
for (i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, TRELLIS_STATES); i++) {
float cost;
int minv, maxv;
if (isinf(paths[idx - 1][i].cost))
continue;
cost = paths[idx - 1][i].cost + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO];
minv = FFMIN(paths[idx - 1][i].min_val, q);
maxv = FFMAX(paths[idx - 1][i].max_val, q);
if (cost < paths[idx][q].cost && maxv-minv < SCALE_MAX_DIFF) {
paths[idx][q].cost = cost;
paths[idx][q].prev = i;
paths[idx][q].min_val = minv;
paths[idx][q].max_val = maxv;
}
}
}
}
sce->zeroes[w*16+g] = !nz;
start += sce->ics.swb_sizes[g];
idx++;
}
}
idx--;
mincost = paths[idx][0].cost;
minq = 0;
for (i = 1; i < TRELLIS_STATES; i++) {
if (paths[idx][i].cost < mincost) {
mincost = paths[idx][i].cost;
minq = i;
}
}
while (idx) {
sce->sf_idx[bandaddr[idx]] = minq;
minq = paths[idx][minq].prev;
idx--;
}
//set the same quantizers inside window groups
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
for (g = 0; g < sce->ics.num_swb; g++)
for (w2 = 1; w2 < sce->ics.group_len[w]; w2++)
sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g];
}
/**
* Calculate rate distortion cost for quantizing with given codebook
*
* @return quantization distortion
*/
static float quantize_and_encode_band_cost(struct AACEncContext *s,
PutBitContext *pb, const float *in,
const float *scaled, int size, int scale_idx,
int cb, const float lambda, const float uplim,
int *bits)
{
const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512];
const float CLIPPED_ESCAPE = 165140.0f*IQ;
int i, j, k;
float cost = 0;
const int dim = cb < FIRST_PAIR_BT ? 4 : 2;
int resbits = 0;
#ifndef USE_REALLY_FULL_SEARCH
const float Q34 = sqrtf(Q * sqrtf(Q));
const int range = aac_cb_range[cb];
const int maxval = aac_cb_maxval[cb];
int offs[4];
#endif /* USE_REALLY_FULL_SEARCH */
if (!cb) {
for (i = 0; i < size; i++)
cost += in[i]*in[i];
if (bits)
*bits = 0;
return cost * lambda;
}
#ifndef USE_REALLY_FULL_SEARCH
offs[0] = 1;
for (i = 1; i < dim; i++)
offs[i] = offs[i-1]*range;
if (!scaled) {
abs_pow34_v(s->scoefs, in, size);
scaled = s->scoefs;
}
quantize_bands(s->qcoefs, in, scaled, size, Q34, !IS_CODEBOOK_UNSIGNED(cb), maxval);
#endif /* USE_REALLY_FULL_SEARCH */
for (i = 0; i < size; i += dim) {
float mincost;
int minidx = 0;
int minbits = 0;
const float *vec;
#ifndef USE_REALLY_FULL_SEARCH
int (*quants)[2] = &s->qcoefs[i];
mincost = 0.0f;
for (j = 0; j < dim; j++)
mincost += in[i+j]*in[i+j];
minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
minbits = ff_aac_spectral_bits[cb-1][minidx];
mincost = mincost * lambda + minbits;
for (j = 0; j < (1<<dim); j++) {
float rd = 0.0f;
int curbits;
int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40;
int same = 0;
for (k = 0; k < dim; k++) {
if ((j & (1 << k)) && quants[k][0] == quants[k][1]) {
same = 1;
break;
}
}
if (same)
continue;
for (k = 0; k < dim; k++)
curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k];
curbits = ff_aac_spectral_bits[cb-1][curidx];
vec = &ff_aac_codebook_vectors[cb-1][curidx*dim];
#else
mincost = INFINITY;
vec = ff_aac_codebook_vectors[cb-1];
for (j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim) {
float rd = 0.0f;
int curbits = ff_aac_spectral_bits[cb-1][j];
int curidx = j;
#endif /* USE_REALLY_FULL_SEARCH */
if (IS_CODEBOOK_UNSIGNED(cb)) {
for (k = 0; k < dim; k++) {
float t = fabsf(in[i+k]);
float di;
if (vec[k] == 64.0f) { //FIXME: slow
//do not code with escape sequence small values
if (t < 39.0f*IQ) {
rd = INFINITY;
break;
}
if (t >= CLIPPED_ESCAPE) {
di = t - CLIPPED_ESCAPE;
curbits += 21;
} else {
int c = av_clip(quant(t, Q), 0, 8191);
di = t - c*cbrtf(c)*IQ;
curbits += av_log2(c)*2 - 4 + 1;
}
} else {
di = t - vec[k]*IQ;
}
if (vec[k] != 0.0f)
curbits++;
rd += di*di;
}
} else {
for (k = 0; k < dim; k++) {
float di = in[i+k] - vec[k]*IQ;
rd += di*di;
}
}
rd = rd * lambda + curbits;
if (rd < mincost) {
mincost = rd;
minidx = curidx;
minbits = curbits;
}
}
cost += mincost;
resbits += minbits;
if (cost >= uplim)
return uplim;
if (pb) {
put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]);
if (IS_CODEBOOK_UNSIGNED(cb))
for (j = 0; j < dim; j++)
if (ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f)
put_bits(pb, 1, in[i+j] < 0.0f);
if (cb == ESC_BT) {
for (j = 0; j < 2; j++) {
if (ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f) {
int coef = av_clip(quant(fabsf(in[i+j]), Q), 0, 8191);
int len = av_log2(coef);
put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
put_bits(pb, len, coef & ((1 << len) - 1));
}
}
}
}
}
if (bits)
*bits = resbits;
return cost;
}
static float quantize_band_cost(struct AACEncContext *s, const float *in,
const float *scaled, int size, int scale_idx,
int cb, const float lambda, const float uplim,
int *bits)
{
return quantize_and_encode_band_cost(s, NULL, in, scaled, size, scale_idx,
cb, lambda, uplim, bits);
}
static void quantize_and_encode_band(struct AACEncContext *s, PutBitContext *pb,
const float *in, int size, int scale_idx,
int cb, const float lambda)
{
quantize_and_encode_band_cost(s, pb, in, NULL, size, scale_idx, cb, lambda,
INFINITY, NULL);
}
static int testf(float float_x, long double long_double_x, /*float complex float_complex_x,*/ int int_x, long long_x)
{
int r = 0;
r += acosf(float_x);
r += acoshf(float_x);
r += asinf(float_x);
r += asinhf(float_x);
r += atan2f(float_x, float_x);
r += atanf(float_x);
r += atanhf(float_x);
/*r += cargf(float_complex_x); - will fight with complex numbers later */
r += cbrtf(float_x);
r += ceilf(float_x);
r += copysignf(float_x, float_x);
r += cosf(float_x);
r += coshf(float_x);
r += erfcf(float_x);
r += erff(float_x);
r += exp2f(float_x);
r += expf(float_x);
r += expm1f(float_x);
r += fabsf(float_x);
r += fdimf(float_x, float_x);
r += floorf(float_x);
r += fmaf(float_x, float_x, float_x);
r += fmaxf(float_x, float_x);
r += fminf(float_x, float_x);
r += fmodf(float_x, float_x);
r += frexpf(float_x, &int_x);
r += gammaf(float_x);
r += hypotf(float_x, float_x);
r += ilogbf(float_x);
r += ldexpf(float_x, int_x);
r += lgammaf(float_x);
r += llrintf(float_x);
r += llroundf(float_x);
r += log10f(float_x);
r += log1pf(float_x);
r += log2f(float_x);
r += logbf(float_x);
r += logf(float_x);
r += lrintf(float_x);
r += lroundf(float_x);
r += modff(float_x, &float_x);
r += nearbyintf(float_x);
r += nexttowardf(float_x, long_double_x);
r += powf(float_x, float_x);
r += remainderf(float_x, float_x);
r += remquof(float_x, float_x, &int_x);
r += rintf(float_x);
r += roundf(float_x);
#ifdef __UCLIBC_SUSV3_LEGACY__
r += scalbf(float_x, float_x);
#endif
r += scalblnf(float_x, long_x);
r += scalbnf(float_x, int_x);
r += significandf(float_x);
r += sinf(float_x);
r += sinhf(float_x);
r += sqrtf(float_x);
r += tanf(float_x);
r += tanhf(float_x);
r += tgammaf(float_x);
r += truncf(float_x);
return r;
}
//------------------------------------------------------------------------------
double Cmath::cbrt( double x )
{
return (double) cbrtf( (float) x );
}
void sph_simulation::load_settings(std::string fluid_file_name,
std::string parameters_file_name) {
int particles_inside_influence_radius = 0;
{
picojson::value fluid_params;
std::ifstream fluid_stream(fluid_file_name);
fluid_stream >> fluid_params;
parameters.fluid_density =
(float)(fluid_params.get<picojson::object>()["fluid_density"]
.get<double>());
parameters.dynamic_viscosity =
(float)(fluid_params.get<picojson::object>()["dynamic_viscosity"]
.get<double>());
parameters.restitution =
(float)(fluid_params.get<picojson::object>()["restitution"]
.get<double>());
if (parameters.restitution < 0 || parameters.restitution > 1) {
throw std::runtime_error("Restitution has an invalid value!");
}
parameters.K =
(float)(fluid_params.get<picojson::object>()["k"].get<double>());
parameters.surface_tension_threshold =
(float)(fluid_params
.get<picojson::object>()["surface_tension_threshold"]
.get<double>());
parameters.surface_tension =
(float)(fluid_params.get<picojson::object>()["surface_tension"]
.get<double>());
particles_inside_influence_radius =
(int)(fluid_params
.get<picojson::object>()["particles_inside_influence_radius"]
.get<double>());
}
{
picojson::value sim_params;
std::ifstream sim_stream(parameters_file_name);
sim_stream >> sim_params;
parameters.particles_count =
(unsigned int)(sim_params.get<picojson::object>()["particles_count"]
.get<double>());
if (parameters.particles_count % kPreferredWorkGroupSizeMultiple != 0) {
std::cout << std::endl
<< "\033[1;31m You should choose a number of particles that is "
"divisble by the preferred work group size.\033[0m";
std::cout << std::endl
<< "\033[1;31m Performances will be sub-optimal.\033[0m"
<< std::endl;
}
parameters.particle_mass =
(float)(sim_params.get<picojson::object>()["particle_mass"]
.get<double>());
parameters.simulation_time =
(float)(sim_params.get<picojson::object>()["simulation_time"]
.get<double>());
parameters.target_fps =
(float)(sim_params.get<picojson::object>()["target_fps"].get<double>());
parameters.simulation_scale =
(float)(sim_params.get<picojson::object>()["simulation_scale"]
.get<double>());
parameters.constant_acceleration.s[0] =
(float)(sim_params.get<picojson::object>()["constant_acceleration"]
.get<picojson::object>()["x"]
.get<double>());
parameters.constant_acceleration.s[1] =
(float)(sim_params.get<picojson::object>()["constant_acceleration"]
.get<picojson::object>()["y"]
.get<double>());
parameters.constant_acceleration.s[2] =
(float)(sim_params.get<picojson::object>()["constant_acceleration"]
.get<picojson::object>()["z"]
.get<double>());
write_intermediate_frames =
sim_params.get<picojson::object>()["write_all_frames"].get<bool>();
serialize = sim_params.get<picojson::object>()["serialize"].get<bool>();
}
parameters.total_mass = parameters.particles_count * parameters.particle_mass;
initial_volume = parameters.total_mass / parameters.fluid_density;
parameters.h = cbrtf(3.f * (particles_inside_influence_radius *
(initial_volume / parameters.particles_count)) /
(4.f * M_PI));
parameters.time_delta = 1.f / parameters.target_fps;
parameters.max_velocity = 0.8f * parameters.h / parameters.time_delta;
precomputed_terms.poly_6 = 315.f / (64.f * M_PI * pow(parameters.h, 9.f));
precomputed_terms.poly_6_gradient =
-945.f / (32.f * M_PI * pow(parameters.h, 9.f));
precomputed_terms.poly_6_laplacian =
-945.f / (32.f * M_PI * pow(parameters.h, 9.f));
precomputed_terms.spiky = -45.f / (M_PI * pow(parameters.h, 6.f));
precomputed_terms.viscosity = 45.f / (M_PI * pow(parameters.h, 6.f));
}
void ParticleSystem::update(){
int repielValue = 2;
float averfft=1.0f;
float fftMax = 0.0f;
float trigger;
// fft is uninitialized
if(!fftSize || !fft)
return ;
// compute fft average value and maximum
for(int i=0; i< fftSize; i++){ averfft+=fft[i];
if(fft[i]>fftMax)
fftMax = fft[i];
}
averfft/=fftSize;
int x, y;// -----
if (particles.size() > maxParticles & fftFlowFactor > 0){
fftFlowFactor-=0.1;
}
if (particles.size() < maxParticles ) {
fftFlowFactor+=0.1;
}
for(int k = fftSize; k >0 ; --k){
int i = ofRandom(0, fftSize);
float factor = fftFlowFactor * cbrtf(fft[i]*(float)(i+1)/20.0);
//factor = sqrtf(factor);
int numPart = factor * fft[i];
for(int j=0; j< numPart; j++){
float angle;
if (rotatingAngle){
angle = ofGetFrameNum()%100;
} else{
angle = 0;
}
if(particleRandomPos){
x = ofRandom(0, width);
y = ofRandom(0, height);
} else {
x = width/2;
y = height/2;
}
Particle p=Particle(x, y, particleSizeMin, particleSizeMax);
p.alpha = 0.0f;
p.id = i;
p.velocityMultiplier = particleVelocityMult;
p.angle = angle + i/(fftSize-1)*TWO_PI;
p.update();
particles.push_back(p);
}
}
// FIXME: rewrite this
// number of particles to add
/*int count = fftSize; // averfft * fftMax * 2000;
while ( count >= 0){
int i = ofRandom(0, fftSize-1);
// apply correct threshold depending on frequency
if ( i < fftSize/3 && ! averageTrigger ){
trigger = lowThresh;
} else if(i >= fftSize/3 && i < 2*(fftSize/3) && ! averageTrigger){
trigger = midThresh;
} else if( ! averageTrigger ){
trigger = highThresh;
} else { // then averfft trigger enabled
trigger = averfft;
}
count--;
// if fft value and not reaching max paticles number
if(fft[i] > trigger && particles.size() < maxParticles){
//cout<<fft[i]*10<<endl;
for(float cnt = 0.0f; cnt < fft[i] * 10.0f ; cnt+=0.1f){
//cout<<count<<endl;
//count--;
// add new particle
int x = width/2;
int y = height/2;
float angle;
if (rotatingAngle){
angle = ofGetFrameNum()%100;
} else{
angle = 0;
}
if(particleRandomPos){
x = ofRandom(0, width);
y = ofRandom(0, height);
} else {
}
Particle p=Particle(x, y, particleSizeMin, particleSizeMax);
p.alpha = 0.0f;
p.id = i;
p.velocityMultiplier = particleVelocityMult;
p.angle = angle + i/(fftSize-1)*TWO_PI;
p.update();
particles.push_back(p);
}
}
}*/
// for each particles
for(int j=0; j < particles.size(); j++){
particles[j].accel *= accelDamp;
particles[j].velocity *= velocityDamp;
//particles[j].addDamping();
if(! particleDeserveToLive(j)){
//cout<<"ERASING"<<endl;
particles.erase(particles.begin()+j);
continue;
//addParticle();
}
//---------------------------------------------------- Move Particles with a perlin noise
if(enableNoise){
float noiseAngle = noise->Get(particles[j].position.x, particles[j].position.y) * TWO_PI;
float noiseAngle2 = noise->Get(particles[j].position.x, particles[j].position.y) * TWO_PI;
ofxVec3f noiseForce = ofxVec3f(cos(noiseAngle), -sin(noiseAngle2), 0);
particles[j].accel += noiseMult * noiseForce;
}
int fftid=particles[j].id;
float sign=(ofRandom(0,1))?-1.0f:1.0f;
float fftAngle = sign * ((float )(fft[fftid]/fftMax)*PI + particles[j].angle);
// apply force
ofxVec2f fftForce;
if( averFFTSpeed)
fftForce = ofxVec3f(averfft * fftMult * cos(fftAngle), averfft * fftMult * sin(fftAngle), 0);
else
fftForce = ofxVec3f(sqrt(fft[fftid]) * averfft * fftMult * cos(fftAngle), averfft * sqrt(fft[fftid]) * fftMult * sin(fftAngle), 0);
particles[j].accel += fftForce;
// set particle color
msaColor color;
int h = (lroundf(particles[j].id * 360.0f / (float)fftSize));
color.setHSV(h, MIN(2.0f*fft[fftid], 1.0), MIN(2.0f* averfft, 1.0) );
particles[j].r = color.r;
particles[j].g = color.g;
particles[j].b = color.b;
particles[j].alpha = 0.5 + 0.5 * fft[fftid];
particles[j].update();
}
}
__host__ void single_precision_math_functions()
{
int iX;
float fX, fY;
acosf(1.0f);
acoshf(1.0f);
asinf(0.0f);
asinhf(0.0f);
atan2f(0.0f, 1.0f);
atanf(0.0f);
atanhf(0.0f);
cbrtf(0.0f);
ceilf(0.0f);
copysignf(1.0f, -2.0f);
cosf(0.0f);
coshf(0.0f);
//cospif(0.0f);
//cyl_bessel_i0f(0.0f);
//cyl_bessel_i1f(0.0f);
erfcf(0.0f);
//erfcinvf(2.0f);
//erfcxf(0.0f);
erff(0.0f);
//erfinvf(1.0f);
exp10f(0.0f);
exp2f(0.0f);
expf(0.0f);
expm1f(0.0f);
fabsf(1.0f);
fdimf(1.0f, 0.0f);
//fdividef(0.0f, 1.0f);
floorf(0.0f);
fmaf(1.0f, 2.0f, 3.0f);
fmaxf(0.0f, 0.0f);
fminf(0.0f, 0.0f);
fmodf(0.0f, 1.0f);
frexpf(0.0f, &iX);
hypotf(1.0f, 0.0f);
ilogbf(1.0f);
isfinite(0.0f);
isinf(0.0f);
isnan(0.0f);
///j0f(0.0f);
///j1f(0.0f);
///jnf(-1.0f, 1.0f);
ldexpf(0.0f, 0);
///lgammaf(1.0f);
///llrintf(0.0f);
///llroundf(0.0f);
log10f(1.0f);
log1pf(-1.0f);
log2f(1.0f);
logbf(1.0f);
logf(1.0f);
///lrintf(0.0f);
///lroundf(0.0f);
modff(0.0f, &fX);
///nanf("1");
nearbyintf(0.0f);
//nextafterf(0.0f);
//norm3df(1.0f, 0.0f, 0.0f);
//norm4df(1.0f, 0.0f, 0.0f, 0.0f);
//normcdff(0.0f);
//normcdfinvf(1.0f);
//fX = 1.0f; normf(1, &fX);
powf(1.0f, 0.0f);
//rcbrtf(1.0f);
remainderf(2.0f, 1.0f);
remquof(1.0f, 2.0f, &iX);
//rhypotf(0.0f, 1.0f);
///rintf(1.0f);
//rnorm3df(0.0f, 0.0f, 1.0f);
//rnorm4df(0.0f, 0.0f, 0.0f, 1.0f);
//fX = 1.0f; rnormf(1, &fX);
roundf(0.0f);
//rsqrtf(1.0f);
///scalblnf(0.0f, 1);
scalbnf(0.0f, 1);
signbit(1.0f);
sincosf(0.0f, &fX, &fY);
//sincospif(0.0f, &fX, &fY);
sinf(0.0f);
sinhf(0.0f);
//sinpif(0.0f);
sqrtf(0.0f);
tanf(0.0f);
tanhf(0.0f);
tgammaf(2.0f);
truncf(0.0f);
///y0f(1.0f);
///y1f(1.0f);
///ynf(1, 1.0f);
}
void
domathf (void)
{
#ifndef NO_FLOAT
float f1;
float f2;
int i1;
f1 = acosf(0.0);
fprintf( stdout, "acosf : %f\n", f1);
f1 = acoshf(0.0);
fprintf( stdout, "acoshf : %f\n", f1);
f1 = asinf(1.0);
fprintf( stdout, "asinf : %f\n", f1);
f1 = asinhf(1.0);
fprintf( stdout, "asinhf : %f\n", f1);
f1 = atanf(M_PI_4);
fprintf( stdout, "atanf : %f\n", f1);
f1 = atan2f(2.3, 2.3);
fprintf( stdout, "atan2f : %f\n", f1);
f1 = atanhf(1.0);
fprintf( stdout, "atanhf : %f\n", f1);
f1 = cbrtf(27.0);
fprintf( stdout, "cbrtf : %f\n", f1);
f1 = ceilf(3.5);
fprintf( stdout, "ceilf : %f\n", f1);
f1 = copysignf(3.5, -2.5);
fprintf( stdout, "copysignf : %f\n", f1);
f1 = cosf(M_PI_2);
fprintf( stdout, "cosf : %f\n", f1);
f1 = coshf(M_PI_2);
fprintf( stdout, "coshf : %f\n", f1);
f1 = erff(42.0);
fprintf( stdout, "erff : %f\n", f1);
f1 = erfcf(42.0);
fprintf( stdout, "erfcf : %f\n", f1);
f1 = expf(0.42);
fprintf( stdout, "expf : %f\n", f1);
f1 = exp2f(0.42);
fprintf( stdout, "exp2f : %f\n", f1);
f1 = expm1f(0.00042);
fprintf( stdout, "expm1f : %f\n", f1);
f1 = fabsf(-1.123);
fprintf( stdout, "fabsf : %f\n", f1);
f1 = fdimf(1.123, 2.123);
fprintf( stdout, "fdimf : %f\n", f1);
f1 = floorf(0.5);
fprintf( stdout, "floorf : %f\n", f1);
f1 = floorf(-0.5);
fprintf( stdout, "floorf : %f\n", f1);
f1 = fmaf(2.1, 2.2, 3.01);
fprintf( stdout, "fmaf : %f\n", f1);
f1 = fmaxf(-0.42, 0.42);
fprintf( stdout, "fmaxf : %f\n", f1);
f1 = fminf(-0.42, 0.42);
fprintf( stdout, "fminf : %f\n", f1);
f1 = fmodf(42.0, 3.0);
fprintf( stdout, "fmodf : %f\n", f1);
/* no type-specific variant */
i1 = fpclassify(1.0);
fprintf( stdout, "fpclassify : %d\n", i1);
f1 = frexpf(42.0, &i1);
fprintf( stdout, "frexpf : %f\n", f1);
f1 = hypotf(42.0, 42.0);
fprintf( stdout, "hypotf : %f\n", f1);
i1 = ilogbf(42.0);
fprintf( stdout, "ilogbf : %d\n", i1);
/* no type-specific variant */
i1 = isfinite(3.0);
fprintf( stdout, "isfinite : %d\n", i1);
/* no type-specific variant */
i1 = isgreater(3.0, 3.1);
fprintf( stdout, "isgreater : %d\n", i1);
/* no type-specific variant */
i1 = isgreaterequal(3.0, 3.1);
fprintf( stdout, "isgreaterequal : %d\n", i1);
/* no type-specific variant */
i1 = isinf(3.0);
fprintf( stdout, "isinf : %d\n", i1);
/* no type-specific variant */
i1 = isless(3.0, 3.1);
fprintf( stdout, "isless : %d\n", i1);
/* no type-specific variant */
i1 = islessequal(3.0, 3.1);
fprintf( stdout, "islessequal : %d\n", i1);
/* no type-specific variant */
i1 = islessgreater(3.0, 3.1);
fprintf( stdout, "islessgreater : %d\n", i1);
/* no type-specific variant */
i1 = isnan(0.0);
fprintf( stdout, "isnan : %d\n", i1);
/* no type-specific variant */
i1 = isnormal(3.0);
fprintf( stdout, "isnormal : %d\n", i1);
/* no type-specific variant */
f1 = isunordered(1.0, 2.0);
fprintf( stdout, "isunordered : %d\n", i1);
f1 = j0f(1.2);
fprintf( stdout, "j0f : %f\n", f1);
f1 = j1f(1.2);
fprintf( stdout, "j1f : %f\n", f1);
f1 = jnf(2,1.2);
fprintf( stdout, "jnf : %f\n", f1);
f1 = ldexpf(1.2,3);
fprintf( stdout, "ldexpf : %f\n", f1);
f1 = lgammaf(42.0);
fprintf( stdout, "lgammaf : %f\n", f1);
f1 = llrintf(-0.5);
fprintf( stdout, "llrintf : %f\n", f1);
f1 = llrintf(0.5);
fprintf( stdout, "llrintf : %f\n", f1);
f1 = llroundf(-0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = llroundf(0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = logf(42.0);
fprintf( stdout, "logf : %f\n", f1);
f1 = log10f(42.0);
fprintf( stdout, "log10f : %f\n", f1);
f1 = log1pf(42.0);
fprintf( stdout, "log1pf : %f\n", f1);
f1 = log2f(42.0);
fprintf( stdout, "log2f : %f\n", f1);
f1 = logbf(42.0);
fprintf( stdout, "logbf : %f\n", f1);
f1 = lrintf(-0.5);
fprintf( stdout, "lrintf : %f\n", f1);
f1 = lrintf(0.5);
fprintf( stdout, "lrintf : %f\n", f1);
f1 = lroundf(-0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = lroundf(0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = modff(42.0,&f2);
fprintf( stdout, "lmodff : %f\n", f1);
f1 = nanf("");
fprintf( stdout, "nanf : %f\n", f1);
f1 = nearbyintf(1.5);
fprintf( stdout, "nearbyintf : %f\n", f1);
f1 = nextafterf(1.5,2.0);
fprintf( stdout, "nextafterf : %f\n", f1);
f1 = powf(3.01, 2.0);
fprintf( stdout, "powf : %f\n", f1);
f1 = remainderf(3.01,2.0);
fprintf( stdout, "remainderf : %f\n", f1);
f1 = remquof(29.0,3.0,&i1);
fprintf( stdout, "remquof : %f\n", f1);
f1 = rintf(0.5);
fprintf( stdout, "rintf : %f\n", f1);
f1 = rintf(-0.5);
fprintf( stdout, "rintf : %f\n", f1);
f1 = roundf(0.5);
fprintf( stdout, "roundf : %f\n", f1);
f1 = roundf(-0.5);
fprintf( stdout, "roundf : %f\n", f1);
f1 = scalblnf(1.2,3);
fprintf( stdout, "scalblnf : %f\n", f1);
f1 = scalbnf(1.2,3);
fprintf( stdout, "scalbnf : %f\n", f1);
/* no type-specific variant */
i1 = signbit(1.0);
fprintf( stdout, "signbit : %i\n", i1);
f1 = sinf(M_PI_4);
fprintf( stdout, "sinf : %f\n", f1);
f1 = sinhf(M_PI_4);
fprintf( stdout, "sinhf : %f\n", f1);
f1 = sqrtf(9.0);
fprintf( stdout, "sqrtf : %f\n", f1);
f1 = tanf(M_PI_4);
fprintf( stdout, "tanf : %f\n", f1);
f1 = tanhf(M_PI_4);
fprintf( stdout, "tanhf : %f\n", f1);
f1 = tgammaf(2.1);
fprintf( stdout, "tgammaf : %f\n", f1);
f1 = truncf(3.5);
fprintf( stdout, "truncf : %f\n", f1);
f1 = y0f(1.2);
fprintf( stdout, "y0f : %f\n", f1);
f1 = y1f(1.2);
fprintf( stdout, "y1f : %f\n", f1);
f1 = ynf(3,1.2);
fprintf( stdout, "ynf : %f\n", f1);
#endif
}
void test (float f, double d, long double ld)
{
if (sqrt (0.0) != 0.0)
link_error ();
if (sqrt (1.0) != 1.0)
link_error ();
if (cbrt (0.0) != 0.0)
link_error ();
if (cbrt (1.0) != 1.0)
link_error ();
if (cbrt (-1.0) != -1.0)
link_error ();
if (exp (0.0) != 1.0)
link_error ();
if (exp (1.0) <= 2.71 || exp (1.0) >= 2.72)
link_error ();
if (log (1.0) != 0.0)
link_error ();
if (sin (0.0) != 0.0)
link_error ();
if (cos (0.0) != 1.0)
link_error ();
if (tan (0.0) != 0.0)
link_error ();
if (atan (0.0) != 0.0)
link_error ();
if (4.0*atan (1.0) <= 3.14 || 4.0*atan (1.0) >= 3.15)
link_error ();
if (pow (d, 0.0) != 1.0)
link_error ();
if (pow (1.0, d) != 1.0)
link_error ();
if (sqrtf (0.0F) != 0.0F)
link_error ();
if (sqrtf (1.0F) != 1.0F)
link_error ();
if (cbrtf (0.0F) != 0.0F)
link_error ();
if (cbrtf (1.0F) != 1.0F)
link_error ();
if (cbrtf (-1.0F) != -1.0F)
link_error ();
if (expf (0.0F) != 1.0F)
link_error ();
if (expf (1.0F) <= 2.71F || expf (1.0F) >= 2.72F)
link_error ();
if (logf (1.0F) != 0.0F)
link_error ();
if (sinf (0.0F) != 0.0F)
link_error ();
if (cosf (0.0F) != 1.0F)
link_error ();
if (tanf (0.0F) != 0.0F)
link_error ();
if (atanf (0.0F) != 0.0F)
link_error ();
if (4.0F*atanf (1.0F) <= 3.14F || 4.0F*atanf (1.0F) >= 3.15F)
link_error ();
if (powf (f, 0.0F) != 1.0F)
link_error ();
if (powf (1.0F, f) != 1.0F)
link_error ();
if (sqrtl (0.0L) != 0.0L)
link_error ();
if (sqrtl (1.0L) != 1.0L)
link_error ();
if (cbrtl (0.0L) != 0.0L)
link_error ();
if (cbrtl (1.0L) != 1.0L)
link_error ();
if (cbrtl (-1.0L) != -1.0L)
link_error ();
if (expl (0.0L) != 1.0L)
link_error ();
if (expl (1.0L) <= 2.71L || expl (1.0L) >= 2.72L)
link_error ();
if (logl (1.0L) != 0.0L)
link_error ();
if (sinl (0.0L) != 0.0L)
link_error ();
if (cosl (0.0L) != 1.0L)
link_error ();
if (tanl (0.0L) != 0.0L)
link_error ();
if (atanl (0.0) != 0.0L)
link_error ();
if (4.0L*atanl (1.0L) <= 3.14L || 4.0L*atanl (1.0L) >= 3.15L)
link_error ();
if (powl (ld, 0.0L) != 1.0L)
link_error ();
if (powl (1.0L, ld) != 1.0L)
link_error ();
}
void mri_metrics_pp( MRI_IMAGE *imp , MRI_IMAGE *imq , float *met , byte *mmm )
{
int nvox , nmmm=0 ;
int *rst , *pst , *qst ;
byte *par, *qar ;
float qj,rij , rat,tmp,lrr,rm1,rp1 , fac ;
float esum,tsum,hsum , jsum,dsum,xsum , qsum,asum ;
register int ii,jj,kk ;
MRI_IMAGE *imqq, *impp ;
if( imp == NULL || imq == NULL ) return ;
if( met == NULL || imp->nvox != imq->nvox ) return ;
nvox = imp->nvox ;
impp = (imp->kind==MRI_byte) ? imp : mri_to_byte(imp) ;
imqq = (imq->kind==MRI_byte) ? imq : mri_to_byte(imq) ;
par = MRI_BYTE_PTR(impp) ;
qar = MRI_BYTE_PTR(imqq) ;
pst = (int *)calloc(256 ,sizeof(int)) ;
qst = (int *)calloc(256 ,sizeof(int)) ;
rst = (int *)calloc(256*256,sizeof(int)) ;
if( mmm != NULL ){
for( kk=0 ; kk < nvox ; kk++ ) if( mmm[kk] ) nmmm++ ;
fac = 1.0f / nmmm ;
for( kk=0 ; kk < nvox ; kk++ ){
if( mmm[kk] == 0 ) continue ;
ii = par[kk] ; jj = qar[kk] ;
pst[ii]++ ; qst[jj]++ ; rst[ii+256*jj]++ ;
}
} else {
fac = 1.0f / nvox ;
for( kk=0 ; kk < nvox ; kk++ ){
ii = par[kk] ; jj = qar[kk] ;
pst[ii]++ ; qst[jj]++ ; rst[ii+256*jj]++ ;
}
}
esum = tsum = hsum = 0.0f ;
jsum = dsum = xsum = 0.0f ;
qsum = asum = 0.0f ;
for( jj=0 ; jj < 256 ; jj++ ){
qj = (float)qst[jj] ;
if( qj > 0.0f ){
kk = 256*jj ; qj *= fac ;
for( ii=0 ; ii < 256 ; ii++ ){
rij = (float)rst[ii+kk] ;
if( rij > 0.0f ){
rat = (qj *(float)pst[ii]) / rij ;
lrr = logf(rat) ;
esum += rij * lrr ;
tmp = 1.0f/sqrtf(rat)-1.0f ; hsum += rij * tmp ;
tmp = 1.0f/cbrtf(rat)-1.0f ; qsum += rij * tmp ;
tmp = lrr*lrr*lrr ; asum += rij * tmp ;
}
}
}
}
free((void*)rst); free((void*)qst); free((void*)pst);
if( impp != imp ) mri_free(impp);
if( imqq != imq ) mri_free(imqq);
met[0] = -fac * esum * 0.5f ;
met[1] = fac * hsum * 0.5f ;
met[2] = fac * qsum ;
return ;
}
TEST(math, cbrtf) {
ASSERT_FLOAT_EQ(3.0f, cbrtf(27.0f));
}
