this-city/engine/thirdparty/meshoptimizer/quantization.cpp
2026-07-10 17:04:34 +02:00

127 lines
3.4 KiB
C++

// This file is part of meshoptimizer library; see meshoptimizer.h for version/license details
#include "meshoptimizer.h"
#include <assert.h>
#include <float.h>
#include <math.h>
union FloatBits
{
float f;
unsigned int ui;
};
unsigned short meshopt_quantizeHalf(float v)
{
FloatBits u = {v};
unsigned int ui = u.ui;
int s = (ui >> 16) & 0x8000;
int em = ui & 0x7fffffff;
// bias exponent and round to nearest; 112 is relative exponent bias (127-15)
int h = (em - (112 << 23) + (1 << 12)) >> 13;
// underflow: flush to zero; 113 encodes exponent -14
h = (em < (113 << 23)) ? 0 : h;
// overflow: infinity; 143 encodes exponent 16
h = (em >= (143 << 23)) ? 0x7c00 : h;
// NaN; note that we convert all types of NaN to qNaN
h = (em > (255 << 23)) ? 0x7e00 : h;
return (unsigned short)(s | h);
}
float meshopt_quantizeFloat(float v, int N)
{
assert(N >= 0 && N <= 23);
FloatBits u = {v};
unsigned int ui = u.ui;
const int mask = (1 << (23 - N)) - 1;
const int round = (1 << (23 - N)) >> 1;
int e = ui & 0x7f800000;
unsigned int rui = (ui + round) & ~mask;
// round all numbers except inf/nan; this is important to make sure nan doesn't overflow into -0
ui = e == 0x7f800000 ? ui : rui;
// flush denormals to zero
ui = e == 0 ? 0 : ui;
u.ui = ui;
return u.f;
}
float meshopt_dequantizeHalf(unsigned short h)
{
unsigned int s = unsigned(h & 0x8000) << 16;
int em = h & 0x7fff;
// bias exponent and pad mantissa with 0; 112 is relative exponent bias (127-15)
int r = (em + (112 << 10)) << 13;
// denormal: flush to zero
r = (em < (1 << 10)) ? 0 : r;
// infinity/NaN; note that we preserve NaN payload as a byproduct of unifying inf/nan cases
// 112 is an exponent bias fixup; since we already applied it once, applying it twice converts 31 to 255
r += (em >= (31 << 10)) ? (112 << 23) : 0;
FloatBits u;
u.ui = s | r;
return u.f;
}
int meshopt_computePositionExponent(const float* minv, const float* maxv, int min_exp, int max_bits)
{
assert(min_exp >= -126);
assert(max_bits >= 2 && max_bits <= 24);
int exp = min_exp;
// compute max absolute component to ensure that individual endpoints fit on a 24-bit signed grid
float maxc = 0.f;
for (int k = 0; k < 3; ++k)
{
maxc = maxc < fabsf(minv[k]) ? fabsf(minv[k]) : maxc;
maxc = maxc < fabsf(maxv[k]) ? fabsf(maxv[k]) : maxc;
}
int maxc_exp = 0;
float maxc_fr = frexpf(maxc, &maxc_exp);
// maxc is representable as 2^(maxc_exp-24) * 24-bit *unsigned* integer
// we have to use a 24-bit *signed* grid, so we have to chop off the last bit
// however, rounding in the corner case (mantissa is 1.111...) may increase the exponent by 1 so we need to offset it
int maxc_off = 23 - (maxc_fr >= 1.f - FLT_EPSILON / 2);
exp = exp < maxc_exp - maxc_off ? maxc_exp - maxc_off : exp;
// compute effective range with conservative rounding, to allow for some ambiguity in the caller's rounding direction
float scale = ldexpf(1.f, -exp);
float range = 0.f;
for (int k = 0; k < 3; ++k)
{
float a = floorf(minv[k] * scale);
float v = ceilf(maxv[k] * scale);
range = range < v - a ? v - a : range;
}
// range must be representable as max_bits unsigned integer
int range_exp = 0;
float range_fr = frexpf(range, &range_exp);
exp += (range_exp > max_bits) ? range_exp - max_bits : 0;
// correct range if we are at the rounding boundary that would push us to overflow max_bits
exp += range_exp > max_bits && range_fr >= 1.f - 1.f / float((1 << max_bits) - 1);
return exp;
}