mirror of
https://github.com/nestriness/nestri.git
synced 2026-11-15 11:42:28 +02:00
cf46bf4d12
At this point, why not.. Calling it `nespyro` since it's slightly modified, uses slang shader language instead, along other things.. --------- Co-authored-by: DatCaptainHorse <DatCaptainHorse@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
42 lines
1.6 KiB
Plaintext
42 lines
1.6 KiB
Plaintext
// The CDF 9/7 lifting constants and the shared tile the forward and inverse
|
|
// transforms work in.
|
|
//
|
|
// Ported from PyroWave's dwt_common.h (Copyright (c) 2025 Hans-Kristian
|
|
// Arntzen, MIT) at precision level 1, upstream's default: the lifting maths is
|
|
// FP32 and the shared tile is FP16. `FP16` says how the tile is stored: as
|
|
// native half2 where the device has shaderFloat16, packed into a uint through
|
|
// packHalf2x16 where it does not. The values held are the same either way.
|
|
|
|
#pragma once
|
|
|
|
#include "common.slang"
|
|
|
|
static const int APRON = 4;
|
|
static const int APRON_HALF = APRON / 2;
|
|
static const int BLOCK_SIZE = 32;
|
|
static const int BLOCK_SIZE_HALF = BLOCK_SIZE >> 1;
|
|
|
|
static const float ALPHA = -1.586134342059924;
|
|
static const float BETA = -0.052980118572961;
|
|
static const float GAMMA = 0.882911075530934;
|
|
static const float DELTA = 0.443506852043971;
|
|
static const float K = 1.230174104914001;
|
|
static const float inv_K = 1.0 / 1.230174104914001;
|
|
|
|
// Two image rows packed side by side in each element, so one lane lifts two
|
|
// lines at once.
|
|
#if FP16
|
|
groupshared half2 shared_block[(BLOCK_SIZE + 2 * APRON) / 2][(BLOCK_SIZE + 2 * APRON) + 1];
|
|
float2 load_shared(uint y, uint x) { return float2(shared_block[y][x]); }
|
|
void store_shared(uint y, uint x, float2 v) { shared_block[y][x] = half2(v); }
|
|
#else
|
|
groupshared uint shared_block[(BLOCK_SIZE + 2 * APRON) / 2][(BLOCK_SIZE + 2 * APRON) + 1];
|
|
float2 load_shared(uint y, uint x) { return unpackHalf2x16(shared_block[y][x]); }
|
|
void store_shared(uint y, uint x, float2 v) { shared_block[y][x] = packHalf2x16(v); }
|
|
#endif
|
|
|
|
bool2 band(bool2 a, bool2 b)
|
|
{
|
|
return bool2(a.x && b.x, a.y && b.y);
|
|
}
|