jirayu_sri/stockfish
1
0
Fork 0
mirror of https://github.com/official-stockfish/Stockfish.git synced 2026-07-22 12:47:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove an instruction from find_nnz()

Browse Source 
closes https://github.com/official-stockfish/Stockfish/pull/6212

No functional change
This commit is contained in:
mstembera
2025-08-09 08:54:47 +02:00
committed by Joost VandeVondele
parent 125a0cfe7b
commit 5464f7b114
1 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/nnue/layers/affine_transform_sparse_input.h
+5 -4
View File
@@ -88,8 +88,9 @@ void find_nnz(const std::int32_t* RESTRICT input,
constexpr IndexType SimdWidthOut = 32; // 512 bits / 16 bits constexpr IndexType SimdWidthOut = 32; // 512 bits / 16 bits
constexpr IndexType NumChunks = InputDimensions / SimdWidthOut; constexpr IndexType NumChunks = InputDimensions / SimdWidthOut;
const __m512i increment = _mm512_set1_epi16(SimdWidthOut); const __m512i increment = _mm512_set1_epi16(SimdWidthOut);
__m512i base = _mm512_set_epi16(31, 30, 29, 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, __m512i base = _mm512_set_epi16( // Same permute order as _mm512_packus_epi32()
15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); 31, 30, 29, 28, 15, 14, 13, 12, 27, 26, 25, 24, 11, 10, 9, 8, 23, 22, 21, 20, 7, 6, 5, 4, 19,
18, 17, 16, 3, 2, 1, 0);
IndexType count = 0; IndexType count = 0;
for (IndexType i = 0; i < NumChunks; ++i) for (IndexType i = 0; i < NumChunks; ++i)
@@ -98,8 +99,8 @@ void find_nnz(const std::int32_t* RESTRICT input,
const __m512i inputV1 = _mm512_load_si512(input + i * 2 * SimdWidthIn + SimdWidthIn); const __m512i inputV1 = _mm512_load_si512(input + i * 2 * SimdWidthIn + SimdWidthIn);
// Get a bitmask and gather non zero indices // Get a bitmask and gather non zero indices
const __mmask32 nnzMask = _mm512_kunpackw(_mm512_test_epi32_mask(inputV1, inputV1), const __m512i inputV01 = _mm512_packus_epi32(inputV0, inputV1);
_mm512_test_epi32_mask(inputV0, inputV0)); const __mmask32 nnzMask = _mm512_test_epi16_mask(inputV01, inputV01);
// Avoid _mm512_mask_compressstoreu_epi16() as it's 256 uOps on Zen4 // Avoid _mm512_mask_compressstoreu_epi16() as it's 256 uOps on Zen4
__m512i nnz = _mm512_maskz_compress_epi16(nnzMask, base); __m512i nnz = _mm512_maskz_compress_epi16(nnzMask, base);