coffee/fast_float
1
0
Fork 0
mirror of https://github.com/fastfloat/fast_float.git synced 2026-02-06 17:59:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Avoid redundant load in SSE2 code

Browse Source
This commit is contained in:
Maya Warrier 2023-05-07 17:38:32 -04:00
parent 4cb09b5f59
commit 53b065f38d
1 changed files with 17 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
include/fast_float/ascii_number.h
View File

@ -67,23 +67,28 @@ uint64_t read8_to_u64(const UC *chars) {
#ifdef FASTFLOAT_SSE2 #ifdef FASTFLOAT_SSE2
fastfloat_really_inline fastfloat_really_inline
uint64_t simd_read8_to_u64(const char16_t* chars) { uint64_t simd_read8_to_u64(const __m128i data) {
FASTFLOAT_SIMD_DISABLE_WARNINGS FASTFLOAT_SIMD_DISABLE_WARNINGS
static const char16_t kmasks[] = { 0xff, 0xff, 0xff, 0xff }; static const char16_t kmasks[] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
const __m128i masks = _mm_loadu_si128(reinterpret_cast<const __m128i*>(kmasks)); const __m128i masks = _mm_loadu_si128(reinterpret_cast<const __m128i*>(kmasks));
// pipeline 4 and 4 chars at the same time (since loadu_si64 has high latency)
// todo: with AVX512BW + AVX512VL, can use cvtepi16_epi8 instead of masking + pack // todo: with AVX512BW + AVX512VL, can use cvtepi16_epi8 instead of masking + pack
const char* const p = reinterpret_cast<const char*>(chars); __m128i masked = _mm_and_si128(data, masks);
__m128i i1 = _mm_and_si128(_mm_loadu_si64(p), masks); __m128i packed = _mm_packus_epi16(masked, masked);
__m128i i2 = _mm_and_si128(_mm_loadu_si64(p + 8), masks);
__m128i packed = _mm_packus_epi16(i1, i2);
uint64_t val; uint64_t val;
_mm_storeu_si64(&val, _mm_shuffle_epi32(packed, 0x8)); _mm_storeu_si64(&val, packed);
return val; return val;
FASTFLOAT_SIMD_RESTORE_WARNINGS FASTFLOAT_SIMD_RESTORE_WARNINGS
} }
fastfloat_really_inline
uint64_t simd_read8_to_u64(const char16_t* chars) {
FASTFLOAT_SIMD_DISABLE_WARNINGS
return simd_read8_to_u64(_mm_loadu_si128(reinterpret_cast<const __m128i*>(chars)));
FASTFLOAT_SIMD_RESTORE_WARNINGS
}
#endif #endif
// dummy for compile // dummy for compile
@ -142,11 +147,11 @@ fastfloat_really_inline constexpr bool is_made_of_eight_digits_fast(uint64_t val
fastfloat_really_inline FASTFLOAT_CONSTEXPR20 fastfloat_really_inline FASTFLOAT_CONSTEXPR20
bool parse_if_eight_digits_unrolled(const char* chars, uint64_t& i) noexcept { bool parse_if_eight_digits_unrolled(const char* chars, uint64_t& i) noexcept {
const bool is_digits = is_made_of_eight_digits_fast(read8_to_u64(chars)); if (is_made_of_eight_digits_fast(read8_to_u64(chars))) {
if (is_digits) {
i = i * 100000000 + parse_eight_digits_unrolled(read8_to_u64(chars)); i = i * 100000000 + parse_eight_digits_unrolled(read8_to_u64(chars));
return true;
} }
return is_digits; return false;
} }
// Call this if chars might not be 8 digits. // Call this if chars might not be 8 digits.
@ -171,8 +176,7 @@ FASTFLOAT_SIMD_DISABLE_WARNINGS
const __m128i t1 = _mm_cmpgt_epi16(t0, _mm_set1_epi16(-119)); const __m128i t1 = _mm_cmpgt_epi16(t0, _mm_set1_epi16(-119));
if (_mm_movemask_epi8(t1) == 0) { if (_mm_movemask_epi8(t1) == 0) {
uint64_t digits = simd_read8_to_u64(chars); i = i * 100000000 + parse_eight_digits_unrolled(simd_read8_to_u64(data));
i = i * 100000000 + parse_eight_digits_unrolled(digits);
return true; return true;
} }
else return false; else return false;