coffee/fast_float
1
0
Fork 0
mirror of https://github.com/fastfloat/fast_float.git synced 2025-12-06 16:56:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Automatically detect SSE2

Browse Source
This commit is contained in:
Maya Warrier 2023-05-02 13:05:57 -04:00
parent c811b027ea
commit 4cb09b5f59
2 changed files with 10 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
include/fast_float/ascii_number.h
View File

@ -73,7 +73,7 @@ FASTFLOAT_SIMD_DISABLE_WARNINGS
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
// todo: with AVX512BW + AVX512VL, can use cvtepi16_epi8 instead of masking + pack
const char* const p = reinterpret_cast<const char*>(chars);
__m128i i1 = _mm_and_si128(_mm_loadu_si64(p), masks);
__m128i i2 = _mm_and_si128(_mm_loadu_si64(p + 8), masks);
@ -150,8 +150,8 @@ bool parse_if_eight_digits_unrolled(const char* chars, uint64_t& i) noexcept {
}
// Call this if chars might not be 8 digits.
// Using this (instead of is_made_of_eight_digits_fast() and parse_eight_digits_unrolled())
// ensures we don't load SIMD registers twice.
// Using this style (instead of is_made_of_eight_digits_fast() then parse_eight_digits_unrolled())
// ensures we don't load SIMD registers twice if we don't have to.
//
// Benchmark:
// https://quick-bench.com/q/Bbn0B4WmZsdgS3qDZWpggAY-jgs

8
include/fast_float/float_common.h
View File

@ -78,8 +78,14 @@
#endif
#endif
#if defined(__SSE2__) || \
(defined(FASTFLOAT_VISUAL_STUDIO) && \
(defined(_M_AMD64) || defined(_M_X64) || (defined(_M_IX86_FP) && _M_IX86_FP == 2)))
#define FASTFLOAT_SSE2 1
#endif
#ifdef FASTFLOAT_SSE2
#define FASTFLOAT_HAS_SIMD (1)
#define FASTFLOAT_HAS_SIMD 1
#endif
#if defined(__GNUC__)