coffee/fast_float
1
0
Fork 0
mirror of https://github.com/fastfloat/fast_float.git synced 2026-07-30 16:26:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

EXP-053: 4-digit SWAR follow-up in loop_parse_if_eight_digits, GCC path (ffc EXP-001)

Browse Source 
After the 8-digit block loop, consume a remaining 4-7 digit run in one SWAR step
(reusing fast_float's existing read4_to_u32 / is_made_of_four_digits_fast /
parse_four_digits_unrolled) instead of byte-by-byte. GCC path only: on Clang the
follow-up's presence bloated the 2x-unroll codegen and regressed random -6.2%.

ARM Graviton4 (canonical fast_float MB/s vs EXP-052):
  GCC:   canada +2.6% (948.1 from 924.0, i/f 248.7->229.7), random/mesh flat
  Clang: unchanged (EXP-052 path preserved)
Correctness: 14/14 pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
fcostaoliveira 2026-06-01 01:04:30 +01:00
parent ee84946702
commit a30c1f3d3f
1 changed files with 11 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
include/fast_float/ascii_number.h
View File

@ -294,6 +294,17 @@ loop_parse_if_eight_digits(char const *&p, char const *const pend,
p)); // in rare cases, this will overflow, but that's ok
p += 8;
}
// 4-digit SWAR follow-up (ported from ffc EXP-001): consume a remaining 4-7
// digit run in one step rather than byte-by-byte. GCC path only — on Clang
// the follow-up's presence bloated the 2x-unroll codegen and regressed random.
if ((pend - p) >= 4) {
uint32_t const val4 = read4_to_u32(p);
if (is_made_of_four_digits_fast(val4)) {
i = i * 10000 +
parse_four_digits_unrolled(val4); // in rare cases overflows, that's ok
p += 4;
}
}
#endif
}