coffee/fast_float
1
0
Fork 0
mirror of https://github.com/fastfloat/fast_float.git synced 2026-06-15 08:26:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

Unroll the integer-part digit scan (straight-line for the common 1-5 digit case)

Browse Source 
parse_number_string scans the integer part one byte at a time in a while loop,
while the fraction already uses the 8-digit SWAR loop. Most integer parts are
1-5 digits, so the loop back-edge dominates. Peel the first five iterations into
nested ifs, falling through to the original while for longer runs. Semantics are
identical (i = 10*i + digit, advancing p); no behavior change.

AWS m8g.metal-24xl (Graviton4), -O3 -march=native, simple_fastfloat_benchmark,
from_chars->double. base vs patch measured back-to-back, mean of 2 runs:
  canada: gcc +3.1%, clang +2.8%
  mesh:   gcc +5.4%, clang +5.1%
  random: ~flat (1-digit integer part)
No regression; gcc and clang agree.

Alternatives benchmarked and rejected: reusing loop_parse_if_eight_digits for the
integer part regressed 5-8% (integer parts are too short for 8-digit SWAR setup);
a counted for(k<5) loop matched on gcc but clang optimized it worse (canada -0.9%).
The explicit peel is the only form solidly positive on both compilers.
This commit is contained in:
fcostaoliveira 2026-06-01 00:48:45 +01:00
parent 7790aa6231
commit b64d014e2f
1 changed files with 29 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
include/fast_float/ascii_number.h
View File

@ -354,14 +354,37 @@ parse_number_string(UC const *p, UC const *pend,
uint64_t i = 0; // an unsigned int avoids signed overflows (which are bad) uint64_t i = 0; // an unsigned int avoids signed overflows (which are bad)
// Straight-line unroll of the integer-part scan: most integer parts are
// 1-5 digits, so peeling the first iterations eliminates the loop back-edge
// for the common case. Semantics are identical to the original `while` loop:
// i = 10*i + digit, advancing p.
if ((p != pend) && is_integer(*p)) {
i = uint64_t(*p - UC('0'));
++p;
if ((p != pend) && is_integer(*p)) {
i = 10 * i + uint64_t(*p - UC('0'));
++p;
if ((p != pend) && is_integer(*p)) {
i = 10 * i + uint64_t(*p - UC('0'));
++p;
if ((p != pend) && is_integer(*p)) {
i = 10 * i + uint64_t(*p - UC('0'));
++p;
if ((p != pend) && is_integer(*p)) {
i = 10 * i + uint64_t(*p - UC('0'));
++p;
while ((p != pend) && is_integer(*p)) { while ((p != pend) && is_integer(*p)) {
// a multiplication by 10 is cheaper than an arbitrary integer // a multiplication by 10 is cheaper than an arbitrary integer
// multiplication // multiplication
i = 10 * i + i = 10 * i +
uint64_t(*p - uint64_t(*p - UC('0')); // might overflow, handled later
UC('0')); // might overflow, we will handle the overflow later
++p; ++p;
} }
}
}
}
}
}
UC const *const end_of_integer_part = p; UC const *const end_of_integer_part = p;
int64_t digit_count = int64_t(end_of_integer_part - start_digits); int64_t digit_count = int64_t(end_of_integer_part - start_digits);
answer.integer = span<UC const>(start_digits, size_t(digit_count)); answer.integer = span<UC const>(start_digits, size_t(digit_count));