mirrors/kitty
1
0
Fork 0
mirror of https://github.com/kovidgoyal/kitty synced 2026-06-12 11:39:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

Turns out the simde implementation of movemask is not slow enough to compensate for the speed bump from 256 bit

Browse Source
This commit is contained in:
Kovid Goyal
2024-01-27 12:09:16 +05:30
parent 0bd47d8457
commit 8aa1b112b8
1 changed files with 7 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
kitty/simd-string-impl.h
View File

@@ -83,12 +83,6 @@ typedef int32_t find_mask_t;
static inline int
FUNC(is_zero)(const integer_t a) { return simde_mm_testz_si128(a, a); }
static inline find_mask_t
mask_for_find(const integer_t a) { return movemask_epi8(a); }
static inline unsigned
bytes_to_first_match(const find_mask_t m) { return __builtin_ctz(m); }
#else
#define set1_epi8(x) simde_mm256_set1_epi8((char)(x))
@@ -116,12 +110,6 @@ bytes_to_first_match(const find_mask_t m) { return __builtin_ctz(m); }
static inline int
FUNC(is_zero)(const integer_t a) { return simde_mm256_testz_si256(a, a); }
static inline find_mask_t
mask_for_find(const integer_t a) { return movemask_epi8(a); }
static inline unsigned
bytes_to_first_match(const find_mask_t m) { return __builtin_ctz(m); }
static inline integer_t
shift_right_by_one_byte(const integer_t A) {
return simde_mm256_alignr_epi8(A, simde_mm256_permute2x128_si256(A, A, _MM_SHUFFLE(0, 0, 2, 0)), 16 - 1);
@@ -214,6 +202,13 @@ static inline integer_t shuffle_impl256(const integer_t value, const integer_t s
#define debug(...)
#endif
static inline find_mask_t
mask_for_find(const integer_t a) { return movemask_epi8(a); }
static inline unsigned
bytes_to_first_match(const find_mask_t m) { return __builtin_ctz(m); }
// }}}
static inline integer_t