Commit Graph

696 Commits

Author SHA1 Message Date
dzaima
e2b4dbcf20 use any_hom / all_hom for (any|all)_accumulator 2025-03-21 06:01:27 +02:00
dzaima
05d87dd7df move groupstat to fancy accumulator 2025-03-21 06:01:26 +02:00
dzaima
97260e0051 special-cased x86 SSE2 i8 min/max accumulator 2025-03-21 04:32:26 +02:00
dzaima
f503880cc3 use fancy accumulator for getRange 2025-03-21 04:32:26 +02:00
dzaima
e1b62b95b6 fancy Singeli accumulator; use for integer sum 2025-03-21 04:32:26 +02:00
dzaima
3b1239d499 support requesting a flush every k iterations in for_mu 2025-03-21 04:32:26 +02:00
dzaima
9c2ea18e22 use x86 sad for vfold
not used by anything, but it's here
2025-03-21 04:32:26 +02:00
dzaima
6c6f6e4013 add x86 sad & madd intrinsics 2025-03-21 04:32:26 +02:00
dzaima
9876d1d622 reduce Singeli-generated C noise on constant broadcasts 2025-03-21 04:32:26 +02:00
dzaima
4a5a1fa35b a couple more baseline utilities
assert_hom to be expanded upon later
2025-03-21 01:05:53 +02:00
dzaima
87be577ff7 actually use defined is in for_mu 2025-03-20 20:27:26 +02:00
dzaima
0c2fb1a267 a bit of cleanup
still from the long-ago days when symbols couldn't be assignment targets directly
2025-03-19 01:05:22 +02:00
dzaima
577b09a062 introduce x86.singeli 2025-03-19 01:05:15 +02:00
Marshall Lochbaum
233c88120d Singeli function for initial group statistics pass since auto-vectorization isn't great 2025-03-13 14:21:38 -04:00
dzaima
d48194c3c8 fix comment 2025-03-12 17:20:18 +02:00
dzaima
7a412523dd better SSE2 float getRange 2025-03-12 16:38:20 +02:00
dzaima
a013eb094e select_fn doesn't require SIMD anymore
it's not used too much without SIMD extensions available, but it is used a bit
2025-03-12 04:20:53 +02:00
dzaima
d2ca600e02 actually make try_first_hom return things 2025-03-12 00:20:49 +02:00
dzaima
85ef1038b8 fix usz=32 build 2025-03-11 23:41:11 +02:00
Marshall Lochbaum
b69a2c2469 Move swtab outside of its function; it was being compiled as a dynamic array 2025-03-09 08:44:42 -04:00
Marshall Lochbaum
e898a81d68 Use a possibly-overlapping word and constant-length loop in k/bool, like ⊣` 2025-03-09 08:44:42 -04:00
Marshall Lochbaum
302c2f926f AVX2 k/bool for odd 64<k<256 using shift-based masks 2025-03-09 08:44:42 -04:00
dzaima
bd942894bf fix missing usage of unaligned store 2025-03-06 06:01:44 +02:00
dzaima
d5facc33ad some Singeli cleanup
shuf_ind's staying as such I guess; make it more clear when debug.singeli is left included
2025-03-06 05:32:22 +02:00
dzaima
267ddc4987 clarifying comments 2025-03-06 04:02:27 +02:00
Marshall Lochbaum
a11ed3d79a Mask out values past the end for strided +` overflow checks 2025-03-04 07:39:28 -05:00
Marshall Lochbaum
46300bcb51 Use scalar case for small-stride +` on <SSSE3 to avoid incorrectly overlapping vectors 2025-03-03 08:15:47 -05:00
Marshall Lochbaum
4072bde806 Strided scan with AVX2 [8]i32 permute if possible 2025-03-03 08:14:17 -05:00
Marshall Lochbaum
180c79e751 Strided +`, widening and re-trying on overflow 2025-03-03 06:54:40 -05:00
Marshall Lochbaum
87a7d066c8 Strided ⌊⌈ scans with shuffles 2025-03-03 06:54:40 -05:00
Marshall Lochbaum
360e4e8320 Save a word instead of writing to avoid stall for scan with 64<stride<128 2025-03-03 06:54:40 -05:00
Marshall Lochbaum
f3e0ea7531 Use CLMUL-based ≠` for power-of-two stride up to 64 2025-03-03 06:54:40 -05:00
Marshall Lochbaum
ee6b91be8a Basic strided ∧∨≠= scan implementations 2025-03-03 06:54:40 -05:00
Marshall Lochbaum
b03b26656d Fix usz=32 build 2025-03-03 06:54:40 -05:00
dzaima
445007b38c unify toLast and broadcast_last 2025-03-02 19:31:03 +02:00
dzaima
bdee96a995 remove old blend definitions 2025-02-27 05:49:16 +02:00
Marshall Lochbaum
eda1cd8b50 More tuning based on dzaima's measurements 2025-02-26 22:01:37 -05:00
Marshall Lochbaum
a520ff70a1 Fix some NEON performance issues for row scans 2025-02-26 21:52:49 -05:00
Marshall Lochbaum
422e63bc12 Non-CLMUL SIMD ≠` 2025-02-26 08:21:22 -05:00
Marshall Lochbaum
d85d69e759 Improve baseline for ⊣`˘ to beat non-AVX2 on length ≥64 2025-02-25 20:26:51 -05:00
Marshall Lochbaum
709712f03c Tuning; drop the scalar loose_mask branch for ∧∨ as it never seems faster 2025-02-25 19:33:26 -05:00
Marshall Lochbaum
b5ec2f5fa8 Extend loose_mask system to SSE and NEON (not always fast) 2025-02-25 19:28:51 -05:00
Marshall Lochbaum
fb413b1966 Enable vector max- min-, and wrapping plus-scan on NEON
Changes the results on NaNs, but they weren't consistent before
2025-02-24 22:08:49 -05:00
Marshall Lochbaum
317871db55 Extend unaligned <64-bit row scans from AVX2 to SSE, NEON 2025-02-24 21:47:46 -05:00
Marshall Lochbaum
b131914cbe Factor out some common code 2025-02-24 21:13:19 -05:00
Marshall Lochbaum
0973d9b499 Medium-row ⊣˘ and ≠˘ with mask-aware carry 2025-02-24 21:13:19 -05:00
Marshall Lochbaum
1fef51a39e AVX2 ∧ and ∨ with intermediate row sizes (64≤l<160) 2025-02-23 19:19:46 -05:00
Marshall Lochbaum
94e6b6cb3a Better formula for within-word masked ∧ and ∨ 2025-02-23 08:27:47 -05:00
dzaima
8c1040bd9f cleaner vr?(add|sub)hn NEON defs 2025-02-08 18:37:46 +02:00
dzaima
09f2980ebf more BMI and NEON defs 2025-02-08 17:36:44 +02:00