Commit Graph

  • d1f3efe8db get rid of shuf_ind dzaima 2025-04-28 18:01:52 +0300
  • b2c89979fe exclude [k]u1 from vector type check helpers dzaima 2025-04-28 02:01:19 +0300
  • c811e54666 unroll bitsel dzaima 2025-04-28 04:15:32 +0300
  • 0ae0be135c bump for_build binary dzaima 2025-04-29 03:17:00 +0300
  • ae3981a673 update Singeli submodule dzaima 2025-04-28 18:03:44 +0300
  • 2a9428239a merge sse2.singeli and sse.singeli dzaima 2025-04-27 21:36:47 +0300
  • 07e74d62d8 prepare for merging sse2.singeli and sse.singeli dzaima 2025-04-27 21:35:59 +0300
  • ccb1f5fa34 proper x86 i32→float convert definiton dzaima 2025-04-27 22:13:06 +0300
  • f72f554a54 remove duplicate and_bit_none SSE definition dzaima 2025-04-27 21:31:38 +0300
  • a5840d0eb8 fix AVX-512 widen arch check dzaima 2025-04-27 19:12:51 +0300
  • e3a985c6e2 proper packs128 instead of packQ dzaima 2025-04-27 19:06:48 +0300
  • 7bf55793c5 make build.bqn error on lack of output even if there's no errout dzaima 2025-04-27 18:14:25 +0300
  • 26055d0146 generalized x86 load/store low elts dzaima 2025-04-24 18:38:10 +0300
  • 70705f9bfe move bmi.singeli into base x86 include dzaima 2025-04-24 18:36:10 +0300
  • 98f1c97653 is_pow2 helper dzaima 2025-04-24 18:31:25 +0300
  • 0ec37303ea use specified-output-type fold_addw dzaima 2025-04-24 00:05:38 +0300
  • 3d52d14101 use vecfold.singeli's x86 fold_addw in count.singeli dzaima 2025-04-17 20:43:45 +0300
  • 5df3712748 allow specifying result type of fold_addw, x86 impls for it dzaima 2025-03-24 17:38:45 +0200
  • 7237ad1abb more Singeli renames: no more floor,ceil,abs,adds,subs,vshl dzaima 2025-04-23 23:51:15 +0300
  • 2202726005 rename isvec→vect and isprim→ptrt dzaima 2025-04-23 23:18:44 +0300
  • 6eb454f859 get rid of the single isptr usage dzaima 2025-04-23 23:16:15 +0300
  • c1662c83de move x86 mul32 and mulh defs to x86.singeli dzaima 2025-04-23 22:23:34 +0300
  • e46bbf398e 512-bit & arbitrary-width vector check helpers dzaima 2025-04-23 22:23:14 +0300
  • 64a71688fe use util/kind checks instead of custom isreg/istype/istup dzaima 2025-04-23 23:10:29 +0300
  • ee1668c7ea use __min & __max directly instead of custom min/max dzaima 2025-04-23 23:07:33 +0300
  • 8beff254e2 x86 saturating multiply sum defs dzaima 2025-03-24 00:30:00 +0200
  • 44ee417d40 use pattern matching for NEON vfold cases dzaima 2025-03-24 00:29:19 +0200
  • c95482e8e0 gcc-15 should fix 32-bit x86 miscompilation dzaima 2025-04-18 18:45:27 +0300
  • d2b9f4a02e a bit of src/README.md cleanup dzaima 2025-04-08 18:52:58 +0300
  • 64339466cc more clear messages on empty_free / builtin_free dzaima 2025-04-21 03:39:58 +0300
  • f9b3aba234 fix wrong integer types in varargs dzaima 2025-04-18 18:29:40 +0300
  • 732f52630a use AVX-512 integer narrow if available dzaima 2025-04-03 03:00:49 +0300
  • 528fea645a nicer store-low in bins.singeli dzaima 2025-04-03 03:30:12 +0300
  • 78359afc71 move x86 int widen into x86.singeli dzaima 2025-04-03 01:28:16 +0300
  • dbe97cf0e0 move some basic things into x86.singeli dzaima 2025-04-03 00:56:07 +0300
  • 5928bd91bc note problematic latency of dpbusd/dpwssd dzaima 2025-04-03 00:37:11 +0300
  • 7ed5ba3b2f list available options on wrong clean= / rebuild= dzaima 2025-04-03 00:36:49 +0300
  • 63b7fab72c better w+x & w-x overflow checks dzaima 2025-03-26 04:10:59 +0200
  • 09642a354f do reqsz != (size_t) reqsz check in NO_MMAP builds too dzaima 2025-03-24 18:08:10 +0200
  • 68a969b417 get clang to do less horrible codegen for index calculations dzaima 2025-03-24 17:48:42 +0200
  • ad73e075e0 unroll bitwise arith dzaima 2025-03-24 17:39:51 +0200
  • e3239d29bd enable 32-byte alignment for array data dzaima 2025-03-24 17:39:40 +0200
  • 60b848f1b7 don't require strict subset type for trunc dzaima 2025-03-22 19:09:51 +0200
  • c7632c23a2 actually, can define the x86 sadbw-using vfold for i8 dzaima 2025-03-22 18:20:59 +0200
  • 4e03a6310c fix proper-fills-only error message dzaima 2025-03-21 23:35:38 +0200
  • fa2fccf218 update limitations section in readme dzaima 2025-03-21 21:37:30 +0200
  • ae29c4ad45 fix usz=32 build dzaima 2025-03-21 18:08:12 +0200
  • 9a45762ecd more groupstat tests dzaima 2025-03-21 04:39:37 +0200
  • 609f3418d8 use offset loads instead of vec_merge_shift_right for groupstat dzaima 2025-03-21 04:07:13 +0200
  • e2b4dbcf20 use any_hom / all_hom for (any|all)_accumulator dzaima 2025-03-21 04:05:10 +0200
  • 05d87dd7df move groupstat to fancy accumulator dzaima 2025-03-21 01:04:34 +0200
  • 97260e0051 special-cased x86 SSE2 i8 min/max accumulator dzaima 2025-03-20 21:32:15 +0200
  • f503880cc3 use fancy accumulator for getRange dzaima 2025-03-20 21:34:16 +0200
  • e1b62b95b6 fancy Singeli accumulator; use for integer sum dzaima 2025-03-20 21:32:11 +0200
  • 3b1239d499 support requesting a flush every k iterations in for_mu dzaima 2025-03-20 20:38:28 +0200
  • 9c2ea18e22 use x86 sad for vfold dzaima 2025-03-20 20:35:48 +0200
  • 6c6f6e4013 add x86 sad & madd intrinsics dzaima 2025-03-20 21:17:52 +0200
  • 9876d1d622 reduce Singeli-generated C noise on constant broadcasts dzaima 2025-03-21 02:39:03 +0200
  • d6d1ab286f update Singeli submodule dzaima 2025-03-21 04:29:28 +0200
  • 4a5a1fa35b a couple more baseline utilities dzaima 2025-03-21 01:04:25 +0200
  • 87be577ff7 actually use defined is in for_mu dzaima 2025-03-19 20:46:54 +0200
  • 0c2fb1a267 a bit of cleanup dzaima 2025-03-12 05:22:11 +0200
  • 577b09a062 introduce x86.singeli dzaima 2025-03-19 00:53:13 +0200
  • 108379f121 fix UB shift dzaima 2025-03-16 23:46:59 +0200
  • f1021580ad silence warning on non-Singeli builds dzaima 2025-03-16 20:22:17 +0200
  • e5c53eefd7 update bytecode submodule dzaima 2025-03-16 20:03:09 +0200
  • 9c870ddab9 add more info in heap dump dzaima 2025-03-16 03:52:23 +0200
  • 37911af67a show Singeli show{}s even when erroring dzaima 2025-03-13 23:38:18 +0200
  • 9023f6fe3b
    Merge pull request #138 from mlochbaum/groupstat dzaima 2025-03-13 23:37:44 +0200
  • 82ebd37b3f group statistics test dzaima 2025-03-13 20:41:46 +0200
  • 233c88120d Singeli function for initial group statistics pass since auto-vectorization isn't great Marshall Lochbaum 2025-03-12 19:06:59 -0400
  • 72c4eca6c0 fix float ⌾(l⊸⊏) doing wrong comparisons dzaima 2025-03-12 17:23:17 +0200
  • d48194c3c8 fix comment dzaima 2025-03-12 17:20:18 +0200
  • 7a412523dd better SSE2 float getRange dzaima 2025-03-12 16:08:10 +0200
  • 62b095be89 oneElt⊏𝕩 improvements & fixes dzaima 2025-03-08 01:46:17 +0200
  • 3ae7d134d2 more uniform F¨ fill behavior dzaima 2025-03-11 04:32:14 +0200
  • b14ea82218 accept two error messages for 2⋆63 test dzaima 2025-03-12 04:17:21 +0200
  • a013eb094e select_fn doesn't require SIMD anymore dzaima 2025-03-12 03:42:48 +0200
  • bfce527096 fix 𝕗⌾(-2⋆63)𝕩 dzaima 2025-03-12 02:25:24 +0200
  • e33b3b95ee relax documented RangeFn behavior to align with NEON impl dzaima 2025-03-12 02:00:48 +0200
  • a6202c8ca6 fix error message of ⌾(oob⊸⊏) dzaima 2025-03-12 01:36:54 +0200
  • 0848faf8c7 don't leak memory in TEST_RANGE dzaima 2025-03-12 01:14:24 +0200
  • 76fc956e7c note about test_range test needing a specific build dzaima 2025-03-12 01:04:04 +0200
  • 5f5bb37a92 fix builtin function comparison on ALL_R1 dzaima 2025-03-12 01:02:25 +0200
  • d2ca600e02 actually make try_first_hom return things dzaima 2025-03-11 23:44:15 +0200
  • 85ef1038b8 fix usz=32 build dzaima 2025-03-11 23:38:09 +0200
  • cca858ceee test system improvements dzaima 2025-03-11 23:38:22 +0200
  • dfce405111
    Merge pull request #137 from mlochbaum/scalreduce dzaima 2025-03-11 19:23:57 +0200
  • f5fdd50f75 test alternating fills of ˝ dzaima 2025-03-11 04:32:14 +0200
  • dcecc70c25 More careful fill handling with characters in insert_scal Marshall Lochbaum 2025-03-10 20:26:38 -0400
  • 97e0210c41 remove unnecessary F˜˘ acceptance from F˘ dzaima 2025-03-10 20:30:45 +0200
  • bcf32838f4 some more ˝ tests, use COPY_TO dzaima 2025-03-11 00:15:44 +0200
  • f1972fc408 Faster stride-1 check Marshall Lochbaum 2025-03-10 12:57:25 -0400
  • e42273bfe8 Generic stride-1 𝔽`˘ implementation Marshall Lochbaum 2025-03-10 12:50:54 -0400
  • bbe4549aff Scalar implementation of 𝔽˝ for generic arithmetic 𝔽 and short rows Marshall Lochbaum 2025-03-10 12:00:09 -0400
  • e09fb53fb9
    Merge pull request #136 from mlochbaum/boolrep dzaima 2025-03-09 23:42:12 +0200
  • b69a2c2469 Move swtab outside of its function; it was being compiled as a dynamic array Marshall Lochbaum 2025-03-07 21:25:00 -0500
  • e898a81d68 Use a possibly-overlapping word and constant-length loop in k/bool, like ⊣` Marshall Lochbaum 2025-03-07 18:17:17 -0500
  • 302c2f926f AVX2 k/bool for odd 64<k<256 using shift-based masks Marshall Lochbaum 2025-03-07 15:26:50 -0500
  • bd942894bf fix missing usage of unaligned store dzaima 2025-03-06 05:59:13 +0200