Commit Graph

  • 331d060956 update docs of •FFI dzaima 2022-10-08 19:55:36 +0300
  • 2eecc0feda squeeze *u8/*u16/*u32 result dzaima 2022-10-08 19:38:33 +0300
  • ac041d37f4 fix WARN_SLOW build dzaima 2022-10-08 19:30:08 +0300
  • 86fa4a066e FFI test for mutating a pointer within struct dzaima 2022-10-08 19:21:42 +0300
  • 7e1c9b50cd initial struct support dzaima 2022-10-08 19:06:15 +0300
  • 702d4e979d various utility changes dzaima 2022-10-08 19:03:35 +0300
  • 529d2abae4 add warning about miscompilation for x86 on gcc dzaima 2022-10-08 14:41:22 +0300
  • 7bda795365 fix gcc warnings about pointer casts in 32-bit builds dzaima 2022-10-07 12:42:13 +0300
  • abcb575a53 simplify bit boolean constant replicate start dzaima 2022-10-04 22:11:20 +0300
  • d1783e4706
    Merge pull request #48 from mlochbaum/singelirep dzaima 2022-10-04 22:01:02 +0300
  • ce74e36a19 Fix lots of missing casts for function calls Marshall Lochbaum 2022-09-30 20:20:38 -0400
  • 82ac059b75 disable clmul ≠` when pclmul not available dzaima 2022-09-30 20:06:32 +0300
  • e495efebf8 update Singeli submodule dzaima 2022-09-30 20:00:54 +0300
  • cca64bba70 Move nescan.c into md1.c and constrep.c into slash.c Marshall Lochbaum 2022-09-30 11:40:37 -0400
  • 47cdf02877 Build Singeli files with a filename prefix Marshall Lochbaum 2022-09-30 11:35:25 -0400
  • cad566d547 Not-equals scan with carry-less multiply Marshall Lochbaum 2022-09-29 20:39:46 -0400
  • acc200222d Move constant Replicate Singeli code to its own file Marshall Lochbaum 2022-09-29 20:33:09 -0400
  • 582afe33c9 Merge 3 to 7 replicate float with other types: shuffle instead of permute Marshall Lochbaum 2022-09-29 20:12:45 -0400
  • 29886f355a Combine small-replication shuffle vector tables across types Marshall Lochbaum 2022-09-29 19:58:14 -0400
  • e407286185 BMI2-based constant boolean Replicate for 𝕨<=52 Marshall Lochbaum 2022-09-28 21:56:57 -0400
  • 0c0d967dc3 Adjust Table thresholds for fast constant Replicate Marshall Lochbaum 2022-09-28 17:18:55 -0400
  • 09aba9bc37 slash.singeli is only used with __BMI2__ for now Marshall Lochbaum 2022-09-28 17:18:42 -0400
  • 4836f70e0d Simplify rep_const_broadcast, removing unused kv==0 case Marshall Lochbaum 2022-09-28 14:14:39 -0400
  • d28a74b572 Move 4-shuffle data to a table Marshall Lochbaum 2022-09-28 13:22:55 -0400
  • 9947de0146 Performance improvements when compiling slash.singeli Marshall Lochbaum 2022-09-27 21:29:24 -0400
  • db3e0a6324 Move all the algorithm decision logic into main rep_const and simplify Marshall Lochbaum 2022-09-27 20:47:36 -0400
  • c430922a7e Use shared functions and shuffle array for sizes 3 to 7 in constant Replicate Marshall Lochbaum 2022-09-27 17:38:19 -0400
  • 68aeb0f7cb Add large-constant replicate method always using 4 shuffles Marshall Lochbaum 2022-09-27 15:07:07 -0400
  • 7069a60453 Handle all constant Replicate in Singeli Marshall Lochbaum 2022-09-26 17:19:56 -0400
  • 15c405e7bf Factorize constant replicate lengths under 32 Marshall Lochbaum 2022-09-26 11:15:37 -0400
  • 7535d768dd Singeli shuffle-based code for constant replicates <=7 Marshall Lochbaum 2022-09-26 10:33:54 -0400
  • 851f50635e Some simplifications in base.singeli Marshall Lochbaum 2022-09-26 10:30:10 -0400
  • 46501ac819 update build & packaging information dzaima 2022-09-27 17:47:37 +0300
  • d3ed718a6d branchless sparse Where block size bound was too big dzaima 2022-09-26 23:12:17 +0300
  • bd823839fe fixes for / dzaima 2022-09-25 18:45:27 +0300
  • 1fcd5295eb slight attempt at improving ↕list dzaima 2022-09-25 14:55:49 +0300
  • 5c87c291f9 refactor unroll/vectorization disabling macros dzaima 2022-09-25 14:49:57 +0300
  • 2bb6750d7d better bitarr fill dzaima 2022-09-24 19:31:46 +0300
  • 608489aed1 fix gcc singeli build dzaima 2022-09-24 19:30:10 +0300
  • 4d42e19c27
    Merge pull request #44 from mlochbaum/rep dzaima 2022-09-24 19:21:22 +0300
  • 37a3979c86 clean up generic unit reshape case dzaima 2022-09-24 19:20:32 +0300
  • e8e10790f8 Fast array reshape Marshall Lochbaum 2022-09-23 16:57:41 -0400
  • 218a8b069e Rework reshape of singleton Marshall Lochbaum 2022-09-23 15:32:27 -0400
  • 8e24da8e8e Adjust Replicate thresholds Marshall Lochbaum 2022-09-23 08:22:41 -0400
  • 8a13bf9476 Constant boolean replicate Marshall Lochbaum 2022-09-22 22:32:52 -0400
  • 7f6cf06eea Merge scan-based sparse Indices/Replicate code with macros Marshall Lochbaum 2022-09-22 22:09:56 -0400
  • d647978c8f Type-independent Replicate by constant Marshall Lochbaum 2022-09-22 21:42:18 -0400
  • ec12dd4502 Switch to forward inclusive sums for all radix sorting, and use Singeli Marshall Lochbaum 2022-09-22 15:40:01 -0400
  • 8e1d8bb42c Use Singeli max/min scan for Indices and Sort Marshall Lochbaum 2022-09-21 17:29:10 -0400
  • bc4f5167fb Constant replicate using plus-scan Marshall Lochbaum 2022-09-20 20:33:48 -0400
  • 6ed3c18389 Implement SIMD wrapping plus-scan for Replicate Marshall Lochbaum 2022-09-20 20:24:13 -0400
  • c42f0fd699 revert de4fca85 dzaima 2022-09-22 01:18:41 +0300
  • de4fca859f extract arr_shCopy dzaima 2022-09-22 00:51:04 +0300
  • 06fe437520 bitarr reversing dzaima 2022-09-22 00:38:09 +0300
  • 1e3089ceaa fast ⍉𝕩 when 2=≠𝕩 dzaima 2022-09-20 22:29:32 +0300
  • 7c4673f972 more num_squeezeChk dzaima 2022-09-20 19:27:23 +0300
  • 42ae82eb1f simpler slash_ucw !elInt(w) case dzaima 2022-09-20 19:14:02 +0300
  • 741d7ca31d
    Merge pull request #43 from mlochbaum/cellwidth dzaima 2022-09-20 19:07:21 +0300
  • 42e36851b9 more cleanup dzaima 2022-09-20 18:57:19 +0300
  • f231a30057 fix 𝕨⊔0 dzaima 2022-09-20 17:36:55 +0300
  • cec5b7dc6c slash.c cleanup dzaima 2022-09-20 17:31:39 +0300
  • 01f73dc83b Move usum and bit_sum from numarr.h to stuff.c Marshall Lochbaum 2022-09-19 19:38:26 -0400
  • ac4729bb29 Use grouped compress for all byte sizes if the number of groups is small enough Marshall Lochbaum 2022-09-19 19:31:44 -0400
  • 11e3db1787 Can't call harr withFill before adding data to array Marshall Lochbaum 2022-09-19 11:15:47 -0400
  • 05401547c1 Grouped compress for odd cell widths Marshall Lochbaum 2022-09-18 21:42:24 -0400
  • 9af77e63ee Split up partitioning loop in Group for rank 1 and rank>1 cases Marshall Lochbaum 2022-09-18 16:43:19 -0400
  • 12c4fd0b07 Vectorizable usum code for i8, i16, i32 Marshall Lochbaum 2022-09-18 16:15:20 -0400
  • d525acb51e Merge branch 'master' into cellwidth Marshall Lochbaum 2022-09-18 15:26:34 -0400
  • 9453a9b665 remove old loops for non-bitarr 𝕨 consisting of 0 and 1 from 𝕨/𝕩 dzaima 2022-09-18 22:11:16 +0300
  • 43ddb9759c inline part of depth dzaima 2022-09-18 21:58:53 +0300
  • dd56cfab20 allow using picked element of ⌾(n⊸⊑) in-place dzaima 2022-09-18 21:00:51 +0300
  • 770f7b2f0c check type directly in pick_ucw dzaima 2022-09-18 20:17:48 +0300
  • c0cf58ffc2 use existing function for 8-bit utf8Decode dzaima 2022-09-18 20:17:20 +0300
  • cae65947cd Apply fast Replicate code whenever the cell size is right Marshall Lochbaum 2022-09-17 20:36:43 -0400
  • ea1367e639 Handle cell size 0 correctly in cellWidthLog() Marshall Lochbaum 2022-09-17 20:54:27 -0400
  • c72b9b5abd Redo replicate with sparse and dense cases Marshall Lochbaum 2022-09-17 10:22:34 -0400
  • 5f196c49d9 Extend sparse and dense where to all cases fitting in i32s Marshall Lochbaum 2022-09-16 20:08:23 -0400
  • fb1230bcc2 Sparse Indices prototype Marshall Lochbaum 2022-09-16 20:02:21 -0400
  • 26c3dbfeea Start using type-independent utilities in compress() Marshall Lochbaum 2022-09-16 19:26:55 -0400
  • 3505e1515e Extend Group where 𝕨 is an integer list to handle higher-rank 𝕩 Marshall Lochbaum 2022-09-16 11:42:55 -0400
  • 21033fa355 Move Group to its own file Marshall Lochbaum 2022-09-16 10:59:03 -0400
  • 0b1ded14e0 Type-independent Reverse Marshall Lochbaum 2022-09-14 17:45:48 -0400
  • b84a2d4e1f Special-casing for length 1 in self-search functions Marshall Lochbaum 2022-09-14 15:42:21 -0400
  • 57d33990e3 Make self-search special code only depend on cell size Marshall Lochbaum 2022-09-14 14:48:33 -0400
  • 5bf4e20ead PREFIX and DESTDIR for make install dzaima 2022-09-14 16:57:23 +0300
  • 9c5396e711 consistently output WARN_SLOW information on stderr dzaima 2022-09-12 20:12:32 +0300
  • 89503ba1c1 don't not check top bits in LIKELY & RARE dzaima 2022-09-12 19:50:43 +0300
  • 889c520251 move slash_im setting dzaima 2022-09-12 18:32:56 +0300
  • a46ec7343e not slow ⍉bitarr, fix ⊢⌾⍉ 2‿3‿4⥊↕24 dzaima 2022-09-12 18:27:18 +0300
  • 32d533b6e2 don't print left argument of ⊑ if it's large dzaima 2022-09-12 18:21:04 +0300
  • 47c4a53da4 fix WARN_SLOW build dzaima 2022-09-12 18:16:37 +0300
  • fa4c5424d3
    Merge pull request #42 from mlochbaum/group dzaima 2022-09-12 18:14:12 +0300
  • 6d372959e8 Just return 𝕩 for an all-ones compress Marshall Lochbaum 2022-09-11 21:48:31 -0400
  • 9b17701c66 Sparse 1- and 2-byte compress, with blocking, if total sum is low enough Marshall Lochbaum 2022-09-11 20:08:20 -0400
  • c796670549 Faster trailing zero trim, and apply it to all cases in compress() Marshall Lochbaum 2022-09-11 16:28:06 -0400
  • 2baa88ad31 Compress for 4-byte and 8-byte data using index blocks Marshall Lochbaum 2022-09-11 16:28:06 -0400
  • ab25b08f84 Move compress (bool/list) to its own function Marshall Lochbaum 2022-09-11 15:53:26 -0400
  • 819899c745 Implement f64 Where with u16 blocks Marshall Lochbaum 2022-09-11 15:42:28 -0400
  • fda3efe217 More macros for Where Marshall Lochbaum 2022-09-11 15:31:23 -0400
  • 0d6894ddbc Fix broken non-Singeli dense i32 Where Marshall Lochbaum 2022-09-11 15:15:56 -0400