Commit Graph

  • 7c4599543b test •platform dzaima 2023-08-19 17:56:56 +0300
  • ee1b487624 use exact desired bit pattern for NaN in test dzaima 2023-08-19 17:35:39 +0300
  • b237f170fb bump for_build binary dzaima 2023-08-19 17:58:17 +0300
  • 25c2629842 •platform dzaima 2023-08-19 03:22:55 +0300
  • 568c49ff84 update bytecode submodule dzaima 2023-08-19 17:16:38 +0300
  • 4d19654e3b fix store types dzaima 2023-08-16 18:55:19 +0300
  • acca5c0c7c fix completely wrong dyTable lookup dzaima 2023-08-15 14:45:18 +0300
  • 6e2ec89102 verify assumptions made by tests of ¯0 dzaima 2023-08-13 03:54:29 +0300
  • df56f6b2fc fix 𝕨•FLines𝕩 with invalid elements in 𝕩 dzaima 2023-08-13 03:51:06 +0300
  • a175c48104
    Merge pull request #89 from mlochbaum/pext dzaima 2023-08-08 14:11:52 +0300
  • 2dd3a0fdfa Boolean compress implementation comments Marshall Lochbaum 2023-08-07 15:33:37 -0400
  • 4e9745688b Use generic boolean Compress as the sparse case with Singeli Marshall Lochbaum 2023-08-07 13:44:52 -0400
  • a7908a10b0 Faster, sparse, non-Singeli boolean Compress Marshall Lochbaum 2023-08-07 12:26:40 -0400
  • bbad98bce1 In AVX2 boolean Compress, jump from 8 to 32 with 32-bit shifts Marshall Lochbaum 2023-08-07 11:32:02 -0400
  • 120df26814 mullo works for both signed and unsigned Marshall Lochbaum 2023-08-07 11:22:40 -0400
  • d039562efe Fix disabled BMI2 boolean Compress; cleaner SIMD width system Marshall Lochbaum 2023-08-06 08:09:08 -0400
  • 9046dd6b53 Handle boolean Compress SIMD tail with conditional write, not scalar loop Marshall Lochbaum 2023-08-06 08:03:14 -0400
  • 07ace41d6c Vectorized version of the clmul boolean Compress Marshall Lochbaum 2023-08-06 07:36:05 -0400
  • 371aa6f2ef AVX2 boolean Compress using variable shifts Marshall Lochbaum 2023-08-05 21:50:22 -0400
  • 66c0fe041c Get pext and popcount together, since generic pext emulation includes popc Marshall Lochbaum 2023-08-04 21:43:17 -0400
  • 056b18edd2 Move pext-based compress loop from C to Singeli Marshall Lochbaum 2023-08-04 21:30:06 -0400
  • 7767cf933f Remove boolean Compress overallocation: not needed with w masked off Marshall Lochbaum 2023-08-04 21:26:20 -0400
  • e09d385df5 Simplify pairwise pext implementation; drop 12-bit as no longer competitive Marshall Lochbaum 2023-08-01 21:31:20 -0400
  • 7fe1ab930a FFI: permit 𝕨≡⟨⟩, require 𝕩≡⟨⟩ if no arguments are on 𝕩, check list rank dzaima 2023-08-05 16:05:57 +0300
  • fd4477992f fall back from clang to cc in build.bqn & default makefile to cc dzaima 2023-08-04 21:44:36 +0300
  • 682d5986d8 slightly more accurate trailing newline removal for •GetLine dzaima 2023-08-04 21:09:45 +0300
  • a1e4f9bd6f remove multiple trailing newline characters from •GetLine dzaima 2023-08-04 20:52:07 +0300
  • 8e35241e47 Better error messages for unknown namespace key dzaima 2023-08-03 22:54:38 +0300
  • 0b6c5189d8 move arithdDispatch declarations to a header dzaima 2023-06-21 18:28:27 +0300
  • 627a7e0771 more FFI error checking, plus bugfixes dzaima 2023-07-28 14:43:17 +0300
  • 1bc96f5557 extract u8/u16/u32/f32 reading functions dzaima 2023-07-28 13:59:51 +0300
  • 1454ca3a1a expand & test FFI error checking dzaima 2023-07-28 13:59:25 +0300
  • 79f72d84b8 extract unsigned & f32 copying functions in ffi.c dzaima 2023-07-28 01:47:57 +0300
  • ed07995e68 don't rely on thing that doesn't hold outside of Singeli SIMD-ful builds dzaima 2023-07-26 17:52:56 +0300
  • 0d7de8b190 define ux in Singeli, replacing Size; return ux from popc/clz/ctz dzaima 2023-07-11 18:02:42 +0300
  • f416d58579 compiler version updates in readme dzaima 2023-07-15 22:18:02 +0300
  • f7a0eb94a3 don't use AVX2 buffer copying for a 16-byte buffer dzaima 2023-07-22 21:33:43 +0300
  • 1451395ef5 don't do fancy char stuff for unchecked arith dzaima 2023-07-22 21:13:49 +0300
  • 959614c785 use partial application dzaima 2023-07-22 18:19:31 +0300
  • 033f3de6b9 remove unnecessary lambda wrappers dzaima 2023-07-22 17:58:34 +0300
  • cc160910a2 update Singeli submodule dzaima 2023-07-22 17:43:58 +0300
  • 446b2d09ed Singeli renames & docs dzaima 2023-07-22 14:11:28 +0300
  • 6e96600878 move unaligned load & store to base.singeli dzaima 2023-07-20 00:21:09 +0300
  • 0486adccb5 fix signedness of abs{x} dzaima 2023-07-20 00:19:54 +0300
  • 2243fa76c1 better base case list dzaima 2023-07-19 21:56:35 +0300
  • eb46ec8d9c rename __mulhi & __mul32 dzaima 2023-07-19 21:54:12 +0300
  • 700d69e7dd unpack → zip & mzip; enable slash.singeli SIMD parts for NEON dzaima 2023-07-19 21:18:01 +0300
  • ba2e859dce use existing unpackLo in neq.singeli dzaima 2023-07-19 20:56:08 +0300
  • 0a4a0f3e2a
    Merge pull request #88 from mlochbaum/compress dzaima 2023-07-19 20:46:52 +0300
  • 19b35e1074 fix size comment dzaima 2023-07-19 20:11:14 +0300
  • 9945711b5e AVX2 i32 & i64 slash{} via widening load dzaima 2023-07-19 19:58:41 +0300
  • ff9bb258c2 more comments in slash.singeli dzaima 2023-07-19 18:50:01 +0300
  • 48111d4873 extract iota{} for AVX-512 dzaima 2023-07-19 17:24:20 +0300
  • a0797318ca use has_simd & arch_defvw dzaima 2023-07-19 17:03:11 +0300
  • 03c21112c2 make NEON builds work dzaima 2023-07-19 16:58:35 +0300
  • 6930bc34b6 further update slash.c implementation comments dzaima 2023-07-19 16:51:14 +0300
  • 77188b8e73 Add clmul.singeli Marshall Lochbaum 2023-07-19 09:36:50 -0400
  • 301a23fb1f Update where/compress implementation comments Marshall Lochbaum 2023-07-19 09:35:40 -0400
  • c4364aefa6 SSSE3 4-byte compress with half-width table Marshall Lochbaum 2023-07-19 08:15:53 -0400
  • ba837ba01b Remove table-less where/compress methods, as they're not competitive Marshall Lochbaum 2023-07-19 07:22:55 -0400
  • 64d65ae837 Never use generic dense compress methods if use_table: where is faster Marshall Lochbaum 2023-07-18 21:35:08 -0400
  • b0e0f210c6 2- and 4-byte Where with unpack and table Marshall Lochbaum 2023-07-18 21:10:39 -0400
  • 93e1262864 Avoid using popcount primitive in Singeli if POPCNT isn't present Marshall Lochbaum 2023-07-18 16:52:08 -0400
  • 6df283c379 Always increment by full popcount instead of two halves when splitting byte Marshall Lochbaum 2023-07-17 17:52:59 -0400
  • 2f6a8e55a1 Always use vectors for for_special_buffered buffer copy Marshall Lochbaum 2023-07-17 16:31:01 -0400
  • 8add95b7ab Remove BMI2 where/compress methods Marshall Lochbaum 2023-07-17 16:20:30 -0400
  • f005169518 1- and 2-byte compress with table and SSSE3 Marshall Lochbaum 2023-07-17 16:19:05 -0400
  • 68978f7e10 Table-based 1- and 2-byte Where Marshall Lochbaum 2023-07-17 16:05:06 -0400
  • fc187afdf2 Extend 1-byte SSE2 Where to 2-byte and 4-byte with unpacked writes Marshall Lochbaum 2023-07-17 11:28:04 -0400
  • 8b297ae2dc Dedicated 32-bit Singeli where, taking index offset Marshall Lochbaum 2023-07-16 21:32:48 -0400
  • 3bd8d1de68 Extend all Singeli Compress methods to do Where Marshall Lochbaum 2023-07-16 20:18:54 -0400
  • 4415869496 Generic Singeli where implementation Marshall Lochbaum 2023-07-16 18:38:27 -0400
  • c484f8ee8d Avoid UB from shifting by 64 Marshall Lochbaum 2023-07-16 17:22:26 -0400
  • da5f1faa07 Merge BMI2 slash{} cases, with some simplifications Marshall Lochbaum 2023-07-16 17:05:09 -0400
  • b244ba98f5 Unify slash1{} and slash2{} Marshall Lochbaum 2023-07-16 12:40:40 -0400
  • f5121f6806 Thresholds for compress methods (versus sparse) defined in Singeli Marshall Lochbaum 2023-07-16 08:19:20 -0400
  • 824bb04b4d Plain SSE2 1-byte compress Marshall Lochbaum 2023-07-15 11:24:54 -0400
  • f852d12a5c Generic Singeli compress implementation Marshall Lochbaum 2023-07-15 11:14:52 -0400
  • f36426031a Slightly faster table-less compress method Marshall Lochbaum 2023-07-14 20:44:11 -0400
  • 84a928863b AVX-512 compress implementations Marshall Lochbaum 2023-07-14 19:25:53 -0400
  • 80e93c41c9 Use new undefined{type,len} instead of alloca Marshall Lochbaum 2023-07-14 18:46:02 -0400
  • 13b3bcd74b Table-less AVX2 1-byte compress Marshall Lochbaum 2023-07-13 16:31:59 -0400
  • d956ba921b Non-overwriting 1-byte and 2-byte compress/where Marshall Lochbaum 2023-07-13 13:24:41 -0400
  • 8f4b1966cb Non-overwriting AVX2 4-byte and 8-byte compress, using a buffer Marshall Lochbaum 2023-07-13 12:17:47 -0400
  • 9010f648bb Unify permutevar-based compress implementations Marshall Lochbaum 2023-07-12 21:44:38 -0400
  • 0921fc2c62 AVX2 compress implementations using permutevar and lookup tables Marshall Lochbaum 2023-07-12 21:26:17 -0400
  • 03d8919307 Add has='pclmul' option and avoid using broadcast Marshall Lochbaum 2023-07-11 19:52:47 -0400
  • 871a81ca5f BMI2 pext not-emulation in Singeli Marshall Lochbaum 2023-07-11 15:43:45 -0400
  • 1314859a80 clmul-based pext emulation Marshall Lochbaum 2023-07-11 15:36:43 -0400
  • f315a6d3ea Emulate pext instruction in Singeli Marshall Lochbaum 2023-07-11 10:57:15 -0400
  • 3946f2cc87 use new Singeli builtins dzaima 2023-07-17 02:23:50 +0300
  • cdd3015940 update Singeli submodule dzaima 2023-07-17 02:18:07 +0300
  • 6305a3bd50 change error message of failed path_abs dzaima 2023-07-16 16:26:16 +0300
  • cf6a290e5c •file.RealPath dzaima 2023-07-16 01:38:51 +0300
  • 1a6e0282e4 another update of the Singeli submodule dzaima 2023-07-14 21:30:36 +0300
  • 2a8d13768e update Singeli submodule dzaima 2023-07-14 18:31:26 +0300
  • 8c7e26e999 change BQN link in readme to pages site dzaima 2023-07-14 13:55:47 +0300
  • 08bdf20bb9 update moreCfgs.sh test dzaima 2023-07-13 16:29:05 +0300
  • d1e511bebd add padding before allocations & document it dzaima 2023-07-13 15:44:09 +0300
  • f0e807c40e update expected test result dzaima 2023-07-13 15:47:04 +0300