Releases · exo-lang/ExoBLAS

What's Changed

Restructure tests correctness/bench by @SamirDroubi in #1
Parse json output to generate graph by @yamaguchi1024 in #3
Merge correct and okay fast gemm by @yamaguchi1024 in #5
Revert "testing branch protection" by @SamirDroubi in #6
Fail-fast check in hoist by @SamirDroubi in #7
update graph for GEMM by @yamaguchi1024 in #8
Do instruction selection early in vectorize by @SamirDroubi in #9
N and M must be lowercase for the graph script by @yamaguchi1024 in #10
Rewrite gemm microk and improve vectorizer by @SamirDroubi in #11
Tile_loops and auto_stage_mem operations by @SamirDroubi in #12
Rerwite Trans algos + retune parameters for NonTrans case by @SamirDroubi in #13
Add cmath header for fabs on mac by @yamaguchi1024 in #14
use original algorithm and call_eqv on the scheduled version by @yamaguchi1024 in #17
First refactor of composed schedules by @SamirDroubi in #18
Reshift loops after cut_loop by @SamirDroubi in #19
Change various compile flags by @SamirDroubi in #20
Composed schedule refactor 2 by @SamirDroubi in #21
Rewrite asum by @SamirDroubi in #22
Rewrote copy by @SamirDroubi in #23
Rewrite rot by @SamirDroubi in #24
Rewrite rotm by @SamirDroubi in #25
Rewrite scal by @SamirDroubi in #26
Rewrite swap by @SamirDroubi in #27
Rewrite dsdot by @SamirDroubi in #28
Update remove-if to eliminate-dead-code by @SamirDroubi in #29
Implement clean Gemmini Gemm and preliminary SAD by @yamaguchi1024 in #30
Implement sgemm by @SamirDroubi in #32
Renamed ForSeqCursor to ForCursor by @skeqiqevian in #31
Update github ci workflow files by @yamaguchi1024 in #33
Deprecate CSE argument from bind_expr by @SamirDroubi in #34
Test codegen by @SamirDroubi in #35
Implement a more robust hoist_stmt by @SamirDroubi in #36
Cleanup introspection code by @SamirDroubi in #37
Make names created by ops unique by @SamirDroubi in #38
Cleanup par red by @SamirDroubi in #39
Optimize level 1 operation by @SamirDroubi in #40
Automate reduction parallelism by @SamirDroubi in #41
Implement higher-order functions by @SamirDroubi in #42
Implement hoist_from_loop op by @SamirDroubi in #43
Rewrite gemv and start an optimize_level_2 op by @SamirDroubi in #44
Simplify vectorize and support various tail strats by @SamirDroubi in #45
Rewrite interleave_execution by @SamirDroubi in #46
Implement unroll_buffers and get_inner_loop ops by @SamirDroubi in #47
Implement stage_compute by @SamirDroubi in #48
Vectorization stage compute by @SamirDroubi in #49
Rewrite asum as optimize_level_1 by @SamirDroubi in #50
Rewrite dsdot via optimize level 1 by @SamirDroubi in #51
Support rc mode for hoist operations by @SamirDroubi in #53
Remove unsafe_disable_checks from expand_dim by @yamaguchi1024 in #54
Restructure vectorization by @SamirDroubi in #55
Deprecate bound_alloc by @SamirDroubi in #56
Automate nth_loop in parallelize_all_reductions by @SamirDroubi in #57
Interleave restructure by @SamirDroubi in #58
Rewrite optimize_level_2 by @SamirDroubi in #59
Generate gemv variant by @SamirDroubi in #60
Rewrite auto-stage-mem to handle general blocks by @SamirDroubi in #61
Check codegen hash instead of sources by @SamirDroubi in #62
Add Preliminary CSE support by @SamirDroubi in #63
Update CI to use the new M1 macOS runner by @yamaguchi1024 in #64
Fix Github workflows by @SamirDroubi in #65
Support triangular matrices in optimize_level_2 by @SamirDroubi in #66
Deprecate parameters class by @SamirDroubi in #68
Minimize Build/Test time by @SamirDroubi in #69
Automation of precision-stride variants generation by @SamirDroubi in #70
Rewrite ger by @SamirDroubi in #71
Rewrite symv by @SamirDroubi in #72
Rewrite syr2 by @SamirDroubi in #73
Rewrite syr by @SamirDroubi in #74
Rewrite trsv by @SamirDroubi in #75
New graphing script and start metaprogramming tests by @SamirDroubi in #76
Restructure lib and deprecate legacy code by @SamirDroubi in #77
Parameterize Tail Strategies for level 1 and 2 by @SamirDroubi in #78
Changes to speedup compilation time by @SamirDroubi in #79
Enable inner tails in symmetric/triangular matrices by @SamirDroubi in #80
Fix extract_subproc call by @SamirDroubi in #81
Make workflows reusable by @SamirDroubi in #84
Generate various performance features with kernels by @SamirDroubi in #83
Apply yapf fix to workflow by @SamirDroubi in #85
New scheduling flow for gemm by @SamirDroubi in #87
Remove Gemmini matmul code from BLAS by @yamaguchi1024 in #86
Rewrite gemm testing by @SamirDroubi in #88
Implement filter_cursors op by @SamirDroubi in #89
Gemm macro micro by @SamirDroubi in #90
Add AVX512 target by @SamirDroubi in #91
Bump black from 22.10.0 to 24.3.0 by @dependabot in #92
Write the uk schedule by hand to manage comp time by @SamirDroubi in #93
syrk schedule (missing calls to gemm after tiling) by @SamirDroubi in #94
call gemms from syrk after tiling by @SamirDroubi in #95
Implement symm algorithms and testing by @SamirDroubi in #96
Schedule symm by @SamirDroubi in #97
Generate symm variants by @SamirDroubi in #98
Implement syr2k algorithm and testing by @SamirDroubi in #99
Level 3 Testing Fixes by @SamirDroubi in #100
Support output matrix scaling in level 3 by @SamirDroubi in #101
Support transposition options in gemm and syrk by @SamirDroubi in #102
Automate cutting of iteration space for Uplo by @SamirDroubi in #103
LoC analytics script by @SamirDroubi in #104
Rewrite trmv, trsv, and ger in the new testing format by @SamirDroubi in #105
Heatmap summary plots by @SamirDroubi in #106
Various improvements to gemm by @SamirDroubi in #108
Profile tool by @SamirDroubi in #109
Fix exo_gemm hash based on circular buffer PR by @skeqiqevian in https://github.com/exo-lan...