Philippe Tillet 4215086931 [BACKEND] no longer uses shared mem or barriers for single-warp reductions (#1915) ...

0-bytes shared mem buffers don't materialize empty allocation buffers;
this could lead to unnecessary barriers.

note: reduceop code has become quite messy and will require some cleanup

2023-08-15 11:51:20 +00:00

..

Analysis

[BACKEND] no longer uses shared mem or barriers for single-warp reductions (#1915)

2023-08-15 11:51:20 +00:00

Conversion

[BACKEND] no longer uses shared mem or barriers for single-warp reductions (#1915)

2023-08-15 11:51:20 +00:00

Dialect

Fix FA v2 hanging issue when BLOCK_N=32 (#274)

2023-08-10 12:52:09 -05:00

rocm

get Arch Info using HSA

2023-04-03 13:58:02 -05:00

Target

Use optimal number of VGPRs (#281)

2023-08-04 10:46:53 -05:00

CMakeLists.txt

Merge triton-mlir branch - Complete rewrite of the backend from scratch (#1004)

2022-12-21 01:30:50 -08:00