Yuheng XIE 1e093fbfff [OPTIMIZER] Calculate a proper divisibility for ExpandDims (#2397) ...

Previously ExpandDims always inserts 1 as the new divisibility, which
makes writing (x * stride)[:, None] far more slower than (x[:, None] *
stride). A better divisibility can be afforded by computing the GCD of
the old dims. Now the two code above are equally fast. E.g. the conv
inductor in pytorch may be faster.

---------

Co-authored-by: Yuheng XIE <thinelephant@gmail.com>

2023-09-27 23:10:01 -07:00

..

Alias.cpp

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

Allocation.cpp

[BACKEND] Support MMA V3 with register operand (#2375)

2023-09-25 10:43:54 -07:00

AxisInfo.cpp

[OPTIMIZER] Calculate a proper divisibility for ExpandDims (#2397)

2023-09-27 23:10:01 -07:00

CMakeLists.txt

[BUILD] Fix few dependencies and layering issues to make lld work (#2307)

2023-09-15 17:00:58 +00:00

Membar.cpp

[HOPPER][WS] fix missing WS attrs when lowering to llvm (#2063)

2023-08-09 15:45:44 +08:00

Utility.cpp

[BACKEND] Support MMA V3 with register operand (#2375)

2023-09-25 10:43:54 -07:00