tree 0e74f3d96d131f3e81654846bfdee844f42fca4b
parent 41a23ade9c5857ec56a70e3cb907ffaa9feb6c5e
author Han-Chung Wang <hanhan0912@gmail.com> 1699379151 -0800
committer GitHub <noreply@github.com> 1699379151 -0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJlSnfPCRBK7hj4Ov3rIwAAcWUIAEL+2FWERjXjRnbshGCxY+hP
 ztgmkrhyi5ZpakcYQRObox0BPtlOP8HIHvN/pg/HL7hln9gzMQkNvqdvCj7vS886
 6dEONV3IQpYEIhsJ4DUbFRHWtSoOx9DT/E4csaJy1mlgzv0eLvLpmscvVIqwwlCz
 uh1f7cjYwfmAbaHfA3YG5KemzOUw1d/6Y7IgOPNfuGb8aATQ/wS8HBcZs8xpiESW
 BDuXK46p5AtWXZFfr25ybdX7aqcp8zPAM6pvP9vsgIjrywqGfjf5nlM1/HL2t7wU
 BE8T1gqbVnxRihudIqngU1P33M4Emb6lEc6tawNhpsqKXLRUKwSmWWRH50tUFjA=
 =qw00
 -----END PGP SIGNATURE-----
 

[CPU] Improve tile sizes selection for tensor.pack ops. (#15397)

It disables special vector sizes for non-f32 cases because the logic is
only for 16x16 transpose cases. The improvements of dispatch sizes are
from vectorization. We are not able to vectorize named ops if they have
dynamic shapes, which is fixed by
https://github.com/llvm/llvm-project/commit/03529b99b36788ca836b7ce238ea9400ce89847b.
The change allows backends to vectorize them because they become static
shapes (by tiling with size=1). It is not a hard condition; we track it
in https://github.com/openxla/iree/issues/15441

The revision takes the number of threads into account, so we have better
performance on multi-threaded. It also reduces runtime overheads.

This is a step toward to https://github.com/openxla/iree/issues/15391
and https://github.com/openxla/iree/issues/15349

It improves the performance of
[tensor.pack](https://github.com/openxla/iree/issues/15349) op from 420
ms to 170 ms on 8-threaded x86 CPU.