Extend XDL kernel to Support RDNA3/4 - Part 4 (#2724)

* Fix example

* fix build error

* update pk_i4 & moe test case

* fix all instance build (examples)

* fix batched_gemm_gemm (example)

* disable example_gemm_bias_softmax_gemm_permute on gfx11

* remove unnecessary disable gfx11

* update tests

* update tests2

[ROCm/composable_kernel commit: 321627aec5]
This commit is contained in:
linqunAMD
2025-09-12 23:17:07 +08:00
committed by GitHub
parent adc66b9b0e
commit 07def6b13d
123 changed files with 848 additions and 574 deletions

View File

@@ -54,10 +54,10 @@ using DeviceOpInstance =
32,
8,
8,
32,
32,
16,
16,
8,
4,
2,
S<4, 64, 1>,
S<1, 0, 2>,
S<1, 0, 2>,
@@ -75,7 +75,7 @@ using DeviceOpInstance =
1,
1,
S<1, 32, 1, 8>,
8>;
4>;
#include "run_gemm_add_example_xdl.inc"

View File

@@ -54,10 +54,10 @@ using DeviceOpInstance =
32,
8,
8,
32,
32,
16,
16,
8,
4,
2,
S<4, 64, 1>,
S<1, 0, 2>,
S<1, 0, 2>,
@@ -75,7 +75,7 @@ using DeviceOpInstance =
1,
1,
S<1, 32, 1, 8>,
8>;
4>;
#include "run_gemm_add_example_xdl.inc"