Merge branch 'transpose_opt' of https://github.com/ROCm/composable_kernel into rowwise_opt

4885c38a · aska-0096 · cbf14ee1 · 7c8e92fa · 4885c38a · 4885c38a
Commit 4885c38a authored Sep 03, 2024 by aska-0096
20 changed files
--- a/include/ck_tile/core/numeric/math.hpp
+++ b/include/ck_tile/core/numeric/math.hpp
--- a/include/ck_tile/core/tensor/tile_window.hpp
+++ b/include/ck_tile/core/tensor/tile_window.hpp
--- a/include/ck_tile/core/utility/type_traits.hpp
+++ b/include/ck_tile/core/utility/type_traits.hpp
--- a/include/ck_tile/host.hpp
+++ b/include/ck_tile/host.hpp
--- a/include/ck_tile/host/host_tensor.hpp
+++ b/include/ck_tile/host/host_tensor.hpp
--- a/include/ck_tile/host/kernel_launch.hpp
+++ b/include/ck_tile/host/kernel_launch.hpp
--- a/include/ck_tile/host/reference/reference_batched_rotary_position_embedding.hpp
+++ b/include/ck_tile/host/reference/reference_batched_rotary_position_embedding.hpp
--- a/include/ck_tile/ops/fmha.hpp
+++ b/include/ck_tile/ops/fmha.hpp
--- a/include/ck_tile/ops/fmha/block/block_position_encoding.hpp
+++ b/include/ck_tile/ops/fmha/block/block_position_encoding.hpp
--- a/include/ck_tile/ops/fmha/block/block_rotary_embedding.hpp
+++ b/include/ck_tile/ops/fmha/block/block_rotary_embedding.hpp
--- a/include/ck_tile/ops/fmha/block/page_block_navigator.hpp
+++ b/include/ck_tile/ops/fmha/block/page_block_navigator.hpp
--- a/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp
+++ b/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp
--- a/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_tile_partitioner.hpp
+++ b/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_tile_partitioner.hpp
--- a/include/ck_tile/ops/fmha/kernel/fmha_fwd_splitkv_kernel.hpp
+++ b/include/ck_tile/ops/fmha/kernel/fmha_fwd_splitkv_kernel.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline_default_policy.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_appendkv_pipeline_default_policy.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs_async.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs_async.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs_async_default_policy.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_fwd_splitkv_pipeline_qr_ks_vs_async_default_policy.hpp
--- a/include/ck_tile/ops/fmha/pipeline/block_fmha_pipeline_problem.hpp
+++ b/include/ck_tile/ops/fmha/pipeline/block_fmha_pipeline_problem.hpp