Simply create 2wave pipeline/policy files

include/ck_tile/ops/fmha.hpp

@@ -34,6 +34,8 @@
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_enum.hpp"
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_problem.hpp"
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs.hpp"
+#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_2wave.hpp"
+#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_2wave_default_policy.hpp"
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_async.hpp"
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_async_default_policy.hpp"
 #include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_default_policy.hpp"

include/ck_tile/ops/fmha/pipeline/block_fmha_pipeline_enum.hpp

@@ -11,6 +11,7 @@ enum class BlockFmhaPipelineEnum
     QRKSVS = 0,
     QRKSVS_ASYNC,
     QSKSVS,
+    QRKSVS_2WAVE,
 };
 
 template <BlockFmhaPipelineEnum>
@@ -22,6 +23,11 @@ struct BlockFmhaPipelineEnumToStr<BlockFmhaPipelineEnum::QRKSVS>
     static constexpr const char* name = "qr";
 };
 template <>
+struct BlockFmhaPipelineEnumToStr<BlockFmhaPipelineEnum::QRKSVS_2WAVE>
+{
+    static constexpr const char* name = "qr_2wave";
+};
+template <>
 struct BlockFmhaPipelineEnumToStr<BlockFmhaPipelineEnum::QRKSVS_ASYNC>
 {
     static constexpr const char* name = "qr_async";

include/ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qr_ks_vs_2wave_default_policy.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include "ck_tile/core.hpp"
+#include "ck_tile/ops/fmha/pipeline/block_fmha_pipeline_qx_ks_vs_custom_policy.hpp"
+
+namespace ck_tile {
+
+// This pipeline is qkv all located in LDS
+using BlockFmhaPipelineQRKSVS2WaveDefaultPolicy =
+    BlockFmhaPipelineQXKSVSCustomPolicy</* QLoadOnce = */ true,
+                                        /* AsyncCopyK = */ false,
+                                        /* AsyncCopyV = */ false,
+                                        /* NumPrefetchK = */ 1,
+                                        /* NumPrefetchV = */ 1>;
+
+} // namespace ck_tile