[CK_Tile] Fix gemm kernel for 4,64,16 and 64,4,16 warp tile sizes (#2262)

* debugging issue * debugging issue * debugging * debugging * reverting debugging code * clang formatted * updating default_config.json * fix ci failure * clang formatted [ROCm/composable_kernel commit: 59a85cb4bc]
2026-05-14 02:02:46 +00:00 · 2025-06-03 20:16:10 -07:00
parent fd2b0e741a
commit c395db8926
6 changed files with 46 additions and 17 deletions
--- a/example/ck_tile/03_gemm/universal_gemm.cpp
+++ b/example/ck_tile/03_gemm/universal_gemm.cpp
@@ -121,8 +121,11 @@ float gemm_calc(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config&

            if(s.log_level_ > 0)
            {
-                std::cout << "Launching kernel with args:"
-                          << " grid: {" << grids.x << ", " << grids.y << ", " << grids.z << "}"
+                std::cout << "Launching kernel with args: " << Kernel::GetName() << '\n'
+                          << "shape: " << GemmShape::GetName() << '\n'
+                          << "problem: " << GemmPipelineProblem::GetName() << '\n'
+                          << "pipeline: " << GemmPipeline::GetName() << '\n'
+                          << "grid: {" << grids.x << ", " << grids.y << ", " << grids.z << "}"
                          << ", blocks: {" << blocks.x << ", " << blocks.y << ", " << blocks.z
                          << "}" << std::endl;
            }
--- a/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp
+++ b/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp
@@ -92,7 +92,20 @@ struct CShuffleEpilogue
    CK_TILE_HOST_DEVICE static constexpr auto GetVectorSizeC()
    {
        constexpr index_t MaxVectorStoreSize = 16;
-        return MaxVectorStoreSize / sizeof(ODataType);
+        if constexpr(std::is_same_v<CLayout, tensor_layout::gemm::RowMajor>)
+        {
+            return std::min(static_cast<int>(kNPerIteration),
+                            static_cast<int>(MaxVectorStoreSize / sizeof(ODataType)));
+        }
+        else if constexpr(std::is_same_v<CLayout, tensor_layout::gemm::ColumnMajor>)
+        {
+            return std::min(static_cast<int>(kMPerIteration),
+                            static_cast<int>(MaxVectorStoreSize / sizeof(ODataType)));
+        }
+        else
+        {
+            static_assert(false, "Unsupported CLayout!");
+        }
    }

    template <typename Problem>
--- a/include/ck_tile/ops/epilogue/default_2d_epilogue.hpp
+++ b/include/ck_tile/ops/epilogue/default_2d_epilogue.hpp
@@ -149,7 +149,9 @@ struct DefaultGemm2DEpilogue : public Default2DEpilogue<Problem_, Policy_>
            else
            {
                // In this case each thread has just a single item in Ndim
-                return WG::WarpGemmAttribute::Impl::kCNLane / WG::kN;
+                return (WG::WarpGemmAttribute::Impl::kCNLane *
+                        WG::WarpGemmAttribute::Impl::kBNBlock) /
+                       WG::kN;
            }
        }
        // M is contiguous dimension
@@ -158,7 +160,9 @@ struct DefaultGemm2DEpilogue : public Default2DEpilogue<Problem_, Policy_>
            if constexpr(isCTransposed)
            {
                // In this case each thread has just a single item in Mdim
-                return WG::WarpGemmAttribute::Impl::kCNLane / WG::kN;
+                return (WG::WarpGemmAttribute::Impl::kCNLane *
+                        WG::WarpGemmAttribute::Impl::kAMBlock) /
+                       WG::kN;
            }
            else
            {
--- a/tile_engine/ops/gemm/codegen_utils.py
+++ b/tile_engine/ops/gemm/codegen_utils.py
@@ -167,20 +167,20 @@ def BOOL_MAP(b_): return {True: 'true', False: 'false'}[bool(b_)]
 # To Do: add some more supported combinations
 warp_tile_supported_combinations = {
    "gfx90a": {
-        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
-        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
+        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
+        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32]]
    },
    "gfx942": {
-        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
-        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
+        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
+        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32], [16, 16, 32], [16, 16, 64]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32], [16, 16, 64], [16, 16, 32]]
    },
    "gfx950": {
-        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
-        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32]],
+        'fp16_fp16_fp16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
+        'bf16_bf16_bf16': [[32, 32, 8], [16, 16, 16], [32, 32, 16], [16, 16, 32], [4, 64, 16], [64, 4, 16]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32], [16, 16, 32], [16, 16, 64], [16, 16, 128], [32, 32, 64]],
        'fp8_fp8_fp16': [[32, 32, 16], [32, 32, 32], [16, 16, 64], [16, 16, 32], [16, 16, 128], [32, 32, 64]]
    }
--- a/tile_engine/ops/gemm/configs/default_config.json
+++ b/tile_engine/ops/gemm/configs/default_config.json
@@ -48,7 +48,7 @@
            "max": 512,
            "min": 64,
            "step": 64,
-            "exclude": []
+            "exclude": [192]
        },
        "warp_m": {
            "values": [
@@ -71,14 +71,20 @@
        },
        "warp_tile_m": {
            "values": [
+                4,
+                8,
                16,
-                32
+                32,
+                64
            ]
        },
        "warp_tile_n": {
            "values": [
+                4,
+                8,
                16,
-                32
+                32,
+                64
            ]
        },
        "warp_tile_k": {
--- a/tile_engine/ops/gemm/gemm_instance_builder.py
+++ b/tile_engine/ops/gemm/gemm_instance_builder.py
@@ -420,12 +420,12 @@ struct GemmKernel {{

        # LDS capacity verification
        matrix_a_size = (tile_m * tile_k) * \
-            pow(2, element_size(self.config.problem.datatype_map['matrix_a']))
+            element_size(self.config.problem.datatype_map['matrix_a'])
        matrix_b_size = (tile_n * tile_k) * \
-            pow(2, element_size(self.config.problem.datatype_map['matrix_b']))
+            element_size(self.config.problem.datatype_map['matrix_b'])
        total_tile_in_lds = matrix_a_size + matrix_b_size

-        max_tile_size = 2**16 if pipeline == "compv4" else 2**15
+        max_tile_size = 2**15 if pipeline == "compv4" else 2**16
        if total_tile_in_lds > max_tile_size:
            logging.debug(
                f"LDS capacity exceeded [{trait}]: Total required {total_tile_in_lds:,}B ({total_tile_in_lds/1024:.1f}KB) > "
@@ -493,6 +493,9 @@ struct GemmKernel {{
        for trait in self.valid_trait_names:
            tile_valid_params = list(
                filter(lambda t: self.is_tile_valid(t, trait), tile_params))
+
+            # if len(tile_valid_params) == 0:
+            #     raise RuntimeError(f"No valid kernel instance selected for trait: {trait}")
            if trait not in self.valid_trait_tile_combinations:
                self.valid_trait_tile_combinations[trait] = []
            self.valid_trait_tile_combinations[trait].append(tile_valid_params)