Implement grouped gemm tile loop for RDNA4 (#3304)

* feat: grouped gemm tile loop support for RDNA4

* fix: removed extra parameter from grouped gemm example instance

* fix: FP8 check incorrectly enabling FP8 on RDNA3

Author: ErwinTerpstra

example/15_grouped_gemm/CMakeLists.txt

@@ -44,6 +44,9 @@ add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_wmma_spl
 add_example_executable(example_grouped_gemm_wmma_splitk_bf16 grouped_gemm_wmma_splitk_bf16.cpp)
 add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_wmma_splitk_bf16)
 
+add_example_executable(example_grouped_gemm_multiple_d_wmma_fp16 grouped_gemm_multiple_d_wmma_fp16.cpp)
+add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_multiple_d_wmma_fp16)
+
 list(APPEND gpu_list_tf32 gfx942 gfx950)
 set(target 0)
 foreach(gpu IN LISTS GPU_TARGETS)

example/15_grouped_gemm/grouped_gemm_multiple_d_wmma_fp16.cpp

@@ -0,0 +1,76 @@
+// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
+// SPDX-License-Identifier: MIT
+
+#include <iostream>
+#include <numeric>
+#include <initializer_list>
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_wmma_cshuffle_tile_loop_v3.hpp"
+#include "ck/tensor_operation/gpu/device/device_grouped_gemm_tile_loop.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include <ck/utility/data_type.hpp>
+#include <ck/utility/tuple.hpp>
+
+#include "ck/library/utility/check_err.hpp"
+#include "ck/library/utility/device_memory.hpp"
+#include "ck/library/utility/host_tensor.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_gemm_multiple_d.hpp"
+
+using ::ck::DeviceMem;
+using ::ck::hip_check_error;
+using ::ck::HostTensorDescriptor;
+using ::ck::Tensor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+using AddAdd      = ck::tensor_operation::element_wise::AddAdd;
+
+using ADataType        = F16;
+using BDataType        = F16;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using DDataType        = F16;
+using DsDataType       = ck::Tuple<DDataType, DDataType>;
+using EDataType        = F16;
+
+using ALayout  = Row;
+using BLayout  = Col;
+using DLayout  = Row;
+using DsLayout = ck::Tuple<DLayout, DLayout>;
+using ELayout  = Row;
+
+using AElementOp   = PassThrough;
+using BElementOp   = PassThrough;
+using CDEElementOp = AddAdd;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr int NumDs     = 2;
+
+using DeviceGemmInstance =
+    ck::tensor_operation::device::DeviceGroupedGemmMultipleD_Wmma_CShuffle_TileLoop_V3
+    // clang-format off
+//######| ALayout| BLayout| DsLayout| ELayout|     AData|     BData|     AccData|         CShuffle|     DsData|     EData|           A|           B|          CDE|           GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MRepeat| NRepeat|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+//######|        |        |         |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Spacialization|  Size| Block| Block| Block|    |    | Wmma| Wmma|        |        |   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|     MRepeat|     NRepeat|              _MBlock_MRepeat| ScalarPerVector|
+//######|        |        |         |        |          |          |            |                 |           |          |   Operation|   Operation|    Operation|               |      |      |      |      |    |    |     |     |        |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|              _NBlock_NRepeat|        _NRepeat|
+//######|        |        |         |        |          |          |            |                 |           |          |            |            |             |               |      |      |      |      |    |    |     |     |        |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        < ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,   256,   128,   128,    64,   8,   8,   16,   16,       2,       4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 64, 1, 4>,      S<4, 4, 4>>;
+// clang-format on
+
+#include "run_grouped_gemm_multiple_d_example.inc"
+
+int main(int argc, char* argv[]) { return !run_grouped_gemm_example(argc, argv); }