Universal streamk with atomics (#1360)

* universal streamk with atomics with ckprofiler support. grid_size and streamk strategy are tunable. grid_size of -1 leads to #WGs = maximum occupancy X num_CUs. implementation supports many different streamk policies: 1-tile, 2-tile, 3-tile and 4-tile. streamk strategy of -1 leads to default streamk policy (4-tile). 

* Update README.md

* fixing clang-format issues

* removed conflicts in struct members between streamk and universal streamk

* corrected arg parsing for streamk and universal streamk

* added stream-k policies for 3 tile and 4 tile

* fixed argument type issue with parsing cmd args

* changes suggested in PR review are made- removing comments and correcting copyright

* file permissions updated

* added default value support for grid_size and streamk-policy selection set to -1

* print messages for arguments

* print messages for arguments

* print messages for arguments1

example/01_gemm/CMakeLists.txt

@@ -22,6 +22,8 @@ add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16)
 add_example_executable(example_gemm_xdl_fp16_v2 gemm_xdl_fp16_v2.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_v2)
 
+add_example_executable(example_gemm_xdl_fp16_streamk_v3 gemm_xdl_fp16_streamk_v3.cpp)
+add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_streamk_v3)
 add_example_executable(example_gemm_xdl_fp16_v3 gemm_xdl_fp16_v3.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_v3)
 add_example_executable(example_gemm_xdl_fp8_v3 gemm_xdl_fp8_v3.cpp)

example/01_gemm/README.md

@@ -7,3 +7,21 @@
 #arg3: run kernel # of times (>1)
 ./bin/example_gemm_xdl 0 1 5
 ```
+
+# Instructions for ```example_gemm_xdl_fp16_streamk_v3```
+
+## Run ```example_gemm_xdl_fp16_streamk_v3```
+```bash
+arg1: verification (0=no, 1=yes)
+arg2: initialization (0=no init, 1=integer value, 2=decimal value)
+arg3: time kernel (0=no, 1=yes)
+arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideC
+arg10: stream-k select (-1: default config, 0: all DP, 1: 1-tile SK, 2: 2-tile SK)
+arg11: Grid_size(-1 for max occupancy)
+bin/example_gemm_xdl_fp16_streamk_v3 1 2 1 3840 4096 4096 4096 4096 4096 1 -1
+a_m_k: dim 2, lengths {3840, 4096}, strides {4096, 1}
+b_k_n: dim 2, lengths {4096, 4096}, strides {4096, 1}
+c_m_n: dim 2, lengths {3840, 4096}, strides {4096, 1}
+problem {M:3840, N:4096, K:4096, SA:4096, SB:4096, SC:4096, MP:4032, NP:4096, KRead:4096, KP:4096, AK0:512, BK0:2048, MBlock: 18, NBlock: 16, Stream-K Selection:1, Grid size:-1}
+Perf: 0.292022 ms, 441.23 TFlops, 330.348 GB/s, DeviceGemmXdlUniversal<MNPadding, RRR> BlkSize: 256, BlkTile: 224x256x64, WaveTile: 16x16, WaveMap: 7x8, VmemReadVec: 8x8, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v3, BlkGemmPipelinePrefetchStages: 2
+```

example/01_gemm/common.hpp

@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -45,6 +45,19 @@ struct ProblemSizeStreamK final
 
     ck::index_t NumSKBlocks = -1;
 };
+struct ProblemSizeStreamK_universal final
+{
+    ck::index_t M = 3840;
+    ck::index_t N = 4096;
+    ck::index_t K = 4096;
+
+    ck::index_t StrideA = 4096;
+    ck::index_t StrideB = 4096;
+    ck::index_t StrideC = 4096;
+
+    ck::index_t Grid_size   = -1; // defaults to max occupancy
+    ck::index_t Streamk_sel = 1;  // defaults to 1-tile SK
+};
 
 struct ProblemSizeSplitK final
 {
@@ -123,6 +136,57 @@ bool parse_cmd_args<ProblemSize>(int argc,
     return true;
 }
 
+template <>
+bool parse_cmd_args<ProblemSizeStreamK_universal>(int argc,
+                                                  char* argv[],
+                                                  ProblemSizeStreamK_universal& problem_size,
+                                                  ExecutionConfig& config)
+{
+    if(argc == 1)
+    {
+        // use default case
+    }
+    else if(argc == 4)
+    {
+        config.do_verification = std::stoi(argv[1]);
+        config.init_method     = std::stoi(argv[2]);
+        config.time_kernel     = std::stoi(argv[3]);
+    }
+    else if(argc >= 10)
+    {
+        config.do_verification = std::stoi(argv[1]);
+        config.init_method     = std::stoi(argv[2]);
+        config.time_kernel     = std::stoi(argv[3]);
+
+        problem_size.M = std::stoi(argv[4]);
+        problem_size.N = std::stoi(argv[5]);
+        problem_size.K = std::stoi(argv[6]);
+
+        problem_size.StrideA = std::stoi(argv[7]);
+        problem_size.StrideB = std::stoi(argv[8]);
+        problem_size.StrideC = std::stoi(argv[9]);
+
+        if(argc >= 11)
+        {
+            problem_size.Streamk_sel = std::stoi(argv[10]);
+            problem_size.Grid_size   = std::stoi(argv[11]);
+        }
+    }
+    else
+    {
+        std::cerr
+            << "arg1: verification (0=no, 1=yes)" << std::endl
+            << "arg2: initialization (0=no init, 1=integer value, 2=decimal value)" << std::endl
+            << "arg3: time kernel (0=no, 1=yes)" << std::endl
+            << "arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideC" << std::endl
+            << "arg10: stream-k select (-1: default config, 0: all DP, 1: 1-tile SK, 2: 2-tile SK)"
+            << "\narg11: Grid_size(-1 for max occupancy)" << std::endl;
+        return false;
+    }
+
+    return true;
+}
+
 template <>
 bool parse_cmd_args<ProblemSizeStreamK>(int argc,
                                         char* argv[],
@@ -165,7 +229,8 @@ bool parse_cmd_args<ProblemSizeStreamK>(int argc,
                   << std::endl
                   << "arg3: time kernel (0=no, 1=yes)" << std::endl
                   << "arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideC" << std::endl
-                  << "arg10: NumSKBlocks(optional)" << std::endl;
+                  << "arg10: stream-k select (0: all DP, 1: 1-tile SK, 2: 2-tile SK)"
+                  << "\narg11: Grid_size(-1 for max occupancy)" << std::endl;
         return false;
     }
 

example/01_gemm/gemm_xdl_fp16_streamk_v3.cpp

@@ -0,0 +1,48 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_streamk_v3.hpp"
+
+using ADataType        = ck::half_t;
+using BDataType        = ck::half_t;
+using AccDataType      = float;
+using CShuffleDataType = ck::half_t;
+using CDataType        = ck::half_t;
+
+using ALayout = Row;
+using BLayout = Row;
+using CLayout = Row;
+
+using AElementOp = PassThrough;
+using BElementOp = PassThrough;
+using CElementOp = PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::MNPadding;
+
+// clang-format off
+using DeviceGemmV2_Streamk_Instance = 
+    ck::tensor_operation::device::DeviceGemm_Xdl_CShuffle_Streamk_V3<
+        ALayout,   BLayout,  CLayout,   
+        ADataType,   BDataType,  CDataType,  AccDataType,  CShuffleDataType, 
+        PassThrough, PassThrough, PassThrough, GemmDefault, 
+        256,
+        224, 256, 
+        64, 8, 2,
+        16,   16,
+        7,    8,
+        S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 
+        2, 8, 8, 0,
+        S<8, 32, 1>,  S<0, 2, 1>,  S<0, 2, 1>, 
+        1, 8, 2, 0,
+        1, 2, S<1, 32, 1, 8>, 8,
+        ck::BlockGemmPipelineScheduler::Intrawave,ck::BlockGemmPipelineVersion::v3>;
+// clang-format on
+
+using ReferenceGemmInstance = ck::tensor_operation::host::
+    ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
+
+#include "run_gemm_example_streamk_v2.inc"
+
+int main(int argc, char* argv[]) { return !run_gemm_universal_streamk_example(argc, argv); }