/* Copyright (c) 2022 Advanced Micro Devices, Inc. All rights reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANNTY OF ANY KIND, EXPRESS OR IMPLIED, INNCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANNY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER INN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR INN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include #include "hip/hip_ext.h" #include #include #include #include #if !defined(S_IFREG) && defined(_S_IFREG) #define S_IFREG _S_IFREG #endif struct gridblockDim { unsigned int gridX; unsigned int gridY; unsigned int gridZ; unsigned int blockX; unsigned int blockY; unsigned int blockZ; }; class GraphModuleLaunchKernel { int N = 64; int SIZE = N * N; int *A, *B, *C; hipDeviceptr_t *Ad, *Bd; hipStream_t stream1, stream2; hipModule_t module; hipFunction_t multKernel; struct { void* _Ad; void* _Bd; void* _Cd; int _n; } args1, args2; size_t size1, size2; static constexpr char matmulK[] = "matmulK"; public: GraphModuleLaunchKernel() { allocateMemory(); moduleLoad(); } ~GraphModuleLaunchKernel() { deAllocateMemory(); } void allocateMemory(); void deAllocateMemory(); void moduleLoad(); bool extModuleKernelExecutionMatmul(); bool extModuleKernelExecutionMatmulwithStreamCapture(bool LaunchByDifferentStream = false); static constexpr char fileName[] = "hipMatMul.code"; }; void GraphModuleLaunchKernel::allocateMemory() { A = new int[N * N * sizeof(int)]; B = new int[N * N * sizeof(int)]; for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { A[i * N + j] = 1; B[i * N + j] = 1; } } HIPCHECK(hipStreamCreate(&stream1)); HIPCHECK(hipStreamCreate(&stream2)); HIPCHECK(hipMalloc(reinterpret_cast(&Ad), SIZE * sizeof(int))); HIPCHECK(hipMalloc(reinterpret_cast(&Bd), SIZE * sizeof(int))); HIPCHECK(hipHostMalloc(reinterpret_cast(&C), SIZE * sizeof(int))); HIPCHECK(hipMemcpy(Ad, A, SIZE * sizeof(int), hipMemcpyHostToDevice)); HIPCHECK(hipMemcpy(Bd, B, SIZE * sizeof(int), hipMemcpyHostToDevice)); args1._Ad = Ad; args1._Bd = Bd; args1._Cd = C; args1._n = N; args2._Ad = NULL; args2._Bd = NULL; args2._Cd = NULL; args2._n = 0; size1 = sizeof(args1); size2 = sizeof(args2); } void GraphModuleLaunchKernel::moduleLoad() { HIPCHECK(hipModuleLoad(&module, fileName)); HIPCHECK(hipModuleGetFunction(&multKernel, module, matmulK)); } void GraphModuleLaunchKernel::deAllocateMemory() { HIPCHECK(hipStreamDestroy(stream1)); HIPCHECK(hipStreamDestroy(stream2)); delete[] A; delete[] B; HIPCHECK(hipFree(Ad)); HIPCHECK(hipFree(Bd)); HIPCHECK(hipHostFree(C)); HIPCHECK(hipModuleUnload(module)); } bool GraphModuleLaunchKernel::extModuleKernelExecutionMatmul() { bool testStatus = true; int mismatch = 0; void* config1[] = {HIP_LAUNCH_PARAM_BUFFER_POINTER, &args1, HIP_LAUNCH_PARAM_BUFFER_SIZE, &size1, HIP_LAUNCH_PARAM_END}; HIPCHECK(hipExtModuleLaunchKernel(multKernel, N, N, 1, 32, 32, 1, 0, stream1, NULL, reinterpret_cast(&config1), NULL, NULL, 0)); HIPCHECK(hipStreamSynchronize(stream1)); for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { if (C[i * N + j] != N) mismatch++; } } if (mismatch) { printf("Test failed: the result of matrix multiplications incorrect.\n"); testStatus = false; } return testStatus; } bool GraphModuleLaunchKernel::extModuleKernelExecutionMatmulwithStreamCapture( bool LaunchByDifferentStream) { bool testStatus = true; int mismatch = 0; hipGraph_t graph{nullptr}; hipGraphExec_t graphExec{nullptr}; HIP_CHECK(hipStreamBeginCapture(stream1, hipStreamCaptureModeGlobal)); void* config1[] = {HIP_LAUNCH_PARAM_BUFFER_POINTER, &args1, HIP_LAUNCH_PARAM_BUFFER_SIZE, &size1, HIP_LAUNCH_PARAM_END}; HIPCHECK(hipExtModuleLaunchKernel(multKernel, N, N, 1, 32, 32, 1, 0, stream1, NULL, reinterpret_cast(&config1), NULL, NULL, 0)); HIP_CHECK(hipStreamEndCapture(stream1, &graph)); // Validate end capture is successful REQUIRE(graph != nullptr); HIP_CHECK(hipGraphInstantiate(&graphExec, graph, nullptr, nullptr, 0)); REQUIRE(graphExec != nullptr); // Replay the recorded sequence HIP_CHECK(hipGraphLaunch(graphExec, LaunchByDifferentStream ? stream2 : stream1)); HIP_CHECK(hipStreamSynchronize(LaunchByDifferentStream ? stream2 : stream1)); HIP_CHECK(hipGraphExecDestroy(graphExec)); HIP_CHECK(hipGraphDestroy(graph)); for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { if (C[i * N + j] != N) mismatch++; } } if (mismatch) { printf("Test failed: the result of matrix multiplications incorrect.\n"); testStatus = false; } return testStatus; } TEST_CASE("Unit_hipStreamCapture_ExtModuleLaunchKernel") { struct stat fileStat; if (stat(GraphModuleLaunchKernel::fileName, &fileStat) || !(fileStat.st_mode & S_IFREG)) { FAIL("module file " << GraphModuleLaunchKernel::fileName << " doesn't exist! aborted! \n" << "To generate the file, type\n" << "/opt/rocm/hip/bin/hipcc --cuda-device-only hipMatMul.cc -o hipMatMul.code"); return; } HIPCHECK(hipSetDevice(0)); GraphModuleLaunchKernel kernelLaunch; SECTION("extModuleKernelExecutionMatmul") { REQUIRE(kernelLaunch.extModuleKernelExecutionMatmul()); } SECTION("extModuleKernelExecutionMatmul_withStreamCapture") { REQUIRE(kernelLaunch.extModuleKernelExecutionMatmulwithStreamCapture()); } SECTION("extModuleKernelExecutionMatmul_withStreamCapture_launchByDifferentStream") { REQUIRE(kernelLaunch.extModuleKernelExecutionMatmulwithStreamCapture(true)); } }