/* ************************************************************************ * Copyright (c) 2020 Advanced Micro Devices, Inc. * * 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 WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN * THE SOFTWARE. * * ************************************************************************ */ #pragma once #ifndef TESTING_SPMV_COO_HPP #define TESTING_SPMV_COO_HPP #include "hipsparse_test_unique_ptr.hpp" #include "unit.hpp" #include "utility.hpp" #include #include #include using namespace hipsparse_test; void testing_spmv_coo_bad_arg(void) { #if(!defined(CUDART_VERSION) || CUDART_VERSION >= 10010) int64_t m = 100; int64_t n = 100; int64_t nnz = 100; int64_t safe_size = 100; float alpha = 0.6; float beta = 0.2; hipsparseOperation_t transA = HIPSPARSE_OPERATION_NON_TRANSPOSE; hipsparseIndexBase_t idxBase = HIPSPARSE_INDEX_BASE_ZERO; hipsparseIndexType_t idxType = HIPSPARSE_INDEX_32I; hipDataType dataType = HIP_R_32F; hipsparseSpMVAlg_t alg = HIPSPARSE_MV_ALG_DEFAULT; std::unique_ptr unique_ptr_handle(new handle_struct); hipsparseHandle_t handle = unique_ptr_handle->handle; auto drow_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * safe_size), device_free}; auto dcol_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * safe_size), device_free}; auto dval_managed = hipsparse_unique_ptr{device_malloc(sizeof(float) * safe_size), device_free}; auto dx_managed = hipsparse_unique_ptr{device_malloc(sizeof(float) * safe_size), device_free}; auto dy_managed = hipsparse_unique_ptr{device_malloc(sizeof(float) * safe_size), device_free}; auto dbuf_managed = hipsparse_unique_ptr{device_malloc(sizeof(char) * safe_size), device_free}; int* drow = (int*)drow_managed.get(); int* dcol = (int*)dcol_managed.get(); float* dval = (float*)dval_managed.get(); float* dx = (float*)dx_managed.get(); float* dy = (float*)dy_managed.get(); void* dbuf = (void*)dbuf_managed.get(); if(!dval || !drow || !dcol || !dx || !dy || !dbuf) { PRINT_IF_HIP_ERROR(hipErrorOutOfMemory); return; } // SpMV structures hipsparseSpMatDescr_t A; hipsparseDnVecDescr_t x, y; size_t bsize; // Create SpMV structures verify_hipsparse_status_success( hipsparseCreateCoo(&A, m, n, nnz, drow, dcol, dval, idxType, idxBase, dataType), "success"); verify_hipsparse_status_success(hipsparseCreateDnVec(&x, n, dx, dataType), "success"); verify_hipsparse_status_success(hipsparseCreateDnVec(&y, m, dy, dataType), "success"); // SpMV buffer verify_hipsparse_status_invalid_handle( hipsparseSpMV_bufferSize(nullptr, transA, &alpha, A, x, &beta, y, dataType, alg, &bsize)); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize(handle, transA, nullptr, A, x, &beta, y, dataType, alg, &bsize), "Error: alpha is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize( handle, transA, &alpha, nullptr, x, &beta, y, dataType, alg, &bsize), "Error: A is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize( handle, transA, &alpha, A, nullptr, &beta, y, dataType, alg, &bsize), "Error: x is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize(handle, transA, &alpha, A, x, nullptr, y, dataType, alg, &bsize), "Error: beta is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize( handle, transA, &alpha, A, x, &beta, nullptr, dataType, alg, &bsize), "Error: y is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV_bufferSize(handle, transA, &alpha, A, x, &beta, y, dataType, alg, nullptr), "Error: bsize is nullptr"); // SpMV verify_hipsparse_status_invalid_handle( hipsparseSpMV(nullptr, transA, &alpha, A, x, &beta, y, dataType, alg, dbuf)); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, nullptr, A, x, &beta, y, dataType, alg, dbuf), "Error: alpha is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, &alpha, nullptr, x, &beta, y, dataType, alg, dbuf), "Error: A is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, &alpha, A, nullptr, &beta, y, dataType, alg, dbuf), "Error: x is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, &alpha, A, x, nullptr, y, dataType, alg, dbuf), "Error: beta is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, &alpha, A, x, &beta, nullptr, dataType, alg, dbuf), "Error: y is nullptr"); verify_hipsparse_status_invalid_pointer( hipsparseSpMV(handle, transA, &alpha, A, x, &beta, nullptr, dataType, alg, nullptr), "Error: dbuf is nullptr"); // Destruct verify_hipsparse_status_success(hipsparseDestroySpMat(A), "success"); verify_hipsparse_status_success(hipsparseDestroyDnVec(x), "success"); verify_hipsparse_status_success(hipsparseDestroyDnVec(y), "success"); #endif } template hipsparseStatus_t testing_spmv_coo(void) { #if(!defined(CUDART_VERSION) || CUDART_VERSION >= 10010) I safe_size = 100; T h_alpha = make_DataType(2.0); T h_beta = make_DataType(1.0); hipsparseOperation_t transA = HIPSPARSE_OPERATION_NON_TRANSPOSE; hipsparseIndexBase_t idx_base = HIPSPARSE_INDEX_BASE_ZERO; hipsparseSpMVAlg_t alg = HIPSPARSE_COOMV_ALG; // Matrices are stored at the same path in matrices directory std::string filename = hipsparse_exepath() + "../matrices/nos3.bin"; // Index and data type hipsparseIndexType_t typeI = (typeid(I) == typeid(int32_t)) ? HIPSPARSE_INDEX_32I : HIPSPARSE_INDEX_64I; hipDataType typeT = (typeid(T) == typeid(float)) ? HIP_R_32F : ((typeid(T) == typeid(double)) ? HIP_R_64F : ((typeid(T) == typeid(hipComplex) ? HIP_C_32F : HIP_C_64F))); // hipSPARSE handle std::unique_ptr test_handle(new handle_struct); hipsparseHandle_t handle = test_handle->handle; // Host structures std::vector hrow_ptr; std::vector hcol_ind; std::vector hval; // Initial Data on CPU srand(12345ULL); I m; I n; I nnz; if(read_bin_matrix(filename.c_str(), m, n, nnz, hrow_ptr, hcol_ind, hval, idx_base) != 0) { fprintf(stderr, "Cannot open [read] %s\n", filename.c_str()); return HIPSPARSE_STATUS_INTERNAL_ERROR; } std::vector hrow_ind(nnz); // Convert to COO for(I i = 0; i < m; ++i) { for(I j = hrow_ptr[i]; j < hrow_ptr[i + 1]; ++j) { hrow_ind[j - idx_base] = i + idx_base; } } std::vector hx(n); std::vector hy_1(m); std::vector hy_2(m); std::vector hy_gold(m); hipsparseInit(hx, 1, n); hipsparseInit(hy_1, 1, m); // copy vector is easy in STL; hy_gold = hx: save a copy in hy_gold which will be output of CPU hy_2 = hy_1; hy_gold = hy_1; // allocate memory on device auto drow_managed = hipsparse_unique_ptr{device_malloc(sizeof(I) * nnz), device_free}; auto dcol_managed = hipsparse_unique_ptr{device_malloc(sizeof(I) * nnz), device_free}; auto dval_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * nnz), device_free}; auto dx_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * n), device_free}; auto dy_1_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * m), device_free}; auto dy_2_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * m), device_free}; auto d_alpha_managed = hipsparse_unique_ptr{device_malloc(sizeof(T)), device_free}; auto d_beta_managed = hipsparse_unique_ptr{device_malloc(sizeof(T)), device_free}; I* drow = (I*)drow_managed.get(); I* dcol = (I*)dcol_managed.get(); T* dval = (T*)dval_managed.get(); T* dx = (T*)dx_managed.get(); T* dy_1 = (T*)dy_1_managed.get(); T* dy_2 = (T*)dy_2_managed.get(); T* d_alpha = (T*)d_alpha_managed.get(); T* d_beta = (T*)d_beta_managed.get(); if(!dval || !drow || !dcol || !dx || !dy_1 || !dy_2 || !d_alpha || !d_beta) { verify_hipsparse_status_success(HIPSPARSE_STATUS_ALLOC_FAILED, "!dval || !drow || !dcol || !dx || " "!dy_1 || !dy_2 || !d_alpha || !d_beta"); return HIPSPARSE_STATUS_ALLOC_FAILED; } // copy data from CPU to device CHECK_HIP_ERROR(hipMemcpy(drow, hrow_ind.data(), sizeof(I) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dcol, hcol_ind.data(), sizeof(I) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dval, hval.data(), sizeof(T) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dx, hx.data(), sizeof(T) * n, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dy_1, hy_1.data(), sizeof(T) * m, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dy_2, hy_2.data(), sizeof(T) * m, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(d_alpha, &h_alpha, sizeof(T), hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(d_beta, &h_beta, sizeof(T), hipMemcpyHostToDevice)); // Create matrices hipsparseSpMatDescr_t A; CHECK_HIPSPARSE_ERROR( hipsparseCreateCoo(&A, m, n, nnz, drow, dcol, dval, typeI, idx_base, typeT)); // Create dense vectors hipsparseDnVecDescr_t x, y1, y2; CHECK_HIPSPARSE_ERROR(hipsparseCreateDnVec(&x, n, dx, typeT)); CHECK_HIPSPARSE_ERROR(hipsparseCreateDnVec(&y1, m, dy_1, typeT)); CHECK_HIPSPARSE_ERROR(hipsparseCreateDnVec(&y2, m, dy_2, typeT)); // Query SpMV buffer size_t bufferSize; CHECK_HIPSPARSE_ERROR(hipsparseSpMV_bufferSize( handle, transA, &h_alpha, A, x, &h_beta, y1, typeT, alg, &bufferSize)); void* buffer; CHECK_HIP_ERROR(hipMalloc(&buffer, bufferSize)); // ROCSPARSE pointer mode host CHECK_HIPSPARSE_ERROR(hipsparseSetPointerMode(handle, HIPSPARSE_POINTER_MODE_HOST)); CHECK_HIPSPARSE_ERROR( hipsparseSpMV(handle, transA, &h_alpha, A, x, &h_beta, y1, typeT, alg, buffer)); // ROCSPARSE pointer mode device CHECK_HIPSPARSE_ERROR(hipsparseSetPointerMode(handle, HIPSPARSE_POINTER_MODE_DEVICE)); CHECK_HIPSPARSE_ERROR( hipsparseSpMV(handle, transA, d_alpha, A, x, d_beta, y2, typeT, alg, buffer)); // copy output from device to CPU CHECK_HIP_ERROR(hipMemcpy(hy_1.data(), dy_1, sizeof(T) * m, hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy(hy_2.data(), dy_2, sizeof(T) * m, hipMemcpyDeviceToHost)); // Host SpMV #ifdef _OPENMP #pragma omp parallel for schedule(dynamic, 1024) #endif for(I i = 0; i < m; ++i) { hy_gold[i] = hy_gold[i] * h_beta; } for(I i = 0; i < nnz; ++i) { hy_gold[hrow_ind[i] - idx_base] = testing_fma( h_alpha * hval[i], hx[hcol_ind[i] - idx_base], hy_gold[hrow_ind[i] - idx_base]); } unit_check_near(1, m, 1, hy_gold.data(), hy_1.data()); unit_check_near(1, m, 1, hy_gold.data(), hy_2.data()); CHECK_HIP_ERROR(hipFree(buffer)); CHECK_HIPSPARSE_ERROR(hipsparseDestroySpMat(A)); CHECK_HIPSPARSE_ERROR(hipsparseDestroyDnVec(x)); CHECK_HIPSPARSE_ERROR(hipsparseDestroyDnVec(y1)); CHECK_HIPSPARSE_ERROR(hipsparseDestroyDnVec(y2)); #endif return HIPSPARSE_STATUS_SUCCESS; } #endif // TESTING_SPMV_COO_HPP