/* ************************************************************************ * Copyright (C) 2020 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 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_BSRMV_HPP #define TESTING_BSRMV_HPP #include "display.hpp" #include "flops.hpp" #include "gbyte.hpp" #include "hipsparse.hpp" #include "hipsparse_arguments.hpp" #include "hipsparse_graph.hpp" #include "hipsparse_test_unique_ptr.hpp" #include "unit.hpp" #include "utility.hpp" #include #include #include using namespace hipsparse; using namespace hipsparse_test; template void testing_bsrmv_bad_arg(const Arguments& argus) { #if(!defined(CUDART_VERSION)) int safe_size = 100; int safe_dim = 2; T alpha = make_DataType(0.6); T beta = make_DataType(0.2); hipsparseOperation_t transA = HIPSPARSE_OPERATION_NON_TRANSPOSE; hipsparseDirection_t dirA = HIPSPARSE_DIRECTION_COLUMN; hipsparseLocalHandle_t handle; std::unique_ptr unique_ptr_descr(new descr_struct); hipsparseMatDescr_t descr = unique_ptr_descr->descr; auto dptr_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(T) * safe_size), device_free}; auto dx_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * safe_size), device_free}; auto dy_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * safe_size), device_free}; int* dptr = (int*)dptr_managed.get(); int* dcol = (int*)dcol_managed.get(); T* dval = (T*)dval_managed.get(); T* dx = (T*)dx_managed.get(); T* dy = (T*)dy_managed.get(); // Test hipsparseXbsrmv verify_hipsparse_status_invalid_handle(hipsparseXbsrmv((hipsparseHandle_t) nullptr, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, dx, &beta, dy)); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, (T*)nullptr, descr, dval, dptr, dcol, safe_dim, dx, &beta, dy), "Error: alpha is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, (hipsparseMatDescr_t) nullptr, dval, dptr, dcol, safe_dim, dx, &beta, dy), "Error: descr is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, (T*)nullptr, dptr, dcol, safe_dim, dx, &beta, dy), "Error: bsr_val is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, (int*)nullptr, dcol, safe_dim, dx, &beta, dy), "Error: bsr_row_ptr is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, (int*)nullptr, safe_dim, dx, &beta, dy), "Error: bsr_col_ind is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, (T*)nullptr, &beta, dy), "Error: x is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, dx, (T*)nullptr, dy), "Error: beta is nullptr"); verify_hipsparse_status_invalid_pointer(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, dx, &beta, (T*)nullptr), "Error: y is nullptr"); verify_hipsparse_status_invalid_size(hipsparseXbsrmv(handle, dirA, transA, -1, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, dx, &beta, dy), "Error: mb is invalid"); verify_hipsparse_status_invalid_size(hipsparseXbsrmv(handle, dirA, transA, safe_size, -1, safe_size, &alpha, descr, dval, dptr, dcol, safe_dim, dx, &beta, dy), "Error: nb is invalid"); verify_hipsparse_status_invalid_size(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, -1, &alpha, descr, dval, dptr, dcol, safe_dim, dx, &beta, dy), "Error: nnzb is invalid"); verify_hipsparse_status_invalid_size(hipsparseXbsrmv(handle, dirA, transA, safe_size, safe_size, safe_size, &alpha, descr, dval, dptr, dcol, -1, dx, &beta, dy), "Error: block_dim is invalid"); #endif } template void testing_bsrmv(Arguments argus) { int m = argus.M; int n = argus.N; int block_dim = argus.block_dim; T h_alpha = make_DataType(argus.alpha); T h_beta = make_DataType(argus.beta); hipsparseOperation_t transA = argus.transA; hipsparseIndexBase_t idx_base = argus.baseA; hipsparseDirection_t dir = argus.dirA; std::string filename = argus.filename; hipsparseLocalHandle_t handle(argus); std::unique_ptr unique_ptr_descr(new descr_struct); hipsparseMatDescr_t descr = unique_ptr_descr->descr; // Set matrix index base CHECK_HIPSPARSE_ERROR(hipsparseSetMatIndexBase(descr, idx_base)); int mb = (m + block_dim - 1) / block_dim; int nb = (n + block_dim - 1) / block_dim; srand(12345ULL); // Host structures std::vector hcsr_row_ptr; std::vector hcsr_col_ind; std::vector hcsr_val; // Read or construct CSR matrix int nnz = 0; CHECK_GENERATE_MATRIX_ERROR( generate_csr_matrix(filename, m, n, nnz, hcsr_row_ptr, hcsr_col_ind, hcsr_val, idx_base)); mb = (m + block_dim - 1) / block_dim; nb = (n + block_dim - 1) / block_dim; std::vector hx(nb * block_dim); std::vector hy_1(mb * block_dim); std::vector hy_2(mb * block_dim); std::vector hy_gold(mb * block_dim); hipsparseInit(hx, 1, nb * block_dim); hipsparseInit(hy_1, 1, mb * block_dim); // 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 dcsr_row_ptr_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * (m + 1)), device_free}; auto dcsr_col_ind_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * nnz), device_free}; auto dcsr_val_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * nnz), device_free}; auto dbsr_row_ptr_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * (mb + 1)), device_free}; auto dx_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * nb * block_dim), device_free}; auto dy_1_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * mb * block_dim), device_free}; auto dy_2_managed = hipsparse_unique_ptr{device_malloc(sizeof(T) * mb * block_dim), 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}; int* dcsr_row_ptr = (int*)dcsr_row_ptr_managed.get(); int* dcsr_col_ind = (int*)dcsr_col_ind_managed.get(); T* dcsr_val = (T*)dcsr_val_managed.get(); int* dbsr_row_ptr = (int*)dbsr_row_ptr_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(); // copy data from CPU to device CHECK_HIP_ERROR( hipMemcpy(dcsr_row_ptr, hcsr_row_ptr.data(), sizeof(int) * (m + 1), hipMemcpyHostToDevice)); CHECK_HIP_ERROR( hipMemcpy(dcsr_col_ind, hcsr_col_ind.data(), sizeof(int) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dcsr_val, hcsr_val.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(d_alpha, &h_alpha, sizeof(T), hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(d_beta, &h_beta, sizeof(T), hipMemcpyHostToDevice)); // Convert to BSR int nnzb; CHECK_HIPSPARSE_ERROR(hipsparseXcsr2bsrNnz(handle, dir, m, n, descr, dcsr_row_ptr, dcsr_col_ind, block_dim, descr, dbsr_row_ptr, &nnzb)); auto dbsr_col_ind_managed = hipsparse_unique_ptr{device_malloc(sizeof(int) * nnzb), device_free}; auto dbsr_val_managed = hipsparse_unique_ptr{ device_malloc(sizeof(T) * nnzb * block_dim * block_dim), device_free}; int* dbsr_col_ind = (int*)dbsr_col_ind_managed.get(); T* dbsr_val = (T*)dbsr_val_managed.get(); CHECK_HIPSPARSE_ERROR(hipsparseXcsr2bsr(handle, dir, m, n, descr, dcsr_val, dcsr_row_ptr, dcsr_col_ind, block_dim, descr, dbsr_val, dbsr_row_ptr, dbsr_col_ind)); if(argus.unit_check) { CHECK_HIP_ERROR(hipMemcpy(dy_2, hy_2.data(), sizeof(T) * m, hipMemcpyHostToDevice)); // HIPSPARSE pointer mode host CHECK_HIPSPARSE_ERROR(hipsparseSetPointerMode(handle, HIPSPARSE_POINTER_MODE_HOST)); CHECK_HIPSPARSE_ERROR(testing::hipsparseXbsrmv(handle, dir, transA, mb, nb, nnzb, &h_alpha, descr, dbsr_val, dbsr_row_ptr, dbsr_col_ind, block_dim, dx, &h_beta, dy_1)); // HIPSPARSE pointer mode device CHECK_HIPSPARSE_ERROR(hipsparseSetPointerMode(handle, HIPSPARSE_POINTER_MODE_DEVICE)); CHECK_HIPSPARSE_ERROR(testing::hipsparseXbsrmv(handle, dir, transA, mb, nb, nnzb, d_alpha, descr, dbsr_val, dbsr_row_ptr, dbsr_col_ind, block_dim, dx, d_beta, dy_2)); // 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 bsrmv std::vector hbsr_row_ptr(mb + 1); std::vector hbsr_col_ind(nnzb); std::vector hbsr_val(nnzb * block_dim * block_dim); CHECK_HIP_ERROR(hipMemcpy( hbsr_row_ptr.data(), dbsr_row_ptr, sizeof(int) * (mb + 1), hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy( hbsr_col_ind.data(), dbsr_col_ind, sizeof(int) * nnzb, hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy(hbsr_val.data(), dbsr_val, sizeof(T) * nnzb * block_dim * block_dim, hipMemcpyDeviceToHost)); host_bsrmv(dir, transA, mb, nb, nnzb, h_alpha, hbsr_row_ptr.data(), hbsr_col_ind.data(), hbsr_val.data(), block_dim, hx.data(), h_beta, hy_gold.data(), 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()); } if(argus.timing) { int number_cold_calls = 2; int number_hot_calls = argus.iters; CHECK_HIPSPARSE_ERROR(hipsparseSetPointerMode(handle, HIPSPARSE_POINTER_MODE_HOST)); // Warm up for(int iter = 0; iter < number_cold_calls; ++iter) { CHECK_HIPSPARSE_ERROR(testing::hipsparseXbsrmv(handle, dir, transA, mb, nb, nnzb, &h_alpha, descr, dbsr_val, dbsr_row_ptr, dbsr_col_ind, block_dim, dx, &h_beta, dy_1)); } double gpu_time_used = get_time_us(); // Performance run for(int iter = 0; iter < number_hot_calls; ++iter) { CHECK_HIPSPARSE_ERROR(testing::hipsparseXbsrmv(handle, dir, transA, mb, nb, nnzb, &h_alpha, descr, dbsr_val, dbsr_row_ptr, dbsr_col_ind, block_dim, dx, &h_beta, dy_1)); } gpu_time_used = (get_time_us() - gpu_time_used) / number_hot_calls; double gflop_count = spmv_gflop_count(m, nnzb * block_dim * block_dim, h_beta != make_DataType(0.0)); double gbyte_count = bsrmv_gbyte_count(mb, nb, nnzb, block_dim, h_beta != make_DataType(0.0)); double gpu_gflops = get_gpu_gflops(gpu_time_used, gflop_count); double gpu_gbyte = get_gpu_gbyte(gpu_time_used, gbyte_count); display_timing_info(display_key_t::M, m, display_key_t::N, n, display_key_t::block_dim, block_dim, display_key_t::direction, hipsparse_direction2string(dir), display_key_t::alpha, h_alpha, display_key_t::beta, h_beta, display_key_t::gflops, gpu_gflops, display_key_t::bandwidth, gpu_gbyte, display_key_t::time_ms, get_gpu_time_msec(gpu_time_used)); } } #endif // TESTING_BSRMV_HPP