/*! \file */ /* ************************************************************************ * Copyright (c) 2021 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. * * ************************************************************************ */ #include "testing.hpp" #include "auto_testing_bad_arg.hpp" template void testing_gemvi_bad_arg(const Arguments& arg) { static const size_t safe_size = 100; // Create rocsparse handle rocsparse_local_handle local_handle; T h_alpha = static_cast(1); T h_beta = static_cast(1); rocsparse_handle handle = local_handle; rocsparse_operation trans = rocsparse_operation_none; rocsparse_index_base base = rocsparse_index_base_zero; rocsparse_int m = safe_size; rocsparse_int n = safe_size; rocsparse_int nnz = safe_size; const T* alpha = &h_alpha; const T* A = (const T*)0x4; rocsparse_int lda = safe_size; const rocsparse_int* x_ind = (const rocsparse_int*)0x4; const T* x_val = (const T*)0x4; const T* beta = &h_beta; T* y = (T*)0x4; void* buffer = (void*)0x4; #define PARAMS handle, trans, m, n, alpha, A, lda, nnz, x_val, x_ind, beta, y, base, buffer auto_testing_bad_arg(rocsparse_gemvi, PARAMS); { auto tmp = trans; trans = rocsparse_operation_transpose; EXPECT_ROCSPARSE_STATUS(rocsparse_gemvi(PARAMS), rocsparse_status_not_implemented); trans = tmp; } // nnz cannot be larger than n { auto tmp = nnz; nnz = n + 1; EXPECT_ROCSPARSE_STATUS(rocsparse_gemvi(PARAMS), rocsparse_status_invalid_size); nnz = tmp; } // lda cannot be lesser than m in non transposed case { auto tmp = lda; lda = m - 1; EXPECT_ROCSPARSE_STATUS(rocsparse_gemvi(PARAMS), rocsparse_status_invalid_size); lda = tmp; } // lda cannot be lesser than n in transposed case { auto tmp1 = lda; auto tmp2 = trans; lda = n - 1; trans = rocsparse_operation_transpose; EXPECT_ROCSPARSE_STATUS(rocsparse_gemvi(PARAMS), rocsparse_status_invalid_size); lda = tmp1; trans = tmp2; } } template void testing_gemvi(const Arguments& arg) { rocsparse_int M = arg.M; rocsparse_int N = arg.N; rocsparse_operation trans = arg.transA; rocsparse_index_base base = arg.baseA; T h_alpha = arg.get_alpha(); T h_beta = arg.get_beta(); // Create rocsparse handle rocsparse_local_handle handle; // Vector sparsity of 33% rocsparse_int nnz = N * 0.33; rocsparse_int lda = (trans == rocsparse_operation_none) ? M : N; // Argument sanity check before allocating invalid memory if(M <= 0 || N < 0) { device_vector dy(std::max(100, M)); CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host)); size_t buffer_size; CHECK_ROCSPARSE_ERROR( rocsparse_gemvi_buffer_size(handle, trans, M, N, nnz, &buffer_size)); void* buffer; CHECK_HIP_ERROR(hipMalloc(&buffer, buffer_size)); EXPECT_ROCSPARSE_STATUS(rocsparse_gemvi(handle, trans, M, N, &h_alpha, nullptr, lda, nnz, nullptr, nullptr, &h_beta, dy, base, buffer), (M < 0 || N < 0) ? rocsparse_status_invalid_size : rocsparse_status_success); CHECK_HIP_ERROR(hipFree(buffer)); return; } // Allocate host memory host_vector hA(M * N); host_vector hx_val(nnz); host_vector hx_ind(nnz); host_vector hy_1(M); host_vector hy_2(M); host_vector hy_gold(M); // Initialize data on CPU rocsparse_seedrand(); rocsparse_init_index(hx_ind, nnz, 1, std::max(N, ((rocsparse_int)1))); rocsparse_init(hx_val, 1, nnz, 1); rocsparse_init(hy_1, 1, M, 1); rocsparse_init(hA, M, N, lda, 1); hy_2 = hy_1; hy_gold = hy_1; // Allocate device memory device_vector dx_ind(nnz); device_vector dx_val(nnz); device_vector dA(M * N); device_vector dy_1(M); device_vector dy_2(M); device_vector d_alpha(1); device_vector d_beta(1); // Copy data from CPU to device CHECK_HIP_ERROR(hipMemcpy(dx_ind, hx_ind, sizeof(rocsparse_int) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dx_val, hx_val, sizeof(T) * nnz, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dA, hA, sizeof(T) * M * N, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(dy_1, hy_1, sizeof(T) * M, hipMemcpyHostToDevice)); // Obtain buffer size size_t buffer_size; CHECK_ROCSPARSE_ERROR(rocsparse_gemvi_buffer_size(handle, trans, M, N, nnz, &buffer_size)); void* buffer; CHECK_HIP_ERROR(hipMalloc(&buffer, buffer_size)); if(arg.unit_check) { // Copy data from CPU to device CHECK_HIP_ERROR(hipMemcpy(dy_2, hy_2, 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)); // Pointer mode host CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host)); CHECK_ROCSPARSE_ERROR(rocsparse_gemvi(handle, trans, M, N, &h_alpha, dA, lda, nnz, dx_val, dx_ind, &h_beta, dy_1, base, buffer)); // Pointer mode device CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_device)); CHECK_ROCSPARSE_ERROR(rocsparse_gemvi(handle, trans, M, N, d_alpha, dA, lda, nnz, dx_val, dx_ind, d_beta, dy_2, base, buffer)); // Copy output to host CHECK_HIP_ERROR(hipMemcpy(hy_1, dy_1, sizeof(T) * M, hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy(hy_2, dy_2, sizeof(T) * M, hipMemcpyDeviceToHost)); // CPU gemvi host_gemvi( M, N, h_alpha, hA, lda, nnz, hx_val, hx_ind, h_beta, hy_gold, base); hy_gold.near_check(hy_1); hy_gold.near_check(hy_2); } if(arg.timing) { int number_cold_calls = 2; int number_hot_calls = arg.iters; CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host)); // Warm up for(int iter = 0; iter < number_cold_calls; ++iter) { CHECK_ROCSPARSE_ERROR(rocsparse_gemvi(handle, trans, M, N, &h_alpha, dA, lda, nnz, dx_val, dx_ind, &h_beta, dy_1, base, buffer)); } double gpu_time_used = get_time_us(); // Performance run for(int iter = 0; iter < number_hot_calls; ++iter) { CHECK_ROCSPARSE_ERROR(rocsparse_gemvi(handle, trans, M, N, &h_alpha, dA, lda, nnz, dx_val, dx_ind, &h_beta, dy_1, base, buffer)); } gpu_time_used = (get_time_us() - gpu_time_used) / number_hot_calls; double gpu_gflops = gemvi_gflop_count(M, nnz) / gpu_time_used * 1e6; double gpu_gbyte = gemvi_gbyte_count((trans == rocsparse_operation_none) ? M : N, nnz, h_beta != static_cast(0)) / gpu_time_used * 1e6; display_timing_info("M", M, "N", N, "nnz", nnz, "trans", rocsparse_operation2string(trans), "alpha", h_alpha, "beta", h_beta, s_timing_info_perf, gpu_gflops, s_timing_info_bandwidth, gpu_gbyte, s_timing_info_time, gpu_time_used / 1e3, "iter", number_hot_calls, "verified", (arg.unit_check ? "yes" : "no")); } CHECK_HIP_ERROR(hipFree(buffer)); } #define INSTANTIATE(TYPE) \ template void testing_gemvi_bad_arg(const Arguments& arg); \ template void testing_gemvi(const Arguments& arg) INSTANTIATE(float); INSTANTIATE(double); INSTANTIATE(rocsparse_float_complex); INSTANTIATE(rocsparse_double_complex);