/*! \file */ /* ************************************************************************ * Copyright (c) 2020-2021 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 "auto_testing_bad_arg.hpp" #include "testing.hpp" template void testing_prune_csr2csr_bad_arg(const Arguments& arg) { static const size_t safe_size = 100; static rocsparse_int h_nnz_total_dev_host_ptr = 100; static size_t h_buffer_size = 100; static T h_threshold = static_cast(1); // Create rocsparse handle rocsparse_local_handle local_handle; // Create descriptors rocsparse_local_mat_descr local_csr_descr_A; rocsparse_local_mat_descr local_csr_descr_C; rocsparse_handle handle = local_handle; rocsparse_int m = safe_size; rocsparse_int n = safe_size; rocsparse_int nnz_A = safe_size; const rocsparse_mat_descr csr_descr_A = local_csr_descr_A; const T* csr_val_A = (const T*)0x4; const rocsparse_int* csr_row_ptr_A = (const rocsparse_int*)0x4; const rocsparse_int* csr_col_ind_A = (const rocsparse_int*)0x4; const rocsparse_mat_descr csr_descr_C = local_csr_descr_C; T* csr_val_C = (T*)0x4; rocsparse_int* csr_row_ptr_C = (rocsparse_int*)0x4; rocsparse_int* csr_col_ind_C = (rocsparse_int*)0x4; rocsparse_int* nnz_total_dev_host_ptr = &h_nnz_total_dev_host_ptr; size_t* buffer_size = &h_buffer_size; void* temp_buffer = (void*)0x4; int nargs_to_exclude = 1; const int args_to_exclude[1] = {8}; const T* threshold = &h_threshold; #define PARAMS_BUFFER_SIZE \ handle, m, n, nnz_A, csr_descr_A, csr_val_A, csr_row_ptr_A, csr_col_ind_A, threshold, \ csr_descr_C, csr_val_C, csr_row_ptr_C, csr_col_ind_C, buffer_size #define PARAMS_NNZ \ handle, m, n, nnz_A, csr_descr_A, csr_val_A, csr_row_ptr_A, csr_col_ind_A, threshold, \ csr_descr_C, csr_row_ptr_C, nnz_total_dev_host_ptr, temp_buffer #define PARAMS \ handle, m, n, nnz_A, csr_descr_A, csr_val_A, csr_row_ptr_A, csr_col_ind_A, threshold, \ csr_descr_C, csr_val_C, csr_row_ptr_C, csr_col_ind_C, temp_buffer auto_testing_bad_arg(rocsparse_prune_csr2csr_buffer_size, nargs_to_exclude, args_to_exclude, PARAMS_BUFFER_SIZE); auto_testing_bad_arg( rocsparse_prune_csr2csr_nnz, nargs_to_exclude, args_to_exclude, PARAMS_NNZ); auto_testing_bad_arg(rocsparse_prune_csr2csr, nargs_to_exclude, args_to_exclude, PARAMS); #undef PARAMS #undef PARAMS_NNZ #undef PARAMS_BUFFER_SIZE } template void testing_prune_csr2csr(const Arguments& arg) { rocsparse_matrix_factory matrix_factory(arg); rocsparse_int M = arg.M; rocsparse_int N = arg.N; T threshold = static_cast(arg.threshold); rocsparse_index_base csr_base_A = arg.baseA; rocsparse_index_base csr_base_C = arg.baseB; // Create rocsparse handle rocsparse_local_handle handle; rocsparse_local_mat_descr csr_descr_A; rocsparse_local_mat_descr csr_descr_C; rocsparse_set_mat_index_base(csr_descr_A, csr_base_A); rocsparse_set_mat_index_base(csr_descr_C, csr_base_C); // Argument sanity check before allocating invalid memory if(M <= 0 || N <= 0) { static const size_t safe_size = 100; EXPECT_ROCSPARSE_STATUS(rocsparse_prune_csr2csr(handle, M, N, safe_size, csr_descr_A, nullptr, nullptr, nullptr, nullptr, csr_descr_C, nullptr, nullptr, nullptr, nullptr), (M < 0 || N < 0) ? rocsparse_status_invalid_size : rocsparse_status_success); return; } // Allocate host memory for output CSR matrix host_vector h_csr_row_ptr_A; host_vector h_csr_col_ind_A; host_vector h_csr_val_A; host_vector h_nnz_total_dev_host_ptr(1); // Generate uncompressed CSR matrix on host (or read from file) rocsparse_int nnz_A = 0; matrix_factory.init_csr(h_csr_row_ptr_A, h_csr_col_ind_A, h_csr_val_A, M, N, nnz_A, csr_base_A); // Allocate device memory for input CSR matrix device_vector d_nnz_total_dev_host_ptr(1); device_vector d_csr_row_ptr_C(M + 1); device_vector d_csr_row_ptr_A(M + 1); device_vector d_csr_col_ind_A(nnz_A); device_vector d_csr_val_A(nnz_A); // Copy data from CPU to device CHECK_HIP_ERROR(hipMemcpy( d_csr_row_ptr_A, h_csr_row_ptr_A, sizeof(rocsparse_int) * (M + 1), hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy( d_csr_col_ind_A, h_csr_col_ind_A, sizeof(rocsparse_int) * nnz_A, hipMemcpyHostToDevice)); CHECK_HIP_ERROR(hipMemcpy(d_csr_val_A, h_csr_val_A, sizeof(T) * nnz_A, hipMemcpyHostToDevice)); size_t buffer_size = 0; CHECK_ROCSPARSE_ERROR(rocsparse_prune_csr2csr_buffer_size(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, &threshold, csr_descr_C, nullptr, d_csr_row_ptr_C, nullptr, &buffer_size)); T* d_temp_buffer = nullptr; CHECK_HIP_ERROR(hipMalloc(&d_temp_buffer, buffer_size)); T* d_threshold = nullptr; CHECK_HIP_ERROR(hipMalloc(&d_threshold, sizeof(T))); CHECK_HIP_ERROR(hipMemcpy(d_threshold, &threshold, sizeof(T), hipMemcpyHostToDevice)); CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host)); CHECK_ROCSPARSE_ERROR(rocsparse_prune_csr2csr_nnz(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, &threshold, csr_descr_C, d_csr_row_ptr_C, h_nnz_total_dev_host_ptr, d_temp_buffer)); CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_device)); CHECK_ROCSPARSE_ERROR(rocsparse_prune_csr2csr_nnz(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, d_threshold, csr_descr_C, d_csr_row_ptr_C, d_nnz_total_dev_host_ptr, d_temp_buffer)); device_vector d_csr_col_ind_C(h_nnz_total_dev_host_ptr[0]); device_vector d_csr_val_C(h_nnz_total_dev_host_ptr[0]); if(arg.unit_check) { host_vector h_nnz_total_copied_from_device(1); CHECK_HIP_ERROR(hipMemcpy(h_nnz_total_copied_from_device, d_nnz_total_dev_host_ptr, sizeof(rocsparse_int), hipMemcpyDeviceToHost)); h_nnz_total_dev_host_ptr.unit_check(h_nnz_total_copied_from_device); CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host)); CHECK_ROCSPARSE_ERROR(rocsparse_prune_csr2csr(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, &threshold, csr_descr_C, d_csr_val_C, d_csr_row_ptr_C, d_csr_col_ind_C, d_temp_buffer)); host_vector h_csr_row_ptr_C(M + 1); host_vector h_csr_col_ind_C(h_nnz_total_dev_host_ptr[0]); host_vector h_csr_val_C(h_nnz_total_dev_host_ptr[0]); // Copy output to host CHECK_HIP_ERROR(hipMemcpy(h_csr_row_ptr_C, d_csr_row_ptr_C, sizeof(rocsparse_int) * (M + 1), hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy(h_csr_col_ind_C, d_csr_col_ind_C, sizeof(rocsparse_int) * h_nnz_total_dev_host_ptr[0], hipMemcpyDeviceToHost)); CHECK_HIP_ERROR(hipMemcpy(h_csr_val_C, d_csr_val_C, sizeof(T) * h_nnz_total_dev_host_ptr[0], hipMemcpyDeviceToHost)); // call host and check results host_vector h_csr_row_ptr_C_cpu; host_vector h_csr_col_ind_C_cpu; host_vector h_csr_val_C_cpu; host_vector h_nnz_C_cpu(1); host_prune_csr_to_csr(M, N, nnz_A, h_csr_row_ptr_A, h_csr_col_ind_A, h_csr_val_A, h_nnz_C_cpu[0], h_csr_row_ptr_C_cpu, h_csr_col_ind_C_cpu, h_csr_val_C_cpu, csr_base_A, csr_base_C, threshold); h_nnz_C_cpu.unit_check(h_nnz_total_dev_host_ptr); h_csr_row_ptr_C_cpu.unit_check(h_csr_row_ptr_C); h_csr_col_ind_C_cpu.unit_check(h_csr_col_ind_C); h_csr_val_C_cpu.unit_check(h_csr_val_C); } 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_prune_csr2csr(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, &threshold, csr_descr_C, d_csr_val_C, d_csr_row_ptr_C, d_csr_col_ind_C, d_temp_buffer)); } double gpu_time_used = get_time_us(); // Performance run for(int iter = 0; iter < number_hot_calls; ++iter) { CHECK_ROCSPARSE_ERROR(rocsparse_prune_csr2csr(handle, M, N, nnz_A, csr_descr_A, d_csr_val_A, d_csr_row_ptr_A, d_csr_col_ind_A, &threshold, csr_descr_C, d_csr_val_C, d_csr_row_ptr_C, d_csr_col_ind_C, d_temp_buffer)); } gpu_time_used = (get_time_us() - gpu_time_used) / number_hot_calls; double gbyte_count = prune_csr2csr_gbyte_count(M, nnz_A, h_nnz_total_dev_host_ptr[0]); double gpu_gbyte = get_gpu_gbyte(gpu_time_used, gbyte_count); display_timing_info("M", M, "N", N, "nnz_A", nnz_A, "nnz_C", h_nnz_total_dev_host_ptr[0], "threshold", threshold, s_timing_info_bandwidth, gpu_gbyte, s_timing_info_time, get_gpu_time_msec(gpu_time_used), "iter", number_hot_calls, "verified", (arg.unit_check ? "yes" : "no")); } CHECK_HIP_ERROR(hipFree(d_temp_buffer)); CHECK_HIP_ERROR(hipFree(d_threshold)); } #define INSTANTIATE(TYPE) \ template void testing_prune_csr2csr_bad_arg(const Arguments& arg); \ template void testing_prune_csr2csr(const Arguments& arg) INSTANTIATE(float); INSTANTIATE(double); // INSTANTIATE(rocsparse_float_complex); // INSTANTIATE(rocsparse_double_complex);