1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309
|
/* ************************************************************************
* Copyright (C) 2021-2024 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.
*
* ************************************************************************ */
#include "testing.hpp"
template <typename I, typename T>
void testing_spsv_coo_bad_arg(const Arguments& arg)
{
I m = 100;
I n = 100;
I nnz = 100;
const T local_alpha = T(0.6);
const T* alpha = &local_alpha;
rocsparse_operation trans = rocsparse_operation_none;
rocsparse_index_base base = rocsparse_index_base_zero;
rocsparse_spsv_alg alg = rocsparse_spsv_alg_default;
// Index and data type
rocsparse_indextype itype = get_indextype<I>();
rocsparse_datatype compute_type = get_datatype<T>();
// Create rocsparse handle
rocsparse_local_handle local_handle;
// SpSV structures
rocsparse_local_spmat local_A(
m, n, nnz, (void*)0x4, (void*)0x4, (void*)0x4, itype, base, compute_type);
rocsparse_local_dnvec local_x(m, (void*)0x4, compute_type);
rocsparse_local_dnvec local_y(m, (void*)0x4, compute_type);
int nargs_to_exclude = 2;
const int args_to_exclude[2] = {9, 10};
rocsparse_handle handle = local_handle;
rocsparse_spmat_descr mat = local_A;
rocsparse_dnvec_descr x = local_x;
rocsparse_dnvec_descr y = local_y;
size_t local_buffer_size = 100;
size_t* buffer_size = &local_buffer_size;
void* temp_buffer = (void*)0x4;
rocsparse_spsv_stage stage;
#define PARAMS_BUFFER_SIZE \
handle, trans, alpha, mat, x, y, compute_type, alg, stage, buffer_size, temp_buffer
#define PARAMS_ANALYSIS \
handle, trans, alpha, mat, x, y, compute_type, alg, stage, buffer_size, temp_buffer
#define PARAMS_SOLVE \
handle, trans, alpha, mat, x, y, compute_type, alg, stage, buffer_size, temp_buffer
stage = rocsparse_spsv_stage_buffer_size;
select_bad_arg_analysis(rocsparse_spsv, nargs_to_exclude, args_to_exclude, PARAMS_BUFFER_SIZE);
stage = rocsparse_spsv_stage_preprocess;
select_bad_arg_analysis(rocsparse_spsv, nargs_to_exclude, args_to_exclude, PARAMS_ANALYSIS);
stage = rocsparse_spsv_stage_compute;
select_bad_arg_analysis(rocsparse_spsv, nargs_to_exclude, args_to_exclude, PARAMS_SOLVE);
#undef PARAMS_BUFFER_SIZE
#undef PARAMS_ANALYSIS
#undef PARAMS_SOLVE
}
template <typename I, typename T>
void testing_spsv_coo(const Arguments& arg)
{
I M = arg.M;
I N = arg.N;
rocsparse_operation trans_A = arg.transA;
rocsparse_index_base base = arg.baseA;
rocsparse_spsv_alg alg = arg.spsv_alg;
rocsparse_diag_type diag = arg.diag;
rocsparse_fill_mode uplo = arg.uplo;
rocsparse_matrix_type matrix_type = arg.matrix_type;
rocsparse_spsv_stage buffersize = rocsparse_spsv_stage_buffer_size;
rocsparse_spsv_stage preprocess = rocsparse_spsv_stage_preprocess;
rocsparse_spsv_stage compute = rocsparse_spsv_stage_compute;
T halpha = arg.get_alpha<T>();
// Index and data type
rocsparse_indextype itype = get_indextype<I>();
rocsparse_datatype ttype = get_datatype<T>();
// Create rocsparse handle
rocsparse_local_handle handle(arg);
rocsparse_matrix_factory<T, I, I> matrix_factory(arg);
// Allocate host memory for matrix
host_vector<I> hcoo_row_ind;
host_vector<I> hcoo_col_ind;
host_vector<T> hcoo_val;
// Sample matrix
int64_t nnz_A;
matrix_factory.init_coo(hcoo_row_ind, hcoo_col_ind, hcoo_val, M, N, nnz_A, base);
// Non-squared matrices are not supported
if(M != N)
{
return;
}
// Allocate host memory for vectors
host_vector<T> hx(M);
host_vector<T> hy_1(M);
host_vector<T> hy_2(M);
host_vector<T> hy_gold(M);
// Initialize data on CPU
rocsparse_init<T>(hx, M, 1, 1);
rocsparse_init<T>(hy_1, M, 1, 1);
hy_2 = hy_1;
hy_gold = hy_1;
// Allocate device memory
device_vector<I> dcoo_row_ind(nnz_A);
device_vector<I> dcoo_col_ind(nnz_A);
device_vector<T> dcoo_val(nnz_A);
device_vector<T> dx(M);
device_vector<T> dy_1(M);
device_vector<T> dy_2(M);
device_vector<T> dalpha(1);
// Copy data from CPU to device
CHECK_HIP_ERROR(
hipMemcpy(dcoo_row_ind, hcoo_row_ind.data(), sizeof(I) * nnz_A, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(
hipMemcpy(dcoo_col_ind, hcoo_col_ind.data(), sizeof(I) * nnz_A, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(hipMemcpy(dcoo_val, hcoo_val.data(), sizeof(T) * nnz_A, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(hipMemcpy(dx, hx, sizeof(T) * M, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(hipMemcpy(dy_1, hy_1, sizeof(T) * M, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(hipMemcpy(dy_2, hy_2, sizeof(T) * M, hipMemcpyHostToDevice));
CHECK_HIP_ERROR(hipMemcpy(dalpha, &halpha, sizeof(T), hipMemcpyHostToDevice));
// Create descriptors
rocsparse_local_spmat A(M, N, nnz_A, dcoo_row_ind, dcoo_col_ind, dcoo_val, itype, base, ttype);
rocsparse_local_dnvec x(M, dx, ttype);
rocsparse_local_dnvec y1(M, dy_1, ttype);
rocsparse_local_dnvec y2(M, dy_2, ttype);
CHECK_ROCSPARSE_ERROR(
rocsparse_spmat_set_attribute(A, rocsparse_spmat_fill_mode, &uplo, sizeof(uplo)));
CHECK_ROCSPARSE_ERROR(
rocsparse_spmat_set_attribute(A, rocsparse_spmat_diag_type, &diag, sizeof(diag)));
CHECK_ROCSPARSE_ERROR(rocsparse_spmat_set_attribute(
A, rocsparse_spmat_matrix_type, &matrix_type, sizeof(matrix_type)));
// Query SpSV buffer
size_t buffer_size;
CHECK_ROCSPARSE_ERROR(rocsparse_spsv(
handle, trans_A, &halpha, A, x, y1, ttype, alg, buffersize, &buffer_size, nullptr));
// Allocate buffer
void* dbuffer;
CHECK_HIP_ERROR(rocsparse_hipMalloc(&dbuffer, buffer_size));
// Perform analysis on host
CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host));
CHECK_ROCSPARSE_ERROR(rocsparse_spsv(
handle, trans_A, &halpha, A, x, y1, ttype, alg, preprocess, nullptr, dbuffer));
// Perform analysis on device
CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_device));
CHECK_ROCSPARSE_ERROR(rocsparse_spsv(
handle, trans_A, dalpha, A, x, y2, ttype, alg, preprocess, nullptr, dbuffer));
if(arg.unit_check)
{
// Solve on host
CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_host));
CHECK_ROCSPARSE_ERROR(testing::rocsparse_spsv(
handle, trans_A, &halpha, A, x, y1, ttype, alg, compute, &buffer_size, dbuffer));
// Solve on device
CHECK_ROCSPARSE_ERROR(rocsparse_set_pointer_mode(handle, rocsparse_pointer_mode_device));
CHECK_ROCSPARSE_ERROR(testing::rocsparse_spsv(
handle, trans_A, dalpha, A, x, y2, ttype, alg, compute, &buffer_size, dbuffer));
CHECK_HIP_ERROR(hipDeviceSynchronize());
if(ROCSPARSE_REPRODUCIBILITY)
{
rocsparse_reproducibility::save(
"Y pointer mode host", dy_1, "Y pointer mode device", dy_2);
}
// 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 coosv
I analysis_pivot = -1;
I solve_pivot = -1;
host_coosv(trans_A,
M,
nnz_A,
halpha,
hcoo_row_ind.data(),
hcoo_col_ind.data(),
hcoo_val.data(),
hx.data(),
hy_gold.data(),
diag,
uplo,
base,
&analysis_pivot,
&solve_pivot);
if(analysis_pivot == -1 && solve_pivot == -1)
{
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_spsv(
handle, trans_A, &halpha, A, x, y1, ttype, alg, compute, &buffer_size, dbuffer));
}
double gpu_time_used = get_time_us();
// Performance run
for(int iter = 0; iter < number_hot_calls; ++iter)
{
CHECK_ROCSPARSE_ERROR(rocsparse_spsv(
handle, trans_A, &halpha, A, x, y1, ttype, alg, compute, &buffer_size, dbuffer));
}
gpu_time_used = (get_time_us() - gpu_time_used) / number_hot_calls;
double gflop_count = spsv_gflop_count(M, nnz_A, diag);
double gpu_gflops = get_gpu_gflops(gpu_time_used, gflop_count);
double gbyte_count = coosv_gbyte_count<T>(M, nnz_A);
double gpu_gbyte = get_gpu_gbyte(gpu_time_used, gbyte_count);
display_timing_info(display_key_t::M,
M,
display_key_t::nnz_A,
nnz_A,
display_key_t::alpha,
halpha,
display_key_t::algorithm,
rocsparse_spsvalg2string(alg),
display_key_t::gflops,
gpu_gflops,
display_key_t::bandwidth,
gpu_gbyte,
display_key_t::time_ms,
get_gpu_time_msec(gpu_time_used));
}
CHECK_HIP_ERROR(rocsparse_hipFree(dbuffer));
}
#define INSTANTIATE(ITYPE, TTYPE) \
template void testing_spsv_coo_bad_arg<ITYPE, TTYPE>(const Arguments& arg); \
template void testing_spsv_coo<ITYPE, TTYPE>(const Arguments& arg)
INSTANTIATE(int32_t, float);
INSTANTIATE(int32_t, double);
INSTANTIATE(int32_t, rocsparse_float_complex);
INSTANTIATE(int32_t, rocsparse_double_complex);
INSTANTIATE(int64_t, float);
INSTANTIATE(int64_t, double);
INSTANTIATE(int64_t, rocsparse_float_complex);
INSTANTIATE(int64_t, rocsparse_double_complex);
void testing_spsv_coo_extra(const Arguments& arg) {}
|