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
|
/* ************************************************************************
* Copyright (C) 2018-2023 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 cop-
* ies 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 IM-
* PLIED, 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 CONNE-
* CTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ************************************************************************ */
#pragma once
#include "cblas_interface.hpp"
#include "norm.hpp"
#include "rocblas.hpp"
#include "rocblas_init.hpp"
#include "rocblas_math.hpp"
#include "rocblas_random.hpp"
#include "rocblas_test.hpp"
#include "rocblas_vector.hpp"
#include "unit.hpp"
#include "utility.hpp"
template <typename T, typename U = T>
void testing_rotg_bad_arg(const Arguments& arg)
{
const bool FORTRAN = arg.fortran;
auto rocblas_rotg_fn = FORTRAN ? rocblas_rotg<T, U, true> : rocblas_rotg<T, U, false>;
rocblas_local_handle handle{arg};
// Allocate device memory
device_vector<T> da(1, 1);
device_vector<T> db(1, 1);
device_vector<U> dc(1, 1);
device_vector<T> ds(1, 1);
// Check device memory allocation
CHECK_DEVICE_ALLOCATION(da.memcheck());
CHECK_DEVICE_ALLOCATION(db.memcheck());
CHECK_DEVICE_ALLOCATION(dc.memcheck());
CHECK_DEVICE_ALLOCATION(ds.memcheck());
EXPECT_ROCBLAS_STATUS((rocblas_rotg_fn(nullptr, da, db, dc, ds)),
rocblas_status_invalid_handle);
EXPECT_ROCBLAS_STATUS((rocblas_rotg_fn(handle, nullptr, db, dc, ds)),
rocblas_status_invalid_pointer);
EXPECT_ROCBLAS_STATUS((rocblas_rotg_fn(handle, da, nullptr, dc, ds)),
rocblas_status_invalid_pointer);
EXPECT_ROCBLAS_STATUS((rocblas_rotg_fn(handle, da, db, nullptr, ds)),
rocblas_status_invalid_pointer);
EXPECT_ROCBLAS_STATUS((rocblas_rotg_fn(handle, da, db, dc, nullptr)),
rocblas_status_invalid_pointer);
}
template <typename T, typename U = T>
void testing_rotg(const Arguments& arg)
{
const bool FORTRAN = arg.fortran;
auto rocblas_rotg_fn = FORTRAN ? rocblas_rotg<T, U, true> : rocblas_rotg<T, U, false>;
rocblas_local_handle handle{arg};
double gpu_time_used, cpu_time_used;
double error_host, error_device;
const U rel_error = std::numeric_limits<U>::epsilon() * 10;
host_vector<T> a(1, 1);
host_vector<T> b(1, 1);
host_vector<U> c(1, 1);
host_vector<T> s(1, 1);
// Initialize data on host memory
a[0] = arg.get_alpha<T>(); // reuse alpha in place of a to keep number of arguments small
b[0] = arg.get_beta<T>(); // reuse beta in place of b to keep number of arguments small
c[0] = U(0);
s[0] = T(0);
// CPU BLAS
host_vector<T> ha_gold = a;
host_vector<T> hb_gold = b;
host_vector<U> hc_gold = c;
host_vector<T> hs_gold = s;
cpu_time_used = get_time_us_no_sync();
cblas_rotg<T, U>(ha_gold, hb_gold, hc_gold, hs_gold);
cpu_time_used = get_time_us_no_sync() - cpu_time_used;
// Test rocblas_pointer_mode_host
if(arg.unit_check || arg.norm_check)
{
// Naming: `h` is in CPU (host) memory(eg ha), `d` is in GPU (device) memory (eg da).
// Allocate host memory
host_vector<T> ha = a;
host_vector<T> hb = b;
host_vector<U> hc = c;
host_vector<T> hs = s;
CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_host));
handle.pre_test(arg);
CHECK_ROCBLAS_ERROR((rocblas_rotg_fn(handle, ha, hb, hc, hs)));
handle.post_test(arg);
if(arg.unit_check)
{
near_check_general<T>(1, 1, 1, ha_gold, ha, rel_error);
near_check_general<T>(1, 1, 1, hb_gold, hb, rel_error);
near_check_general<U>(1, 1, 1, hc_gold, hc, rel_error);
near_check_general<T>(1, 1, 1, hs_gold, hs, rel_error);
}
if(arg.norm_check)
{
error_host = norm_check_general<T>('F', 1, 1, 1, ha_gold, ha);
error_host += norm_check_general<T>('F', 1, 1, 1, hb_gold, hb);
error_host += norm_check_general<U>('F', 1, 1, 1, hc_gold, hc);
error_host += norm_check_general<T>('F', 1, 1, 1, hs_gold, hs);
}
}
// Test rocblas_pointer_mode_device
if(arg.unit_check || arg.norm_check)
{
// Allocate device memory
device_vector<T> da(1, 1);
device_vector<T> db(1, 1);
device_vector<U> dc(1, 1);
device_vector<T> ds(1, 1);
// Check device memory allocation
CHECK_DEVICE_ALLOCATION(da.memcheck());
CHECK_DEVICE_ALLOCATION(db.memcheck());
CHECK_DEVICE_ALLOCATION(dc.memcheck());
CHECK_DEVICE_ALLOCATION(ds.memcheck());
CHECK_DEVICE_ALLOCATION(ds.memcheck());
// Transfer from CPU to GPU
CHECK_HIP_ERROR(da.transfer_from(a));
CHECK_HIP_ERROR(db.transfer_from(b));
CHECK_HIP_ERROR(dc.transfer_from(c));
CHECK_HIP_ERROR(ds.transfer_from(s));
CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_device));
handle.pre_test(arg);
CHECK_ROCBLAS_ERROR((rocblas_rotg_fn(handle, da, db, dc, ds)));
handle.post_test(arg);
host_vector<T> ha(1);
host_vector<T> hb(1);
host_vector<U> hc(1);
host_vector<T> hs(1);
// Transfer from GPU to CPU
CHECK_HIP_ERROR(ha.transfer_from(da));
CHECK_HIP_ERROR(hb.transfer_from(db));
CHECK_HIP_ERROR(hc.transfer_from(dc));
CHECK_HIP_ERROR(hs.transfer_from(ds));
if(arg.unit_check)
{
near_check_general<T>(1, 1, 1, ha_gold, ha, rel_error);
near_check_general<T>(1, 1, 1, hb_gold, hb, rel_error);
near_check_general<U>(1, 1, 1, hc_gold, hc, rel_error);
near_check_general<T>(1, 1, 1, hs_gold, hs, rel_error);
}
if(arg.norm_check)
{
error_device = norm_check_general<T>('F', 1, 1, 1, ha_gold, ha);
error_device += norm_check_general<T>('F', 1, 1, 1, hb_gold, hb);
error_device += norm_check_general<U>('F', 1, 1, 1, hc_gold, hc);
error_device += norm_check_general<T>('F', 1, 1, 1, hs_gold, hs);
}
}
if(arg.timing)
{
int number_cold_calls = arg.cold_iters;
int number_hot_calls = arg.iters;
CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_host));
host_vector<T> ha = a;
host_vector<T> hb = b;
host_vector<U> hc = c;
host_vector<T> hs = s;
for(int iter = 0; iter < number_cold_calls; ++iter)
{
rocblas_rotg_fn(handle, ha, hb, hc, hs);
}
hipStream_t stream;
CHECK_ROCBLAS_ERROR(rocblas_get_stream(handle, &stream));
gpu_time_used = get_time_us_sync(stream); // in microseconds
for(int iter = 0; iter < number_hot_calls; ++iter)
{
ha = a;
hb = b;
hc = c;
hs = s;
rocblas_rotg_fn(handle, ha, hb, hc, hs);
}
gpu_time_used = (get_time_us_sync(stream) - gpu_time_used) / number_hot_calls;
rocblas_cout << "rocblas-us,CPU-us";
if(arg.norm_check)
rocblas_cout << ",norm_error_host_ptr,norm_error_dev_ptr";
rocblas_cout << std::endl;
rocblas_cout << gpu_time_used << "," << cpu_time_used;
if(arg.norm_check)
rocblas_cout << ',' << error_host << ',' << error_device;
rocblas_cout << std::endl;
}
}
|