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
|
// MIT License
//
// Copyright (c) 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.
// CUB's implementation of single_pass_scan_operators has maybe uninitialized parameters,
// disable the warning because all warnings are threated as errors:
#include "common_benchmark_header.hpp"
// HIP API
#include "hipcub/device/device_for.hpp"
#ifndef DEFAULT_N
const size_t DEFAULT_N = 1024 * 1024 * 32;
#endif
const unsigned int batch_size = 10;
const unsigned int warmup_size = 5;
template<class T>
struct op_t
{
unsigned int* d_count;
HIPCUB_DEVICE
void operator()(T i)
{
// The data is non zero so atomic will never be activated.
if(i == 0)
{
atomicAdd(d_count, 1);
}
}
};
template<class Value>
void run_benchmark(benchmark::State& state, hipStream_t stream, size_t size)
{
using T = Value;
// Generate data
std::vector<T> values_input(size, 4);
T* d_input;
HIP_CHECK(hipMalloc(&d_input, size * sizeof(T)));
HIP_CHECK(hipMemcpy(d_input, values_input.data(), size * sizeof(T), hipMemcpyHostToDevice));
unsigned int* d_count;
HIP_CHECK(hipMalloc(&d_count, sizeof(T)));
HIP_CHECK(hipMemset(d_count, 0, sizeof(T)));
op_t<T> device_op{d_count};
// Warm-up
for(size_t i = 0; i < warmup_size; i++)
{
HIP_CHECK(hipcub::ForEach(d_input, d_input + size, device_op, stream));
}
HIP_CHECK(hipDeviceSynchronize());
for(auto _ : state)
{
auto start = std::chrono::high_resolution_clock::now();
for(size_t i = 0; i < batch_size; i++)
{
HIP_CHECK(hipcub::ForEach(d_input, d_input + size, device_op, stream));
}
HIP_CHECK(hipStreamSynchronize(stream));
auto end = std::chrono::high_resolution_clock::now();
auto elapsed_seconds
= std::chrono::duration_cast<std::chrono::duration<double>>(end - start);
state.SetIterationTime(elapsed_seconds.count());
}
state.SetBytesProcessed(state.iterations() * batch_size * size * sizeof(T));
state.SetItemsProcessed(state.iterations() * batch_size * size);
HIP_CHECK(hipFree(d_count));
HIP_CHECK(hipFree(d_input));
}
#define CREATE_BENCHMARK(Value) \
benchmark::RegisterBenchmark(("for_each<Datatype:" #Value ">"), \
&run_benchmark<Value>, \
stream, \
size)
int main(int argc, char* argv[])
{
cli::Parser parser(argc, argv);
parser.set_optional<size_t>("size", "size", DEFAULT_N, "number of values");
parser.set_optional<int>("trials", "trials", -1, "number of iterations");
parser.run_and_exit_if_error();
// Parse argv
benchmark::Initialize(&argc, argv);
const size_t size = parser.get<size_t>("size");
const int trials = parser.get<int>("trials");
std::cout << "benchmark_device_reduce_by_key" << std::endl;
// HIP
hipStream_t stream = 0; // default
hipDeviceProp_t devProp;
int device_id = 0;
HIP_CHECK(hipGetDevice(&device_id));
HIP_CHECK(hipGetDeviceProperties(&devProp, device_id));
std::cout << "[HIP] Device name: " << devProp.name << std::endl;
using custom_double2 = benchmark_utils::custom_type<double, double>;
// Add benchmarks
std::vector<benchmark::internal::Benchmark*> benchmarks = {
CREATE_BENCHMARK(float),
CREATE_BENCHMARK(double),
CREATE_BENCHMARK(custom_double2),
CREATE_BENCHMARK(int8_t),
CREATE_BENCHMARK(float),
CREATE_BENCHMARK(double),
CREATE_BENCHMARK(long long),
};
// Use manual timing
for(auto& b : benchmarks)
{
b->UseManualTime();
b->Unit(benchmark::kMillisecond);
}
// Force number of iterations
if(trials > 0)
{
for(auto& b : benchmarks)
{
b->Iterations(trials);
}
}
// Run benchmarks
benchmark::RunSpecifiedBenchmarks();
return 0;
}
|
