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
|
/*
* Copyright 2019 Patrick Stotko
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <iostream>
#include <thrust/reduce.h>
#include <thrust/sequence.h>
#include <stdgpu/atomic.cuh> // stdgpu::atomic
#include <stdgpu/iterator.h> // device_begin, device_end
#include <stdgpu/memory.h> // createDeviceArray, destroyDeviceArray
#include <stdgpu/mutex.cuh> // stdgpu::mutex_array
#include <stdgpu/platform.h> // STDGPU_HOST_DEVICE
#include <stdgpu/vector.cuh> // stdgpu::vector
struct is_odd
{
STDGPU_HOST_DEVICE bool
operator()(const int x) const
{
return x % 2 == 1;
}
};
__global__ void
try_partial_sum(const int* d_input, const stdgpu::index_t n, stdgpu::mutex_array<> locks, int* d_result)
{
stdgpu::index_t i = static_cast<stdgpu::index_t>(blockIdx.x * blockDim.x + threadIdx.x);
if (i >= n)
return;
stdgpu::index_t j = i % locks.size();
// While loops might hang due to internal driver scheduling, so use a fixed number of trials.
// Do not loop over try_lock(). Instead, loop over the whole sequential part to avoid deadlocks.
bool finished = false;
const stdgpu::index_t number_trials = 5;
for (stdgpu::index_t k = 0; k < number_trials; ++k)
{
// --- SEQUENTIAL PART ---
if (!finished && locks[j].try_lock())
{
// START --- critical section --- START
d_result[j] += d_input[i];
// END --- critical section --- END
locks[j].unlock();
finished = true;
}
// --- SEQUENTIAL PART ---
}
}
int
main()
{
//
// EXAMPLE DESCRIPTION
// -------------------
// This example demonstrates how stdgpu::mutex_array can be used to implement spin locks on the GPU.
// Since the correct usage still comes with many implications, this example is oversimplified and just shows the
// deadlock-free looping.
//
const stdgpu::index_t n = 100;
const stdgpu::index_t m = 10;
int* d_input = createDeviceArray<int>(n);
int* d_result = createDeviceArray<int>(m);
stdgpu::mutex_array<> locks = stdgpu::mutex_array<>::createDeviceObject(m);
thrust::sequence(stdgpu::device_begin(d_input), stdgpu::device_end(d_input), 1);
// d_input : 1, 2, 3, ..., 100
stdgpu::index_t threads = 32;
stdgpu::index_t blocks = (n + threads - 1) / threads;
try_partial_sum<<<static_cast<unsigned int>(blocks), static_cast<unsigned int>(threads)>>>(d_input,
n,
locks,
d_result);
cudaDeviceSynchronize();
int sum = thrust::reduce(stdgpu::device_cbegin(d_result), stdgpu::device_cend(d_result), 0, thrust::plus<int>());
const int sum_closed_form = n * (n + 1) / 2;
std::cout << "The sum of all partially computed sums (via mutex locks) is " << sum
<< " which intentionally might not match the expected value of " << sum_closed_form << std::endl;
destroyDeviceArray<int>(d_input);
destroyDeviceArray<int>(d_result);
stdgpu::mutex_array<>::destroyDeviceObject(locks);
}
|
