-
Notifications
You must be signed in to change notification settings - Fork 0
/
cl_reduction_buf.cpp
77 lines (59 loc) · 2.36 KB
/
cl_reduction_buf.cpp
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
// We're using OpenCL C++ API here; there is also C API in <CL/cl.h>
#define CL_HPP_MINIMUM_OPENCL_VERSION 110
#define CL_HPP_TARGET_OPENCL_VERSION 110
#include <CL/opencl.hpp>
// For larger kernels, we can store source in a separate file
static const std::string kernel_source = R"(
__kernel void reduce(__global int* sum, __local int* local_mem) {
// Get work group and work item information
int gsize = get_global_size(0); // global work size
int gid = get_global_id(0); // global work item index
int lsize = get_local_size(0); // local work size
int lid = get_local_id(0); // local work item index
// Store reduced item into local memory
local_mem[lid] = gid; // initialize local memory
barrier(CLK_LOCAL_MEM_FENCE); // synchronize local memory
// Perform reduction across the local work group
for (int s = 1; s < lsize; s *= 2) { // loop over local memory with stride doubling each iteration
if (lid % (2 * s) == 0) {
local_mem[lid] += local_mem[lid + s];
}
barrier(CLK_LOCAL_MEM_FENCE); // synchronize local memory
}
if (lid == 0) { // only one work item per work group
atomic_add(sum, local_mem[0]); // add partial sum to global sum atomically
}
}
)";
int main(int argc, char* argv[]) {
// Initialize OpenCL
cl::Device device = cl::Device::getDefault();
cl::Context context(device);
cl::CommandQueue queue(context, device);
// Print device name
std::string name;
device.getInfo(CL_DEVICE_NAME, &name);
printf("Device: %s\n", name.c_str());
// Compile OpenCL program for found device
cl::Program program(context, kernel_source);
program.build(device);
cl::Kernel kernel_reduce(program, "reduce");
{
// Set problem dimensions
unsigned n = 10;
// Initialize sum variable
int sum = 0;
// Create buffer for sum
cl::Buffer buffer(context, CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR, sizeof(int), &sum);
// Pass arguments to device kernel
kernel_reduce.setArg(0, buffer); // pass buffer to device
kernel_reduce.setArg(1, sizeof(int), NULL); // allocate local memory
// Enqueue kernel
queue.enqueueNDRangeKernel(kernel_reduce, cl::NullRange, cl::NDRange(n), cl::NullRange);
// Read result
queue.enqueueReadBuffer(buffer, CL_TRUE, 0, sizeof(int), &sum);
// Print result
printf("sum = %d\n", sum);
}
return 0;
}