OpenCL Wrapper like CUDA

Usage

include

#include "header/my_clo.hpp"

instatiation

clLikeCUDA clo;///platform_id = 0, device_id = 0
clLikeCUDA clo(deviceNumber);
clLikeCUDA clo(-1);///stdio selection mode

build OpenCL source

cl_kernel kernel1 = clo.clCreateKernelFromFile(fileName, functionName, options);

malloc device memory

cl_mem_flag -> https://www.khronos.org/registry/OpenCL/sdk/1.0/docs/man/xhtml/clCreateBuffer.html

cl_mem *d_a;
clo.clMalloc(&d_a,sizeof(int)*1024*1024,CL_MEM_READ_ONLY);

set arguments

#if __cplusplus >= 201103L  //-std=c++11
    clo.clSetKernelArgs(kernel1, d_a, d_b);
#else
    clSetKernelArg(kernel1,0,sizeof(cl_mem),d_a);
    clSetKernelArg(kernel1,1,sizeof(cl_mem),d_b);
#endif

memory copy

• clMemcpyHostToDevice device -> host
• clMemcpyDeviceToHost host -> device
• clMemcpyDeivceToDevice device -> device

clo.clMemcpy(d_a,h_a,sizeof(int)*1024*1024,clMemcpyHostToDevice);

run kernel

cldim3 blocks(64,64);
cldim3 threads(16,16);

cl_int err = clo.runkernel(kernel1,blocks,threads);

Compile

g++ -o clcheat clcheat.cpp -std=c++11 -lOpenCL
clang++ -o clcheat clcheat.cpp -std=c++11 -framework OpenCL

Run

$./clcheat
[No.] platform = "platformName", device = "deviceName"
----------------------------------------------------------
[  0] platform = "NVIDIA CUDA", device = "GeForce GTX 750 Ti"
[  1] platform = "Intel(R) OpenCL", device = "Intel(R) Core(TM) i5-4440 CPU @ 3.10GHz"
----------------------------------------------------------
input device number : 0
----------------------------------------------------------
CL_PLATFORM_NAME,NVIDIA CUDA,CL_DEVICE_NAME,GeForce GTX 750 Ti
--------------------------------------
[ Matrix Multiplication(1024 x 1024) ]
host->device,  3.799,ms
kernel      , 12.602,ms
device->host,  0.859,ms