高通平台下安卓opencl小例子

http://blog.csdn.net/wcj0626/article/details/26272019

 

先到高通的qdn下载adreno GPU SDK，里面有OpenCL的例子。
https://developer.qualcomm.com/software/adreno-gpu-sdk

例子在以下目录：
AdrenoSDK\Development\Samples\OpenCL

 

1、从高通官网下载Adreon SDK（需要注册为会员）

2、解压以后，把文件夹Development/Inc/内的CL文件夹拷贝到：Android-ndk-r9d/platforms/android-19/arch-arm/usr/include/（请自行选择拷贝到哪个版本下）

3、连接手机到电脑，在终端运行：adb pull /system/vendor/lib/libOpenCL.so,此时会把手机里的libOpenCL.so拷贝到电脑当前文件夹下，把该文件拷贝到：android-ndk-r9d/platforms/android-19/arch-arm/usr/lib/（请自行选择拷贝到哪个版本下）

4、新建安卓工程，NDK开发。不知道NDK怎么弄，请移步这里

5、Java代码：

 

public class MainActivity extends ActionBarActivity {

    static
    {
        System.loadLibrary("ocl");
    }
    public native String  testopencl();
    public native String getPlatformName();
    public native String getDeviceName();
    @Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.activity_main);

      TextView testView2=(TextView)findViewById(R.id.textView2);
      TextView testView4=(TextView)findViewById(R.id.textView4);
      TextView testView6=(TextView)findViewById(R.id.textView6);
      testView6.setText(testopencl());
      testView2.setText(getPlatformName());
       testView4.setText(getDeviceName());

    }



}

6、NDK代码：ocl.cpp文件

 

 

#include
#include
#include
#include
#include
#include"com_example_ocl_MainActivity.h"
#define LEN(arr) sizeof(arr) / sizeof(arr[0])
#define N 1024
#define NUM_THREAD 128

cl_uint num_device;
cl_uint num_platform;
cl_platform_id *platform;
cl_device_id *devices;
cl_int err;
cl_context context;
 cl_command_queue cmdQueue;
 cl_mem buffer,sum_buffer;
 cl_program program ;
 cl_kernel kernel;
 const char* src[] = {
         "  __kernel void redution(  \n"
         "  __global int *data,     \n"
         "  __global int *output,   \n"
         "  __local int *data_local   \n"
         "  )  \n"
        " {   \n"
         "  int gid=get_group_id(0);   \n"
         "  int tid=get_global_id(0);    \n"
         "  int size=get_local_size(0);   \n"
         "  int id=get_local_id(0);     \n"
         "  data_local[id]=data[tid];   \n"
         "  barrier(CLK_LOCAL_MEM_FENCE);   \n"
         "  for(int i=size/2;i>0;i>>=1){    \n"
         "      if(id
         "          data_local[id]+=data_local[id+i];   \n"
         "      }   \n"
         "      barrier(CLK_LOCAL_MEM_FENCE);   \n"
         "  }    \n"
         "  if(id==0){    \n"
         "      output[gid]=data_local[0];   \n"
         "  }    \n"
        " }   \n"

 };
int num_block;


int  test()
{
    int* in,*out;
     num_block=N/NUM_THREAD;
    in=(int*)malloc(sizeof(int)*N);
    out=(int*)malloc(sizeof(int)*num_block);
    for(int i=0;i
        in[i]=1;
    }
    Init_OpenCL();
    Context_cmd();
    Create_Buffer(in);
    Create_program();
    Set_arg();
    Execution();
    CopyOutResult(out);
    int sum=0;
    for(int i=0;i
        sum+=out[i];
    }
    return sum;
}

 JNIEXPORT jstring JNICALL Java_com_example_ocl_MainActivity_testopencl (JNIEnv * env, jobject thisobject)
{
    char result[10];
    sprintf(result,"%d\n",test());
     return env->NewStringUTF(result);
}
 JNIEXPORT jstring JNICALL Java_com_example_ocl_MainActivity_getPlatformName(JNIEnv *env , jobject thisobject)
 {
         char buffer[1024];
         clGetPlatformInfo(platform[0],CL_PLATFORM_NAME,sizeof(buffer),buffer,NULL);
        return env->NewStringUTF(buffer);
 }
 JNIEXPORT jstring JNICALL Java_com_example_ocl_MainActivity_getDeviceName(JNIEnv *env , jobject thisobject)
 {

     char buffer[1024];
     clGetDeviceInfo(devices[0],CL_DEVICE_NAME,sizeof(buffer),buffer,NULL);
    return env->NewStringUTF(buffer);
 }

 void Init_OpenCL()
 {
     size_t nameLen1;
     char platformName[1024];

     err = clGetPlatformIDs(0, 0, &num_platform);
     platform=(cl_platform_id*)malloc(sizeof(cl_platform_id)*num_platform);
     err = clGetPlatformIDs(num_platform, platform, NULL);

     err=clGetDeviceIDs(platform[0],CL_DEVICE_TYPE_GPU,0,NULL,&num_device);
     devices=(cl_device_id*)malloc(sizeof(cl_device_id)*num_device);
     err=clGetDeviceIDs(platform[0],CL_DEVICE_TYPE_GPU,num_device,devices,NULL);

 }

 void Context_cmd()
 {
     context=clCreateContext(NULL,num_device,devices,NULL,NULL,&err);
     cmdQueue=clCreateCommandQueue(context,devices[0],0,&err);
 }

 void Create_Buffer(int *data)
 {

     buffer=clCreateBuffer(context,CL_MEM_READ_ONLY|CL_MEM_COPY_HOST_PTR,sizeof(int)*N,data,&err);
     sum_buffer=clCreateBuffer(context,CL_MEM_WRITE_ONLY,sizeof(int)*num_block,0,&err);
 }

 void Create_program()
 {
     program=clCreateProgramWithSource(context, LEN(src), src, NULL, NULL);
     err=clBuildProgram(program,num_device,devices,NULL,NULL,NULL);
     kernel = clCreateKernel(program, "redution", NULL);
 }

void Set_arg()
{
    err=clSetKernelArg(kernel,0,sizeof(cl_mem),&buffer);
    err=clSetKernelArg(kernel,1,sizeof(cl_mem),&sum_buffer);
    err=clSetKernelArg(kernel,2,sizeof(int)*NUM_THREAD,NULL);
}

void Execution()
{
    const size_t globalWorkSize[1]={N};
        const size_t localWorkSize[1]={NUM_THREAD};
      err=clEnqueueNDRangeKernel(cmdQueue,kernel,1,NULL,globalWorkSize,localWorkSize,0,NULL,NULL);
      clFinish(cmdQueue);
}

void CopyOutResult(int*out)
{
    err=clEnqueueReadBuffer(cmdQueue,sum_buffer,CL_TRUE,0,sizeof(int)*num_block,out,0,NULL,NULL);
}

7、Android.mk代码：

 

 

LOCAL_PATH := $(call my-dir)

include $(CLEAR_VARS)
LOCAL_MODULE    := ocl
LOCAL_SRC_FILES := ocl.cpp
LOCAL_LDFLAGS += -llog  -lOpenCL
include $(BUILD_SHARED_LIBRARY)

8、执行结果：

 

搞定！！！