All cache bugs fixed - Handshaking

benchmarks/new_opencl/bfs/CLHelper.h

@@ -0,0 +1,848 @@
+//------------------------------------------
+//--cambine:helper function for OpenCL
+//--programmer:	Jianbin Fang
+//--date:	27/12/2010
+//------------------------------------------
+#ifndef _CL_HELPER_
+#define _CL_HELPER_
+
+#include <CL/cl.h>
+#include <fstream>
+#include <iostream>
+#include <string>
+#include <vector>
+
+
+using std::string;
+using std::ifstream;
+using std::cerr;
+using std::endl;
+using std::cout;
+//#pragma OPENCL EXTENSION cl_nv_compiler_options:enable
+#define WORK_DIM 2 // work-items dimensions
+
+struct oclHandleStruct {
+  cl_context context;
+  cl_device_id *devices;
+  cl_command_queue queue;
+  cl_program program;
+  cl_int cl_status;
+  std::string error_str;
+  std::vector<cl_kernel> kernel;
+};
+
+struct oclHandleStruct oclHandles;
+
+char kernel_file[100] = "Kernels.cl";
+int total_kernels = 2;
+string kernel_names[2] = {"BFS_1", "BFS_2"};
+int work_group_size = 512;
+int device_id_inused = 0; // deviced id used (default : 0)
+
+int read_kernel_file(const char* filename, uint8_t** data, size_t* size) {
+  if (nullptr == filename || nullptr == data || 0 == size)
+    return -1;
+
+  FILE* fp = fopen(filename, "r");
+  if (NULL == fp) {
+    fprintf(stderr, "Failed to load kernel.");
+    return -1;
+  }
+  fseek(fp , 0 , SEEK_END);
+  long fsize = ftell(fp);
+  rewind(fp);
+
+  *data = (uint8_t*)malloc(fsize);
+  *size = fread(*data, 1, fsize, fp);
+  
+  fclose(fp);
+  
+  return 0;
+}
+
+/*
+ * Converts the contents of a file into a string
+ */
+string FileToString(const string fileName) {
+  ifstream f(fileName.c_str(), ifstream::in | ifstream::binary);
+
+  try {
+    size_t size;
+    char *str;
+    string s;
+
+    if (f.is_open()) {
+      size_t fileSize;
+      f.seekg(0, ifstream::end);
+      size = fileSize = f.tellg();
+      f.seekg(0, ifstream::beg);
+
+      str = new char[size + 1];
+      if (!str)
+        throw(string("Could not allocate memory"));
+
+      f.read(str, fileSize);
+      f.close();
+      str[size] = '\0';
+
+      s = str;
+      delete[] str;
+      return s;
+    }
+  } catch (std::string msg) {
+    cerr << "Exception caught in FileToString(): " << msg << endl;
+    if (f.is_open())
+      f.close();
+  } catch (...) {
+    cerr << "Exception caught in FileToString()" << endl;
+    if (f.is_open())
+      f.close();
+  }
+  string errorMsg = "FileToString()::Error: Unable to open file " + fileName;
+  throw(errorMsg);
+}
+//---------------------------------------
+// Read command line parameters
+//
+void _clCmdParams(int argc, char *argv[]) {
+  for (int i = 0; i < argc; ++i) {
+    switch (argv[i][1]) {
+    case 'g': //--g stands for size of work group
+      if (++i < argc) {
+        sscanf(argv[i], "%u", &work_group_size);
+      } else {
+        std::cerr << "Could not read argument after option " << argv[i - 1]
+                  << std::endl;
+        throw;
+      }
+      break;
+    case 'd': //--d stands for device id used in computaion
+      if (++i < argc) {
+        sscanf(argv[i], "%u", &device_id_inused);
+      } else {
+        std::cerr << "Could not read argument after option " << argv[i - 1]
+                  << std::endl;
+        throw;
+      }
+      break;
+    default:;
+    }
+  }
+}
+
+//---------------------------------------
+// Initlize CL objects
+//--description: there are 5 steps to initialize all the OpenCL objects needed
+//--revised on 04/01/2011: get the number of devices  and
+//  devices have no relationship with context
+void _clInit() {
+  printf("_clInit()\n");
+
+  int DEVICE_ID_INUSED = device_id_inused;
+  cl_int resultCL;
+
+  oclHandles.context = NULL;
+  oclHandles.devices = NULL;
+  oclHandles.queue = NULL;
+  oclHandles.program = NULL;
+
+  cl_uint deviceListSize;
+
+  //-----------------------------------------------
+  //--cambine-1: find the available platforms and select one
+
+  cl_uint numPlatforms = 1;
+  cl_platform_id targetPlatform = NULL;
+
+  cl_platform_id *allPlatforms =
+      (cl_platform_id *)malloc(numPlatforms * sizeof(cl_platform_id));
+
+  resultCL = clGetPlatformIDs(numPlatforms, allPlatforms, NULL);
+  if (resultCL != CL_SUCCESS)
+    throw(string("InitCL()::Error: Getting platform ids (clGetPlatformIDs)"));
+
+  // Select the target platform. Default: first platform
+  targetPlatform = allPlatforms[0];
+
+  /*for (int i = 0; i < numPlatforms; i++)
+{
+char pbuff[128];
+resultCL = clGetPlatformInfo( allPlatforms[i],
+                          CL_PLATFORM_VENDOR,
+                          sizeof(pbuff),
+                          pbuff,
+                          NULL);
+if (resultCL != CL_SUCCESS)
+throw (string("InitCL()::Error: Getting platform info (clGetPlatformInfo)"));
+
+  //printf("vedor is %s\n",pbuff);
+
+}
+free(allPlatforms);*/
+
+  //-----------------------------------------------
+  //--cambine-2: create an OpenCL context
+  /*cl_context_properties cprops[3] = { CL_CONTEXT_PLATFORM,
+ (cl_context_properties)targetPlatform, 0 };
+  oclHandles.context = clCreateContextFromType(cprops,
+                                              CL_DEVICE_TYPE_GPU,
+                                              NULL,
+                                              NULL,
+                                              &resultCL);
+
+  if ((resultCL != CL_SUCCESS) || (oclHandles.context == NULL))
+      throw (string("InitCL()::Error: Creating Context
+ (clCreateContextFromType)"));
+
+        //-----------------------------------------------
+  //--cambine-3: detect OpenCL devices
+  // First, get the size of device list
+ oclHandles.cl_status = clGetDeviceIDs(targetPlatform, CL_DEVICE_TYPE_GPU, 0,
+ NULL, &deviceListSize);
+ if(oclHandles.cl_status!=CL_SUCCESS){
+      throw(string("exception in _clInit -> clGetDeviceIDs"));
+ }
+ if (deviceListSize == 0)
+      throw(string("InitCL()::Error: No devices found."));
+
+              printf("OK1()\n");
+
+  //std::cout<<"device number:"<<deviceListSize<<std::endl;*/
+
+  // Now, allocate the device list
+  deviceListSize = 1;
+  oclHandles.devices =
+      (cl_device_id *)malloc(deviceListSize * sizeof(cl_device_id));
+  if (oclHandles.devices == 0)
+    throw(string("InitCL()::Error: Could not allocate memory."));
+
+  //* Next, get the device list data
+  oclHandles.cl_status =
+      clGetDeviceIDs(targetPlatform, CL_DEVICE_TYPE_DEFAULT, deviceListSize,
+                     oclHandles.devices, NULL);
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(string("exception in _clInit -> clGetDeviceIDs-2"));
+  }
+
+  oclHandles.context = clCreateContext(NULL, deviceListSize, oclHandles.devices,
+                                       NULL, NULL, &resultCL);
+  if ((resultCL != CL_SUCCESS) || (oclHandles.context == NULL))
+    throw(string("InitCL()::Error: Creating Context (clCreateContext)"));
+
+  //-----------------------------------------------
+  //--cambine-4: Create an OpenCL command queue
+  oclHandles.queue = clCreateCommandQueue(
+      oclHandles.context, oclHandles.devices[DEVICE_ID_INUSED], 0, &resultCL);
+  printf("resultCL=%d, queue=0x%x\n", resultCL, oclHandles.queue);
+
+  if ((resultCL != CL_SUCCESS) || (oclHandles.queue == NULL))
+    throw(string("InitCL()::Creating Command Queue. (clCreateCommandQueue)"));
+  //-----------------------------------------------
+  //--cambine-5: Load CL file, build CL program object, create CL kernel object
+  /*std::string  source_str = FileToString(kernel_file);
+  const char * source    = source_str.c_str();
+  size_t sourceSize[]    = { source_str.length() };*/
+
+  //oclHandles.program = clCreateProgramWithBuiltInKernels(
+  //    oclHandles.context, 1, &oclHandles.devices[DEVICE_ID_INUSED],
+  //   "BFS_1;BFS_2", &resultCL);
+  /*oclHandles.program = clCreateProgramWithSource(oclHandles.context,
+                                                  1,
+                                                  &source,
+                                                  sourceSize,
+                                                  &resultCL);*/
+  // read kernel binary from file  
+  uint8_t *kernel_bin = NULL;
+  size_t kernel_size;
+  cl_int binary_status = 0;
+  if (0 != read_kernel_file("kernel.pocl", &kernel_bin, &kernel_size))
+    std::abort();
+
+ oclHandles.program = clCreateProgramWithBinary(
+    oclHandles.context, 1, &oclHandles.devices[DEVICE_ID_INUSED], &kernel_size, &kernel_bin, &binary_status, &resultCL);
+  free(kernel_bin);
+
+  if ((resultCL != CL_SUCCESS) || (oclHandles.program == NULL))
+    throw(string("InitCL()::Error: Loading Binary into cl_program. "
+                 "(clCreateProgramWithBinary)"));
+
+  // insert debug information
+  // std::string options= "-cl-nv-verbose"; //Doesn't work on AMD machines
+  // options += " -cl-nv-opt-level=3";
+  resultCL = clBuildProgram(oclHandles.program, deviceListSize,
+                            oclHandles.devices, NULL, NULL, NULL);
+  if ((resultCL != CL_SUCCESS) || (oclHandles.program == NULL)) {
+    cerr << "InitCL()::Error: In clBuildProgram" << endl;
+
+    size_t length;
+    resultCL = clGetProgramBuildInfo(oclHandles.program,
+                                     oclHandles.devices[DEVICE_ID_INUSED],
+                                     CL_PROGRAM_BUILD_LOG, 0, NULL, &length);
+    if (resultCL != CL_SUCCESS)
+      throw(string("InitCL()::Error: Getting Program build "
+                   "info(clGetProgramBuildInfo)"));
+
+    char *buffer = (char *)malloc(length);
+    resultCL = clGetProgramBuildInfo(
+        oclHandles.program, oclHandles.devices[DEVICE_ID_INUSED],
+        CL_PROGRAM_BUILD_LOG, length, buffer, NULL);
+    if (resultCL != CL_SUCCESS)
+      throw(string("InitCL()::Error: Getting Program build "
+                   "info(clGetProgramBuildInfo)"));
+
+    cerr << buffer << endl;
+    free(buffer);
+
+    throw(string("InitCL()::Error: Building Program (clBuildProgram)"));
+  }
+
+// get program information in intermediate representation
+#ifdef PTX_MSG
+  size_t binary_sizes[deviceListSize];
+  char *binaries[deviceListSize];
+  // figure out number of devices and the sizes of the binary for each device.
+  oclHandles.cl_status =
+      clGetProgramInfo(oclHandles.program, CL_PROGRAM_BINARY_SIZES,
+                       sizeof(size_t) * deviceListSize, &binary_sizes, NULL);
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(string("--cambine:exception in _InitCL -> clGetProgramInfo-2"));
+  }
+
+  std::cout << "--cambine:" << binary_sizes << std::endl;
+  // copy over all of the generated binaries.
+  for (int i = 0; i < deviceListSize; i++)
+    binaries[i] = (char *)malloc(sizeof(char) * (binary_sizes[i] + 1));
+  oclHandles.cl_status =
+      clGetProgramInfo(oclHandles.program, CL_PROGRAM_BINARIES,
+                       sizeof(char *) * deviceListSize, binaries, NULL);
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(string("--cambine:exception in _InitCL -> clGetProgramInfo-3"));
+  }
+  for (int i = 0; i < deviceListSize; i++)
+    binaries[i][binary_sizes[i]] = '\0';
+  std::cout << "--cambine:writing ptd information..." << std::endl;
+  FILE *ptx_file = fopen("cl.ptx", "w");
+  if (ptx_file == NULL) {
+    throw(string("exceptions in allocate ptx file."));
+  }
+  fprintf(ptx_file, "%s", binaries[DEVICE_ID_INUSED]);
+  fclose(ptx_file);
+  std::cout << "--cambine:writing ptd information done." << std::endl;
+  for (int i = 0; i < deviceListSize; i++)
+    free(binaries[i]);
+#endif
+
+  for (int nKernel = 0; nKernel < total_kernels; nKernel++) {
+    /* get a kernel object handle for a kernel with the given name */
+    cl_kernel kernel = clCreateKernel(
+        oclHandles.program, (kernel_names[nKernel]).c_str(), &resultCL);
+
+    if ((resultCL != CL_SUCCESS) || (kernel == NULL)) {
+      string errorMsg = "InitCL()::Error: Creating Kernel (clCreateKernel) \"" +
+                        kernel_names[nKernel] + "\"";
+      throw(errorMsg);
+    }
+
+    oclHandles.kernel.push_back(kernel);
+  }
+// get resource alocation information
+#ifdef RES_MSG
+  char *build_log;
+  size_t ret_val_size;
+  oclHandles.cl_status = clGetProgramBuildInfo(
+      oclHandles.program, oclHandles.devices[DEVICE_ID_INUSED],
+      CL_PROGRAM_BUILD_LOG, 0, NULL, &ret_val_size);
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(string("exceptions in _InitCL -> getting resource information"));
+  }
+
+  build_log = (char *)malloc(ret_val_size + 1);
+  oclHandles.cl_status = clGetProgramBuildInfo(
+      oclHandles.program, oclHandles.devices[DEVICE_ID_INUSED],
+      CL_PROGRAM_BUILD_LOG, ret_val_size, build_log, NULL);
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(string(
+        "exceptions in _InitCL -> getting resources allocation information-2"));
+  }
+  build_log[ret_val_size] = '\0';
+  std::cout << "--cambine:" << build_log << std::endl;
+  free(build_log);
+#endif
+}
+
+//---------------------------------------
+// release CL objects
+void _clRelease() {
+  char errorFlag = false;
+
+  for (int nKernel = 0; nKernel < oclHandles.kernel.size(); nKernel++) {
+    if (oclHandles.kernel[nKernel] != NULL) {
+      cl_int resultCL = clReleaseKernel(oclHandles.kernel[nKernel]);
+      if (resultCL != CL_SUCCESS) {
+        cerr << "ReleaseCL()::Error: In clReleaseKernel" << endl;
+        errorFlag = true;
+      }
+      oclHandles.kernel[nKernel] = NULL;
+    }
+    oclHandles.kernel.clear();
+  }
+
+  if (oclHandles.program != NULL) {
+    cl_int resultCL = clReleaseProgram(oclHandles.program);
+    if (resultCL != CL_SUCCESS) {
+      cerr << "ReleaseCL()::Error: In clReleaseProgram" << endl;
+      errorFlag = true;
+    }
+    oclHandles.program = NULL;
+  }
+
+  if (oclHandles.queue != NULL) {
+    cl_int resultCL = clReleaseCommandQueue(oclHandles.queue);
+    if (resultCL != CL_SUCCESS) {
+      cerr << "ReleaseCL()::Error: In clReleaseCommandQueue" << endl;
+      errorFlag = true;
+    }
+    oclHandles.queue = NULL;
+  }
+
+  free(oclHandles.devices);
+
+  if (oclHandles.context != NULL) {
+    cl_int resultCL = clReleaseContext(oclHandles.context);
+    if (resultCL != CL_SUCCESS) {
+      cerr << "ReleaseCL()::Error: In clReleaseContext" << endl;
+      errorFlag = true;
+    }
+    oclHandles.context = NULL;
+  }
+
+  if (errorFlag)
+    throw(string("ReleaseCL()::Error encountered."));
+}
+//--------------------------------------------------------
+//--cambine:create buffer and then copy data from host to device
+cl_mem _clCreateAndCpyMem(int size, void *h_mem_source) throw(string) {
+  cl_mem d_mem;
+  d_mem = clCreateBuffer(oclHandles.context,
+                         CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, size,
+                         h_mem_source, &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateAndCpyMem()"));
+#endif
+  return d_mem;
+}
+//-------------------------------------------------------
+//--cambine:	create read only  buffer for devices
+//--date:	17/01/2011
+cl_mem _clMallocRW(int size, void *h_mem_ptr) throw(string) {
+  cl_mem d_mem;
+  d_mem = clCreateBuffer(oclHandles.context,
+                         CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR, size,
+                         h_mem_ptr, &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clMallocRW"));
+#endif
+  return d_mem;
+}
+//-------------------------------------------------------
+//--cambine:	create read and write buffer for devices
+//--date:	17/01/2011
+cl_mem _clMalloc(int size, void *h_mem_ptr) throw(string) {
+  cl_mem d_mem;
+  d_mem = clCreateBuffer(oclHandles.context,
+                         CL_MEM_WRITE_ONLY | CL_MEM_COPY_HOST_PTR, size,
+                         h_mem_ptr, &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clMalloc"));
+#endif
+  return d_mem;
+}
+
+//-------------------------------------------------------
+//--cambine:	transfer data from host to device
+//--date:	17/01/2011
+void _clMemcpyH2D(cl_mem d_mem, int size, const void *h_mem_ptr) throw(string) {
+  oclHandles.cl_status = clEnqueueWriteBuffer(
+      oclHandles.queue, d_mem, CL_TRUE, 0, size, h_mem_ptr, 0, NULL, NULL);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clMemcpyH2D"));
+#endif
+}
+//--------------------------------------------------------
+//--cambine:create buffer and then copy data from host to device with pinned
+// memory
+cl_mem _clCreateAndCpyPinnedMem(int size, float *h_mem_source) throw(string) {
+  cl_mem d_mem, d_mem_pinned;
+  float *h_mem_pinned = NULL;
+  d_mem_pinned = clCreateBuffer(oclHandles.context,
+                                CL_MEM_READ_ONLY | CL_MEM_ALLOC_HOST_PTR, size,
+                                NULL, &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateAndCpyMem()->d_mem_pinned"));
+#endif
+  //------------
+  d_mem = clCreateBuffer(oclHandles.context, CL_MEM_READ_ONLY, size, NULL,
+                         &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateAndCpyMem() -> d_mem "));
+#endif
+  //----------
+  h_mem_pinned = (cl_float *)clEnqueueMapBuffer(
+      oclHandles.queue, d_mem_pinned, CL_TRUE, CL_MAP_WRITE, 0, size, 0, NULL,
+      NULL, &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateAndCpyMem() -> clEnqueueMapBuffer"));
+#endif
+  int element_number = size / sizeof(float);
+#pragma omp parallel for
+  for (int i = 0; i < element_number; i++) {
+    h_mem_pinned[i] = h_mem_source[i];
+  }
+  //----------
+  oclHandles.cl_status = clEnqueueWriteBuffer(
+      oclHandles.queue, d_mem, CL_TRUE, 0, size, h_mem_pinned, 0, NULL, NULL);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateAndCpyMem() -> clEnqueueWriteBuffer"));
+#endif
+
+  return d_mem;
+}
+
+//--------------------------------------------------------
+//--cambine:create write only buffer on device
+cl_mem _clMallocWO(int size) throw(string) {
+  cl_mem d_mem;
+  d_mem = clCreateBuffer(oclHandles.context, CL_MEM_WRITE_ONLY, size, 0,
+                         &oclHandles.cl_status);
+#ifdef ERRMSG
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(string("excpetion in _clCreateMem()"));
+#endif
+  return d_mem;
+}
+
+//--------------------------------------------------------
+// transfer data from device to host
+void _clMemcpyD2H(cl_mem d_mem, int size, void *h_mem) throw(string) {
+  oclHandles.cl_status = clEnqueueReadBuffer(oclHandles.queue, d_mem, CL_TRUE,
+                                             0, size, h_mem, 0, 0, 0);
+#ifdef ERRMSG
+  oclHandles.error_str = "excpetion in _clCpyMemD2H -> ";
+  switch (oclHandles.cl_status) {
+  case CL_INVALID_COMMAND_QUEUE:
+    oclHandles.error_str += "CL_INVALID_COMMAND_QUEUE";
+    break;
+  case CL_INVALID_CONTEXT:
+    oclHandles.error_str += "CL_INVALID_CONTEXT";
+    break;
+  case CL_INVALID_MEM_OBJECT:
+    oclHandles.error_str += "CL_INVALID_MEM_OBJECT";
+    break;
+  case CL_INVALID_VALUE:
+    oclHandles.error_str += "CL_INVALID_VALUE";
+    break;
+  case CL_INVALID_EVENT_WAIT_LIST:
+    oclHandles.error_str += "CL_INVALID_EVENT_WAIT_LIST";
+    break;
+  case CL_MEM_OBJECT_ALLOCATION_FAILURE:
+    oclHandles.error_str += "CL_MEM_OBJECT_ALLOCATION_FAILURE";
+    break;
+  case CL_OUT_OF_HOST_MEMORY:
+    oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+    break;
+  default:
+    oclHandles.error_str += "Unknown reason";
+    break;
+  }
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(oclHandles.error_str);
+#endif
+}
+
+//--------------------------------------------------------
+// set kernel arguments
+void _clSetArgs(int kernel_id, int arg_idx, void *d_mem,
+                int size = 0) throw(string) {
+  if (!size) {
+    oclHandles.cl_status = clSetKernelArg(oclHandles.kernel[kernel_id], arg_idx,
+                                          sizeof(d_mem), &d_mem);
+#ifdef ERRMSG
+    oclHandles.error_str = "excpetion in _clSetKernelArg() ";
+    switch (oclHandles.cl_status) {
+    case CL_INVALID_KERNEL:
+      oclHandles.error_str += "CL_INVALID_KERNEL";
+      break;
+    case CL_INVALID_ARG_INDEX:
+      oclHandles.error_str += "CL_INVALID_ARG_INDEX";
+      break;
+    case CL_INVALID_ARG_VALUE:
+      oclHandles.error_str += "CL_INVALID_ARG_VALUE";
+      break;
+    case CL_INVALID_MEM_OBJECT:
+      oclHandles.error_str += "CL_INVALID_MEM_OBJECT";
+      break;
+    case CL_INVALID_SAMPLER:
+      oclHandles.error_str += "CL_INVALID_SAMPLER";
+      break;
+    case CL_INVALID_ARG_SIZE:
+      oclHandles.error_str += "CL_INVALID_ARG_SIZE";
+      break;
+    case CL_OUT_OF_RESOURCES:
+      oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+      break;
+    case CL_OUT_OF_HOST_MEMORY:
+      oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+      break;
+    default:
+      oclHandles.error_str += "Unknown reason";
+      break;
+    }
+    if (oclHandles.cl_status != CL_SUCCESS)
+      throw(oclHandles.error_str);
+#endif
+  } else {
+    oclHandles.cl_status =
+        clSetKernelArg(oclHandles.kernel[kernel_id], arg_idx, size, d_mem);
+#ifdef ERRMSG
+    oclHandles.error_str = "excpetion in _clSetKernelArg() ";
+    switch (oclHandles.cl_status) {
+    case CL_INVALID_KERNEL:
+      oclHandles.error_str += "CL_INVALID_KERNEL";
+      break;
+    case CL_INVALID_ARG_INDEX:
+      oclHandles.error_str += "CL_INVALID_ARG_INDEX";
+      break;
+    case CL_INVALID_ARG_VALUE:
+      oclHandles.error_str += "CL_INVALID_ARG_VALUE";
+      break;
+    case CL_INVALID_MEM_OBJECT:
+      oclHandles.error_str += "CL_INVALID_MEM_OBJECT";
+      break;
+    case CL_INVALID_SAMPLER:
+      oclHandles.error_str += "CL_INVALID_SAMPLER";
+      break;
+    case CL_INVALID_ARG_SIZE:
+      oclHandles.error_str += "CL_INVALID_ARG_SIZE";
+      break;
+    case CL_OUT_OF_RESOURCES:
+      oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+      break;
+    case CL_OUT_OF_HOST_MEMORY:
+      oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+      break;
+    default:
+      oclHandles.error_str += "Unknown reason";
+      break;
+    }
+    if (oclHandles.cl_status != CL_SUCCESS)
+      throw(oclHandles.error_str);
+#endif
+  }
+}
+void _clFinish() throw(string) {
+  oclHandles.cl_status = clFinish(oclHandles.queue);
+#ifdef ERRMSG
+  oclHandles.error_str = "excpetion in _clFinish";
+  switch (oclHandles.cl_status) {
+  case CL_INVALID_COMMAND_QUEUE:
+    oclHandles.error_str += "CL_INVALID_COMMAND_QUEUE";
+    break;
+  case CL_OUT_OF_RESOURCES:
+    oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+    break;
+  case CL_OUT_OF_HOST_MEMORY:
+    oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+    break;
+  default:
+    oclHandles.error_str += "Unknown reasons";
+    break;
+  }
+  if (oclHandles.cl_status != CL_SUCCESS) {
+    throw(oclHandles.error_str);
+  }
+#endif
+}
+//--------------------------------------------------------
+//--cambine:enqueue kernel
+void _clInvokeKernel(int kernel_id, int work_items,
+                     int work_group_size) throw(string) {
+  cl_uint work_dim = WORK_DIM;
+  cl_event e[1];
+  if (work_items % work_group_size != 0) // process situations that work_items
+                                         // cannot be divided by work_group_size
+    work_items =
+        work_items + (work_group_size - (work_items % work_group_size));
+  size_t local_work_size[] = {work_group_size, 1};
+  size_t global_work_size[] = {work_items, 1};
+  oclHandles.cl_status = clEnqueueNDRangeKernel(
+      oclHandles.queue, oclHandles.kernel[kernel_id], work_dim, 0,
+      global_work_size, local_work_size, 0, 0, &(e[0]));
+#ifdef ERRMSG
+  oclHandles.error_str = "excpetion in _clInvokeKernel() -> ";
+  switch (oclHandles.cl_status) {
+  case CL_INVALID_PROGRAM_EXECUTABLE:
+    oclHandles.error_str += "CL_INVALID_PROGRAM_EXECUTABLE";
+    break;
+  case CL_INVALID_COMMAND_QUEUE:
+    oclHandles.error_str += "CL_INVALID_COMMAND_QUEUE";
+    break;
+  case CL_INVALID_KERNEL:
+    oclHandles.error_str += "CL_INVALID_KERNEL";
+    break;
+  case CL_INVALID_CONTEXT:
+    oclHandles.error_str += "CL_INVALID_CONTEXT";
+    break;
+  case CL_INVALID_KERNEL_ARGS:
+    oclHandles.error_str += "CL_INVALID_KERNEL_ARGS";
+    break;
+  case CL_INVALID_WORK_DIMENSION:
+    oclHandles.error_str += "CL_INVALID_WORK_DIMENSION";
+    break;
+  case CL_INVALID_GLOBAL_WORK_SIZE:
+    oclHandles.error_str += "CL_INVALID_GLOBAL_WORK_SIZE";
+    break;
+  case CL_INVALID_WORK_GROUP_SIZE:
+    oclHandles.error_str += "CL_INVALID_WORK_GROUP_SIZE";
+    break;
+  case CL_INVALID_WORK_ITEM_SIZE:
+    oclHandles.error_str += "CL_INVALID_WORK_ITEM_SIZE";
+    break;
+  case CL_INVALID_GLOBAL_OFFSET:
+    oclHandles.error_str += "CL_INVALID_GLOBAL_OFFSET";
+    break;
+  case CL_OUT_OF_RESOURCES:
+    oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+    break;
+  case CL_MEM_OBJECT_ALLOCATION_FAILURE:
+    oclHandles.error_str += "CL_MEM_OBJECT_ALLOCATION_FAILURE";
+    break;
+  case CL_INVALID_EVENT_WAIT_LIST:
+    oclHandles.error_str += "CL_INVALID_EVENT_WAIT_LIST";
+    break;
+  case CL_OUT_OF_HOST_MEMORY:
+    oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+    break;
+  default:
+    oclHandles.error_str += "Unkown reseason";
+    break;
+  }
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(oclHandles.error_str);
+#endif
+  //_clFinish();
+  // oclHandles.cl_status = clWaitForEvents(1, &e[0]);
+  // #ifdef ERRMSG
+  // if (oclHandles.cl_status!= CL_SUCCESS)
+  //     throw(string("excpetion in _clEnqueueNDRange() -> clWaitForEvents"));
+  // #endif
+}
+void _clInvokeKernel2D(int kernel_id, int range_x, int range_y, int group_x,
+                       int group_y) throw(string) {
+  cl_uint work_dim = WORK_DIM;
+  size_t local_work_size[] = {group_x, group_y};
+  size_t global_work_size[] = {range_x, range_y};
+  cl_event e[1];
+  /*if(work_items%work_group_size != 0)	//process situations that work_items
+    cannot be divided by work_group_size
+    work_items = work_items + (work_group_size-(work_items%work_group_size));*/
+  oclHandles.cl_status = clEnqueueNDRangeKernel(
+      oclHandles.queue, oclHandles.kernel[kernel_id], work_dim, 0,
+      global_work_size, local_work_size, 0, 0, &(e[0]));
+#ifdef ERRMSG
+  oclHandles.error_str = "excpetion in _clInvokeKernel() -> ";
+  switch (oclHandles.cl_status) {
+  case CL_INVALID_PROGRAM_EXECUTABLE:
+    oclHandles.error_str += "CL_INVALID_PROGRAM_EXECUTABLE";
+    break;
+  case CL_INVALID_COMMAND_QUEUE:
+    oclHandles.error_str += "CL_INVALID_COMMAND_QUEUE";
+    break;
+  case CL_INVALID_KERNEL:
+    oclHandles.error_str += "CL_INVALID_KERNEL";
+    break;
+  case CL_INVALID_CONTEXT:
+    oclHandles.error_str += "CL_INVALID_CONTEXT";
+    break;
+  case CL_INVALID_KERNEL_ARGS:
+    oclHandles.error_str += "CL_INVALID_KERNEL_ARGS";
+    break;
+  case CL_INVALID_WORK_DIMENSION:
+    oclHandles.error_str += "CL_INVALID_WORK_DIMENSION";
+    break;
+  case CL_INVALID_GLOBAL_WORK_SIZE:
+    oclHandles.error_str += "CL_INVALID_GLOBAL_WORK_SIZE";
+    break;
+  case CL_INVALID_WORK_GROUP_SIZE:
+    oclHandles.error_str += "CL_INVALID_WORK_GROUP_SIZE";
+    break;
+  case CL_INVALID_WORK_ITEM_SIZE:
+    oclHandles.error_str += "CL_INVALID_WORK_ITEM_SIZE";
+    break;
+  case CL_INVALID_GLOBAL_OFFSET:
+    oclHandles.error_str += "CL_INVALID_GLOBAL_OFFSET";
+    break;
+  case CL_OUT_OF_RESOURCES:
+    oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+    break;
+  case CL_MEM_OBJECT_ALLOCATION_FAILURE:
+    oclHandles.error_str += "CL_MEM_OBJECT_ALLOCATION_FAILURE";
+    break;
+  case CL_INVALID_EVENT_WAIT_LIST:
+    oclHandles.error_str += "CL_INVALID_EVENT_WAIT_LIST";
+    break;
+  case CL_OUT_OF_HOST_MEMORY:
+    oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+    break;
+  default:
+    oclHandles.error_str += "Unkown reseason";
+    break;
+  }
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(oclHandles.error_str);
+#endif
+  //_clFinish();
+  /*oclHandles.cl_status = clWaitForEvents(1, &e[0]);
+
+  #ifdef ERRMSG
+
+  if (oclHandles.cl_status!= CL_SUCCESS)
+
+      throw(string("excpetion in _clEnqueueNDRange() -> clWaitForEvents"));
+
+  #endif*/
+}
+
+//--------------------------------------------------------
+// release OpenCL objects
+void _clFree(cl_mem ob) throw(string) {
+  if (ob != NULL)
+    oclHandles.cl_status = clReleaseMemObject(ob);
+#ifdef ERRMSG
+  oclHandles.error_str = "excpetion in _clFree() ->";
+  switch (oclHandles.cl_status) {
+  case CL_INVALID_MEM_OBJECT:
+    oclHandles.error_str += "CL_INVALID_MEM_OBJECT";
+    break;
+  case CL_OUT_OF_RESOURCES:
+    oclHandles.error_str += "CL_OUT_OF_RESOURCES";
+    break;
+  case CL_OUT_OF_HOST_MEMORY:
+    oclHandles.error_str += "CL_OUT_OF_HOST_MEMORY";
+    break;
+  default:
+    oclHandles.error_str += "Unkown reseason";
+    break;
+  }
+  if (oclHandles.cl_status != CL_SUCCESS)
+    throw(oclHandles.error_str);
+#endif
+}
+#endif //_CL_HELPER_

benchmarks/new_opencl/bfs/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = bfs
+
+SRCS = main.cc
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/bfs/kernel.cl

@@ -0,0 +1,53 @@
+/* ============================================================
+//--cambine: kernel funtion of Breadth-First-Search
+//--author:	created by Jianbin Fang
+//--date:	06/12/2010
+============================================================ */
+
+//#pragma OPENCL EXTENSION cl_khr_byte_addressable_store: enable
+
+//Structure to hold a node information
+typedef struct{
+	int starting;
+	int no_of_edges;
+} Node;
+
+//--7 parameters
+__kernel void BFS_1( const __global Node* g_graph_nodes,
+					const __global int* g_graph_edges, 
+					__global char* g_graph_mask, 
+					__global char* g_updating_graph_mask, 
+					__global char* g_graph_visited, 
+					__global int* g_cost, 
+					const  int no_of_nodes){
+	int tid = get_global_id(0);
+	if( tid<no_of_nodes && g_graph_mask[tid]){
+		g_graph_mask[tid]=false;
+		for(int i=g_graph_nodes[tid].starting; i<(g_graph_nodes[tid].no_of_edges + g_graph_nodes[tid].starting); i++){
+			int id = g_graph_edges[i];
+			if(!g_graph_visited[id]){
+				g_cost[id]=g_cost[tid]+1;
+				g_updating_graph_mask[id]=true;
+				}
+			}
+	}	
+}
+
+//--5 parameters
+__kernel void BFS_2(__global char* g_graph_mask, 
+					__global char* g_updating_graph_mask, 
+					__global char* g_graph_visited, 
+					__global char* g_over,
+					const  int no_of_nodes
+					) {
+	int tid = get_global_id(0);
+	if( tid<no_of_nodes && g_updating_graph_mask[tid]){
+
+		g_graph_mask[tid]=true;
+		g_graph_visited[tid]=true;
+		*g_over=true;
+		g_updating_graph_mask[tid]=false;
+	}
+}
+
+

benchmarks/new_opencl/bfs/main.cc

@@ -0,0 +1,297 @@
+//--by Jianbin Fang
+
+#include <cstdlib>
+#include <cstring>
+#include <iostream>
+#include <string>
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/stat.h>
+
+#ifdef PROFILING
+#include "timer.h"
+#endif
+
+#include "CLHelper.h"
+#include "util.h"
+
+#define MAX_THREADS_PER_BLOCK 256
+
+// Structure to hold a node information
+struct Node {
+  int starting;
+  int no_of_edges;
+};
+
+//----------------------------------------------------------
+//--bfs on cpu
+//--programmer:	jianbin
+//--date:	26/01/2011
+//--note: width is changed to the new_width
+//----------------------------------------------------------
+void run_bfs_cpu(int no_of_nodes, Node *h_graph_nodes, int edge_list_size,
+                 int *h_graph_edges, char *h_graph_mask,
+                 char *h_updating_graph_mask, char *h_graph_visited,
+                 int *h_cost_ref) {
+  char stop;
+  int k = 0;
+  do {
+    // if no thread changes this value then the loop stops
+    stop = false;
+    for (int tid = 0; tid < no_of_nodes; tid++) {
+      if (h_graph_mask[tid] == true) {
+        h_graph_mask[tid] = false;
+        for (int i = h_graph_nodes[tid].starting;
+             i < (h_graph_nodes[tid].no_of_edges + h_graph_nodes[tid].starting);
+             i++) {
+          int id =
+              h_graph_edges[i]; //--cambine: node id is connected with node tid
+          if (!h_graph_visited[id]) { //--cambine: if node id has not been
+                                      //visited, enter the body below
+            h_cost_ref[id] = h_cost_ref[tid] + 1;
+            h_updating_graph_mask[id] = true;
+          }
+        }
+      }
+    }
+
+    for (int tid = 0; tid < no_of_nodes; tid++) {
+      if (h_updating_graph_mask[tid] == true) {
+        h_graph_mask[tid] = true;
+        h_graph_visited[tid] = true;
+        stop = true;
+        h_updating_graph_mask[tid] = false;
+      }
+    }
+    k++;
+  } while (stop);
+}
+//----------------------------------------------------------
+//--breadth first search on GPUs
+//----------------------------------------------------------
+void run_bfs_gpu(int no_of_nodes, Node *h_graph_nodes, int edge_list_size,
+                 int *h_graph_edges, char *h_graph_mask,
+                 char *h_updating_graph_mask, char *h_graph_visited,
+                 int *h_cost) throw(std::string) {
+
+  // int number_elements = height*width;
+  char h_over;
+  cl_mem d_graph_nodes, d_graph_edges, d_graph_mask, d_updating_graph_mask,
+      d_graph_visited, d_cost, d_over;
+  try {
+    //--1 transfer data from host to device
+    _clInit();
+    d_graph_nodes = _clMalloc(no_of_nodes * sizeof(Node), h_graph_nodes);
+    d_graph_edges = _clMalloc(edge_list_size * sizeof(int), h_graph_edges);
+    d_graph_mask = _clMallocRW(no_of_nodes * sizeof(char), h_graph_mask);
+    d_updating_graph_mask =
+        _clMallocRW(no_of_nodes * sizeof(char), h_updating_graph_mask);
+    d_graph_visited = _clMallocRW(no_of_nodes * sizeof(char), h_graph_visited);
+
+    d_cost = _clMallocRW(no_of_nodes * sizeof(int), h_cost);
+    d_over = _clMallocRW(sizeof(char), &h_over);
+
+    _clMemcpyH2D(d_graph_nodes, no_of_nodes * sizeof(Node), h_graph_nodes);
+    _clMemcpyH2D(d_graph_edges, edge_list_size * sizeof(int), h_graph_edges);
+    _clMemcpyH2D(d_graph_mask, no_of_nodes * sizeof(char), h_graph_mask);
+    _clMemcpyH2D(d_updating_graph_mask, no_of_nodes * sizeof(char),
+                 h_updating_graph_mask);
+    _clMemcpyH2D(d_graph_visited, no_of_nodes * sizeof(char), h_graph_visited);
+    _clMemcpyH2D(d_cost, no_of_nodes * sizeof(int), h_cost);
+
+//--2 invoke kernel
+#ifdef PROFILING
+    timer kernel_timer;
+    double kernel_time = 0.0;
+    kernel_timer.reset();
+    kernel_timer.start();
+#endif
+    do {
+      h_over = false;
+      _clMemcpyH2D(d_over, sizeof(char), &h_over);
+      //--kernel 0
+      int kernel_id = 0;
+      int kernel_idx = 0;
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_nodes);
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_edges);
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_mask);
+      _clSetArgs(kernel_id, kernel_idx++, d_updating_graph_mask);
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_visited);
+      _clSetArgs(kernel_id, kernel_idx++, d_cost);
+      _clSetArgs(kernel_id, kernel_idx++, &no_of_nodes, sizeof(int));
+
+      // int work_items = no_of_nodes;
+      _clInvokeKernel(kernel_id, no_of_nodes, work_group_size);
+
+      //--kernel 1
+      kernel_id = 1;
+      kernel_idx = 0;
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_mask);
+      _clSetArgs(kernel_id, kernel_idx++, d_updating_graph_mask);
+      _clSetArgs(kernel_id, kernel_idx++, d_graph_visited);
+      _clSetArgs(kernel_id, kernel_idx++, d_over);
+      _clSetArgs(kernel_id, kernel_idx++, &no_of_nodes, sizeof(int));
+
+      // work_items = no_of_nodes;
+      _clInvokeKernel(kernel_id, no_of_nodes, work_group_size);
+
+      _clMemcpyD2H(d_over, sizeof(char), &h_over);
+    } while (h_over);
+
+    _clFinish();
+#ifdef PROFILING
+    kernel_timer.stop();
+    kernel_time = kernel_timer.getTimeInSeconds();
+#endif
+    //--3 transfer data from device to host
+    _clMemcpyD2H(d_cost, no_of_nodes * sizeof(int), h_cost);
+//--statistics
+#ifdef PROFILING
+    std::cout << "kernel time(s):" << kernel_time << std::endl;
+#endif
+    //--4 release cl resources.
+    _clFree(d_graph_nodes);
+    _clFree(d_graph_edges);
+    _clFree(d_graph_mask);
+    _clFree(d_updating_graph_mask);
+    _clFree(d_graph_visited);
+    _clFree(d_cost);
+    _clFree(d_over);
+    _clRelease();
+  } catch (std::string msg) {
+    _clFree(d_graph_nodes);
+    _clFree(d_graph_edges);
+    _clFree(d_graph_mask);
+    _clFree(d_updating_graph_mask);
+    _clFree(d_graph_visited);
+    _clFree(d_cost);
+    _clFree(d_over);
+    _clRelease();
+    std::string e_str = "in run_transpose_gpu -> ";
+    e_str += msg;
+    throw(e_str);
+  }
+  return;
+}
+
+//----------------------------------------------------------
+//--cambine:	main function
+//--author:		created by Jianbin Fang
+//--date:		25/01/2011
+//----------------------------------------------------------
+int main(int argc, char *argv[]) {
+	printf("enter demo main\n");
+
+  int no_of_nodes;
+  int edge_list_size;
+  FILE *fp;
+  Node *h_graph_nodes;
+  char *h_graph_mask, *h_updating_graph_mask, *h_graph_visited;
+  
+  try {
+    char *input_f = "graph4096.txt";
+    printf("Reading File\n");
+    // Read in Graph from a file
+    fp = fopen(input_f, "r");
+    if (!fp) {
+      printf("Error Reading graph file\n");
+      return 0;
+    }
+
+    printf("Reading File completed!\n");
+
+    int source = 0;
+
+    fscanf(fp, "%d", &no_of_nodes);
+
+    int num_of_blocks = 1;
+    int num_of_threads_per_block = no_of_nodes;
+
+    // Make execution Parameters according to the number of nodes
+    // Distribute threads across multiple Blocks if necessary
+    if (no_of_nodes > MAX_THREADS_PER_BLOCK) {
+      num_of_blocks = (int)ceil(no_of_nodes / (double)MAX_THREADS_PER_BLOCK);
+      num_of_threads_per_block = MAX_THREADS_PER_BLOCK;
+    }
+    work_group_size = num_of_threads_per_block;
+    // allocate host memory
+    h_graph_nodes = (Node *)malloc(sizeof(Node) * no_of_nodes);
+    h_graph_mask = (char *)malloc(sizeof(char) * no_of_nodes);
+    h_updating_graph_mask = (char *)malloc(sizeof(char) * no_of_nodes);
+    h_graph_visited = (char *)malloc(sizeof(char) * no_of_nodes);
+
+    int start, edgeno;
+    // initalize the memory
+    for (int i = 0; i < no_of_nodes; i++) {
+      fscanf(fp, "%d %d", &start, &edgeno);
+      h_graph_nodes[i].starting = start;
+      h_graph_nodes[i].no_of_edges = edgeno;
+      h_graph_mask[i] = false;
+      h_updating_graph_mask[i] = false;
+      h_graph_visited[i] = false;
+    }
+    // read the source node from the file
+    fscanf(fp, "%d", &source);
+    source = 0;
+    // set the source node as true in the mask
+    h_graph_mask[source] = true;
+    h_graph_visited[source] = true;
+    fscanf(fp, "%d", &edge_list_size);
+    int id, cost;
+    int *h_graph_edges = (int *)malloc(sizeof(int) * edge_list_size);
+    for (int i = 0; i < edge_list_size; i++) {
+      fscanf(fp, "%d", &id);
+      fscanf(fp, "%d", &cost);
+      h_graph_edges[i] = id;
+    }
+
+    if (fp)
+      fclose(fp);
+    // allocate mem for the result on host side
+    int *h_cost = (int *)malloc(sizeof(int) * no_of_nodes);
+    int *h_cost_ref = (int *)malloc(sizeof(int) * no_of_nodes);
+    for (int i = 0; i < no_of_nodes; i++) {
+      h_cost[i] = -1;
+      h_cost_ref[i] = -1;
+    }
+    h_cost[source] = 0;
+    h_cost_ref[source] = 0;
+    //---------------------------------------------------------
+    //--gpu entry
+    run_bfs_gpu(no_of_nodes, h_graph_nodes, edge_list_size, h_graph_edges,
+                h_graph_mask, h_updating_graph_mask, h_graph_visited, h_cost);
+    //---------------------------------------------------------
+    //--cpu entry
+    // initalize the memory again
+    for (int i = 0; i < no_of_nodes; i++) {
+      h_graph_mask[i] = false;
+      h_updating_graph_mask[i] = false;
+      h_graph_visited[i] = false;
+    }
+    // set the source node as true in the mask
+    source = 0;
+    h_graph_mask[source] = true;
+    h_graph_visited[source] = true;
+    run_bfs_cpu(no_of_nodes, h_graph_nodes, edge_list_size, h_graph_edges,
+                h_graph_mask, h_updating_graph_mask, h_graph_visited,
+                h_cost_ref);
+    //---------------------------------------------------------
+    //--result varification
+    compare_results<int>(h_cost_ref, h_cost, no_of_nodes);
+    // release host memory
+    free(h_graph_nodes);
+    free(h_graph_mask);
+    free(h_updating_graph_mask);
+    free(h_graph_visited);
+
+  } catch (std::string msg) {
+    std::cout << "--cambine: exception in main ->" << msg << std::endl;
+    // release host memory
+    free(h_graph_nodes);
+    free(h_graph_mask);
+    free(h_updating_graph_mask);
+    free(h_graph_visited);
+  }
+
+  return 0;
+}

benchmarks/new_opencl/bfs/run

@@ -0,0 +1 @@
+./bfs ../../data/bfs/graph1MW_6.txt

benchmarks/new_opencl/bfs/timer.cc

@@ -0,0 +1,78 @@
+#include <cstdlib>
+#include <cstring>
+#include <fstream>
+#include <iomanip>
+
+#include "timer.h"
+
+
+using namespace std;
+
+double timer::CPU_speed_in_MHz = timer::get_CPU_speed_in_MHz();
+
+
+double timer::get_CPU_speed_in_MHz()
+{
+#if defined __linux__
+    ifstream infile("/proc/cpuinfo");
+    char     buffer[256], *colon;
+
+    while (infile.good()) {
+	infile.getline(buffer, 256);
+
+	if (strncmp("cpu MHz", buffer, 7) == 0 && (colon = strchr(buffer, ':')) != 0)
+	    return atof(colon + 2);
+    }
+#endif
+
+    return 0.0;
+}
+
+
+void timer::print_time(ostream &str, const char *which, double time) const
+{
+    static const char *units[] = { " ns", " us", " ms", "  s", " ks", 0 };
+    const char	      **unit   = units;
+
+    time = 1000.0 * time / CPU_speed_in_MHz;
+
+    while (time >= 999.5 && unit[1] != 0) {
+	time /= 1000.0;
+	++ unit;
+    }
+
+    str << which << " = " << setprecision(3) << setw(4) << time << *unit;
+}
+
+
+ostream &timer::print(ostream &str)
+{
+    str << left << setw(25) << (name != 0 ? name : "timer") << ": " << right;
+
+    if (CPU_speed_in_MHz == 0)
+	str << "could not determine CPU speed\n";
+    else if (count > 0) {
+	double total = static_cast<double>(total_time);
+
+	print_time(str, "avg", total / static_cast<double>(count));
+	print_time(str, ", total", total);
+	str << ", count = " << setw(9) << count << '\n';
+    }
+    else
+	str << "not used\n";
+
+    return str;
+}
+
+
+ostream &operator << (ostream &str, class timer &timer)
+{
+    return timer.print(str);
+}
+
+double timer::getTimeInSeconds()
+{
+    double total = static_cast<double>(total_time);
+    double res = (total / 1000000.0) / CPU_speed_in_MHz;
+    return res;
+}

benchmarks/new_opencl/bfs/timer.h

@@ -0,0 +1,101 @@
+#ifndef timer_h
+#define timer_h
+
+#include <iostream>
+
+class timer {
+public:
+  timer(const char *name = 0);
+  timer(const char *name, std::ostream &write_on_exit);
+
+  ~timer();
+
+  void start(), stop();
+  void reset();
+  std::ostream &print(std::ostream &);
+
+  double getTimeInSeconds();
+
+private:
+  void print_time(std::ostream &, const char *which, double time) const;
+
+  union {
+    long long total_time;
+    struct {
+#if defined __PPC__
+      int high, low;
+#else
+      int low, high;
+#endif
+    };
+  };
+
+  unsigned long long count;
+  const char *const name;
+  std::ostream *const write_on_exit;
+
+  static double CPU_speed_in_MHz, get_CPU_speed_in_MHz();
+};
+
+std::ostream &operator<<(std::ostream &, class timer &);
+
+inline void timer::reset() {
+  total_time = 0;
+  count = 0;
+}
+
+inline timer::timer(const char *name) : name(name), write_on_exit(0) {
+  reset();
+}
+
+inline timer::timer(const char *name, std::ostream &write_on_exit)
+    : name(name), write_on_exit(&write_on_exit) {
+  reset();
+}
+
+inline timer::~timer() {
+  if (write_on_exit != 0)
+    print(*write_on_exit);
+}
+
+inline void timer::start() {
+#if (defined __PATHSCALE__) && (defined __i386 || defined __x86_64)
+  unsigned eax, edx;
+
+  asm volatile("rdtsc" : "=a"(eax), "=d"(edx));
+
+  total_time -= ((unsigned long long)edx << 32) + eax;
+#elif (defined __GNUC__ || defined __INTEL_COMPILER) &&                        \
+    (defined __i386 || defined __x86_64)
+  asm volatile("rdtsc\n\t"
+               "subl %%eax, %0\n\t"
+               "sbbl %%edx, %1"
+               : "+m"(low), "+m"(high)
+               :
+               : "eax", "edx");
+#else
+#error Compiler/Architecture not recognized
+#endif
+}
+
+inline void timer::stop() {
+#if (defined __PATHSCALE__) && (defined __i386 || defined __x86_64)
+  unsigned eax, edx;
+
+  asm volatile("rdtsc" : "=a"(eax), "=d"(edx));
+
+  total_time += ((unsigned long long)edx << 32) + eax;
+#elif (defined __GNUC__ || defined __INTEL_COMPILER) &&                        \
+    (defined __i386 || defined __x86_64)
+  asm volatile("rdtsc\n\t"
+               "addl %%eax, %0\n\t"
+               "adcl %%edx, %1"
+               : "+m"(low), "+m"(high)
+               :
+               : "eax", "edx");
+#endif
+
+  ++count;
+}
+
+#endif

benchmarks/new_opencl/bfs/util.h

@@ -0,0 +1,72 @@
+#ifndef _C_UTIL_
+#define _C_UTIL_
+#include <math.h>
+#include <iostream>
+
+//-------------------------------------------------------------------
+//--initialize array with maximum limit
+//-------------------------------------------------------------------
+template<typename datatype>
+void fill(datatype *A, const int n, const datatype maxi){
+    for (int j = 0; j < n; j++) 
+    {
+        A[j] = ((datatype) maxi * (rand() / (RAND_MAX + 1.0f)));
+    }
+}
+
+//--print matrix
+template<typename datatype>
+void print_matrix(datatype *A, int height, int width){
+	for(int i=0; i<height; i++){
+		for(int j=0; j<width; j++){
+			int idx = i*width + j;
+			std::cout<<A[idx]<<" ";
+		}
+		std::cout<<std::endl;
+	}
+
+	return;
+}
+//-------------------------------------------------------------------
+//--verify results
+//-------------------------------------------------------------------
+#define MAX_RELATIVE_ERROR  .002
+template<typename datatype>
+void verify_array(const datatype *cpuResults, const datatype *gpuResults, const int size){
+
+    char passed = true; 
+#pragma omp parallel for
+    for (int i=0; i<size; i++){
+      if (fabs(cpuResults[i] - gpuResults[i]) / cpuResults[i] > MAX_RELATIVE_ERROR){
+         passed = false; 
+      }
+    }
+    if (passed){
+        std::cout << "--cambine:passed:-)" << endl;
+    }
+    else{
+        std::cout << "--cambine: failed:-(" << endl;
+    }
+    return ;
+}
+template<typename datatype>
+void compare_results(const datatype *cpu_results, const datatype *gpu_results, const int size){
+
+    char passed = true; 
+//#pragma omp parallel for
+    for (int i=0; i<size; i++){
+      if (cpu_results[i]!=gpu_results[i]){
+         passed = false; 
+      }
+    }
+    if (passed){
+        std::cout << "--cambine:passed:-)" << endl;
+    }
+    else{
+        std::cout << "--cambine: failed:-(" << endl;
+    }
+    return ;
+}
+
+#endif
+

benchmarks/new_opencl/guassian/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = guassian
+
+SRCS = main.cc clutils.cpp utils.cpp
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/guassian/OriginalParallel.c

@@ -0,0 +1,241 @@
+/*-----------------------------------------------------------
+** ge_p.c -- The program is to solve a linear system Ax = b
+**   by using Gaussian Elimination. The algorithm on page 101
+**   ("Foundations of Parallel Programming") is used.  
+**   The sequential version is ge_s.c.  This parallel 
+**   implementation converts three independent for() loops 
+**   into three Fans.  Use the data file ge_3.dat to verify 
+**   the correction of the output. 
+**
+** Written by Andreas Kura, 02/15/95
+** Modified by Chong-wei Xu, /04/20/95
+**-----------------------------------------------------------
+*/
+#include <us.h>
+#include <stdio.h>
+
+int Size, t;
+float **a, *b;
+BEGIN_SHARED_DECL
+  float **m;
+END_SHARED_DECL;
+FILE *fp;
+
+void InitProblemOnce();
+void InitPerRun();
+void ForwardSub();
+void Fan1();
+void Fan2();
+void Fan3();
+void InitMat();
+void InitAry();
+void PrintMat();
+void PrintAry();
+
+main ()
+{
+  InitializeUs();
+  MakeSharedVariables;  /* to make SHARED m */
+
+  InitProblemOnce();
+  InitPerRun();
+  ForwardSub();
+
+  printf("The result of matrix m is: \n");
+  PrintMat(SHARED m, Size, Size);
+  printf("The result of matrix a is: \n");
+  PrintMat(a, Size, Size);
+  printf("The result of array b is: \n");
+  PrintAry(b, Size);
+}
+
+/*------------------------------------------------------
+** InitProblemOnce -- Initialize all of matrices and
+** vectors by opening a data file specified by the user.
+**
+** We used dynamic array **a, *b, and **m to allocate
+** the memory storages.
+**------------------------------------------------------
+*/
+void InitProblemOnce()
+{
+  char filename[30];
+ 
+  printf("Enter the data file name: ");
+  scanf("%s", filename);
+  printf("The file name is: %s\n", filename);
+ 
+  fp = fopen(filename, "r");
+ 
+  fscanf(fp, "%d", &Size);
+  a = (float **) UsAllocScatterMatrix(Size, Size, sizeof(float));
+  /* 
+  a = (float **) malloc(Size * sizeof(float *));
+  for (i=0; i<Size; i++) {
+    a[i] = (float *) malloc(Size * sizeof(float));
+  }
+  */
+  InitMat(a, Size, Size);
+  printf("The input matrix a is:\n");
+  PrintMat(a, Size, Size);
+ 
+  b = (float *) UsAlloc(Size * sizeof(float));
+  /* 
+  b = (float *) malloc(Size * sizeof(float));
+  */
+  InitAry(b, Size);
+  printf("The input array b is:\n");
+  PrintAry(b, Size);
+ 
+  SHARED m = (float **) UsAllocScatterMatrix(Size, Size, sizeof(float));
+  /*
+  m = (float **) malloc(Size * sizeof(float *));
+  for (i=0; i<Size; i++) {
+    m[i] = (float *) malloc(Size * sizeof(float));
+  }
+  */
+
+  Share(&Size);
+  Share(&a);
+  Share(&b);
+}
+
+/*------------------------------------------------------
+** InitPerRun() -- Initialize the contents of the
+** multipier matrix **m
+**------------------------------------------------------
+*/
+void InitPerRun() 
+{
+  int i, j;
+
+  for (i=0; i<Size; i++)
+    for (j=0; j<Size; j++) 
+      SHARED m[i][j] = 0.0;
+}
+
+/*------------------------------------------------------
+** ForwardSub() -- Forward substitution of Gaussian
+** elimination.
+**------------------------------------------------------
+*/
+void ForwardSub()
+{
+  for (t=0; t<(Size-1); t++) {
+    Share(&t);
+    GenOnI(Fan1, Size-1-t);  /* t=0 to (Size-2), the range is
+                             ** Size-2-t+1 = Size-1-t
+                             */   
+    GenOnA(Fan2, Size-1-t, Size-t);
+    GenOnI(Fan3, Size-1-t);
+  }
+}
+
+/*-------------------------------------------------------
+** Fan1() -- Calculate multiplier matrix
+** Pay attention to the index.  Index i give the range
+** which starts from 0 to range-1.  The real values of
+** the index should be adjust and related with the value
+** of t which is defined on the ForwardSub().
+**-------------------------------------------------------
+*/
+void Fan1(dummy, i)
+int dummy, i;
+{   
+  /* Use these printf() to display the nodes and index */
+  printf("from node #%d\n", PhysProcToUsProc(Proc_Node));
+  SHARED m[i+t+1][t] = a[i+t+1][t] / a[t][t];
+  printf("i=%d, a[%d][%d]=%.2f, a[%d][%d]=%.2f, m[%d][%d]=%.2f\n",
+     (i+t+1),t,t,a[t][t],(i+t+1),t,a[i+t+1][t],(i+t+1),t,
+     SHARED m[i+t+1][t]);
+}
+
+/*-------------------------------------------------------
+** Fan2() -- Modify the matrix A into LUD
+**-------------------------------------------------------
+*/ 
+void Fan2(dummy, i, j)
+int dummy, i, j;
+{
+  a[i+1+t][j+t] -= SHARED m[i+1+t][t] * a[t][j+t];
+  Share (&a);
+}
+
+/*-------------------------------------------------------
+** Fan3() -- Modify the array b
+**-------------------------------------------------------
+*/
+void Fan3(dummy, i)
+int dummy, i;
+{
+  b[i+1+t] -= SHARED m[i+1+t][t] * b[t];
+}
+
+/*------------------------------------------------------
+** InitMat() -- Initialize the matrix by reading data
+** from the data file
+**------------------------------------------------------
+*/
+void InitMat(ary, nrow, ncol)
+float **ary;
+int nrow, ncol;
+{
+  int i, j;
+
+  for (i=0; i<nrow; i++) {
+    for (j=0; j<ncol; j++) {
+      fscanf(fp, "%f",  &ary[i][j]);
+    }
+  }  
+}
+
+/*------------------------------------------------------
+** PrintMat() -- Print the contents of the matrix
+**------------------------------------------------------
+*/
+void PrintMat(ary, nrow, ncol)
+float **ary;
+int nrow, ncol;
+{
+  int i, j;
+ 
+  for (i=0; i<nrow; i++) {
+    for (j=0; j<ncol; j++) {
+      printf("%8.2f ", ary[i][j]);
+    }
+    printf("\n");
+  }
+  printf("\n");
+}
+
+/*------------------------------------------------------
+** InitAry() -- Initialize the array (vector) by reading
+** data from the data file
+**------------------------------------------------------
+*/
+void InitAry(ary, ary_size)
+float *ary;
+int ary_size;
+{
+  int i;
+ 
+  for (i=0; i<ary_size; i++) {
+    fscanf(fp, "%f",  &ary[i]);
+  }
+}  
+ 
+/*------------------------------------------------------
+** PrintAry() -- Print the contents of the array (vector)
+**------------------------------------------------------
+*/
+void PrintAry(ary, ary_size)
+float *ary;
+int ary_size;
+{
+  int i;
+ 
+  for (i=0; i<ary_size; i++) {
+    printf("%.2f ", ary[i]);
+  }
+  printf("\n");
+}

benchmarks/new_opencl/guassian/README.txt

@@ -0,0 +1,60 @@
+The Gaussian Elimination application solves systems of equations using the
+gaussian elimination method.
+
+The application analyzes an n x n matrix and an associated 1 x n vector to solve a 
+set of equations with n variables and n unknowns. The matrix and vector describe equations
+of the form:
+
+             a0x + b0y + c0z + d0w = e0
+             a1x + b1y + c1z + d1w = e1
+             a2x + b2y + c2z + d2w = e2
+             a3x + b3y + c3z + d3w = e3
+
+where in this case n=4.  The matrix for the above equations would be as follows:
+
+            [a0 b0 c0 d0]
+            [a1 b1 c1 d1]
+            [a2 b2 c2 d2]
+            [a3 b3 c3 d3]
+            
+and the vector would be:
+
+            [e0]
+            [e1]
+            [e2]
+            [e3]
+
+The application creates a solution vector:
+
+            [x]
+            [y]
+            [z]
+            [w]
+            
+
+The Makefile may need to be adjusted for different machines, but it was written for Mac OS X and
+Linux with either NVIDIA or AMD OpenCL SDKs.
+
+Additional input files can be created with the matrixGenerator.py file in the data folder.
+
+Gaussian Elimination Usage
+
+    gaussianElimination [filename] [-hqt] [-p [int] -d [int]]
+    
+    example:
+    $ ./gaussianElimination matrix4.txt
+    
+    filename     the filename that holds the matrix data
+    
+    -h, --help   Display the help file
+    -q           Quiet mode. Suppress all text output.
+    -t           Print timing information.
+    
+    -p [int]     Choose the platform (must choose both platform and device)
+    -d [int]     Choose the device (must choose both platform and device)
+    
+    
+    Notes: 1. The filename is required as the first parameter.
+           2. If you declare either the device or the platform,
+              you must declare both.
+

benchmarks/new_opencl/guassian/clutils.h

@@ -0,0 +1,281 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#ifndef __CL_UTILS_H__
+#define __CL_UTILS_H__
+
+#include <CL/cl.h>
+
+// The cl_time type is OS specific
+#ifdef _WIN32
+#include <tchar.h>
+#include <Windows.h>
+typedef __int64 cl_time; 
+#else
+#include <sys/time.h>
+typedef double cl_time;
+#endif
+
+//-------------------------------------------------------
+// Initialization and Cleanup
+//-------------------------------------------------------
+
+// Detects platforms and devices, creates context and command queue
+cl_context cl_init(char devicePreference='\0');
+
+// Creates a context given a platform and a device
+cl_context cl_init_context(int platform,int dev,int quiet=0);
+
+// Releases resources used by clutils
+void    cl_cleanup();
+
+// Releases a kernel object
+void    cl_freeKernel(cl_kernel kernel);
+
+// Releases a memory object
+void    cl_freeMem(cl_mem mem);
+
+// Releases a program object
+void    cl_freeProgram(cl_program program);
+
+// Returns the global command queue
+cl_command_queue cl_getCommandQueue();
+
+
+//-------------------------------------------------------
+// Synchronization functions
+//-------------------------------------------------------
+
+// Performs a clFinish on the command queue
+void    cl_sync();
+
+
+//-------------------------------------------------------
+// Memory allocation
+//-------------------------------------------------------
+
+// Allocates a regular buffer on the device
+cl_mem  cl_allocBuffer(size_t mem_size, 
+            cl_mem_flags flags = CL_MEM_READ_WRITE);
+
+// XXX I don't think this does exactly what we want it to do
+// Allocates a read-only buffer and transfers the data
+cl_mem  cl_allocBufferConst(size_t mem_size, void* host_ptr);
+
+// Allocates pinned memory on the host
+cl_mem  cl_allocBufferPinned(size_t mem_size);
+
+// Allocates an image on the device
+cl_mem  cl_allocImage(size_t height, size_t width, char type, 
+            cl_mem_flags flags = CL_MEM_READ_WRITE);
+
+
+
+//-------------------------------------------------------
+// Data transfers
+//-------------------------------------------------------
+
+// Copies a buffer from the device to pinned memory on the host and 
+// maps it so it can be read
+void*   cl_copyAndMapBuffer(cl_mem dst, cl_mem src, size_t size); 
+
+// Copies from one buffer to another
+void    cl_copyBufferToBuffer(cl_mem dst, cl_mem src, size_t size);
+
+// Copies data to a buffer on the device
+void    cl_copyBufferToDevice(cl_mem dst, void *src, size_t mem_size, 
+            cl_bool blocking = CL_TRUE);
+
+// Copies data to an image on the device
+void cl_copyImageToDevice(cl_mem dst, void* src, size_t height, size_t width);
+
+// Copies an image from the device to the host
+void    cl_copyImageToHost(void* dst, cl_mem src, size_t height, size_t width);
+
+// Copies data from a device buffer to the host
+void    cl_copyBufferToHost(void *dst, cl_mem src, size_t mem_size, 
+            cl_bool blocking = CL_TRUE);
+
+// Copies data from a buffer on the device to an image on the device
+void    cl_copyBufferToImage(cl_mem src, cl_mem dst, int height, int width);
+
+// Maps a buffer
+void*   cl_mapBuffer(cl_mem mem, size_t mem_size, cl_mem_flags flags);
+
+// Unmaps a buffer
+void    cl_unmapBuffer(cl_mem mem, void *ptr);
+
+// Writes data to a zero-copy buffer on the device
+void    cl_writeToZCBuffer(cl_mem mem, void* data, size_t size);
+
+//-------------------------------------------------------
+// Program and kernels
+//-------------------------------------------------------
+
+// Compiles a program
+cl_program  cl_compileProgram(char* kernelPath, char* compileoptions, 
+                bool verboseoptions = 0);
+
+// Creates a kernel
+cl_kernel   cl_createKernel(cl_program program, const char* kernelName);
+
+
+// Sets a kernel argument
+void        cl_setKernelArg(cl_kernel kernel, unsigned int index, size_t size, 
+                void* data);
+
+
+//-------------------------------------------------------
+// Profiling/events
+//-------------------------------------------------------
+
+// Computes the execution time (start to end) for an event
+double  cl_computeExecTime(cl_event);
+
+// Compute the elapsed time between two CPU timer values
+double  cl_computeTime(cl_time start, cl_time end); 
+
+// Creates an event from CPU timers
+void    cl_createUserEvent(cl_time start, cl_time end, char* desc);
+
+// Disable logging of events
+void    cl_disableEvents();
+
+// Enable logging of events
+void    cl_enableEvents();
+
+// Query the current system time
+void    cl_getTime(cl_time* time); 
+
+// Calls a function which prints events to the terminal
+void    cl_printEvents();
+
+// Calls a function which writes the events to a file
+void    cl_writeEventsToFile(char* path);
+
+
+//-------------------------------------------------------
+// Error handling
+//-------------------------------------------------------
+
+// Compare a status value to CL_SUCCESS and optionally exit on error
+int     cl_errChk(const cl_int status, const char *msg, bool exitOnErr);
+
+// Queries the supported image formats for the device and prints
+// them to the screen
+void    printSupportedImageFormats();
+
+//-------------------------------------------------------
+// Platform and device information
+//-------------------------------------------------------
+
+bool    cl_deviceIsAMD(cl_device_id dev=NULL);
+bool    cl_deviceIsNVIDIA(cl_device_id dev=NULL);
+bool    cl_platformIsNVIDIA(cl_platform_id plat=NULL);
+char*   cl_getDeviceDriverVersion(cl_device_id dev=NULL);
+char*   cl_getDeviceName(cl_device_id dev=NULL);
+char*   cl_getDeviceVendor(cl_device_id dev=NULL);
+char*   cl_getDeviceVersion(cl_device_id dev=NULL);
+char*   cl_getPlatformName(cl_platform_id platform);
+char*   cl_getPlatformVendor(cl_platform_id platform);
+
+//-------------------------------------------------------
+// Utility functions
+//-------------------------------------------------------
+
+char* catStringWithInt(const char* str, int integer);
+
+char* itoa_portable(int value, char* result, int base);
+
+//-------------------------------------------------------
+// Data types
+//-------------------------------------------------------
+typedef struct{
+    int x;
+    int y;
+} int2;
+
+typedef struct{
+    float x;
+    float y;
+}float2;
+
+typedef struct{
+    float x;
+    float y;
+    float z;
+    float w;
+}float4;
+
+//-------------------------------------------------------
+// Defines
+//-------------------------------------------------------
+
+#define MAX_ERR_VAL 64
+
+#define NUM_PROGRAMS 7
+
+#define NUM_KERNELS 13
+#define KERNEL_INIT_DET 0 
+#define KERNEL_BUILD_DET 1 
+#define KERNEL_SURF_DESC 2
+#define KERNEL_NORM_DESC 3
+#define KERNEL_NON_MAX_SUP 4
+#define KERNEL_GET_ORIENT1 5
+#define KERNEL_GET_ORIENT2 6
+#define KERNEL_NN 7
+#define KERNEL_SCAN 8
+#define KERNEL_SCAN4 9
+#define KERNEL_TRANSPOSE 10
+#define KERNEL_SCANIMAGE 11
+#define KERNEL_TRANSPOSEIMAGE 12
+
+#endif

benchmarks/new_opencl/guassian/gaussianElim.h

@@ -0,0 +1,40 @@
+#ifndef _GAUSSIANELIM
+#define _GAUSSIANELIM
+
+#include <iostream>
+#include <vector>
+#include <float.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <algorithm>
+
+#include "clutils.h"
+
+// All OpenCL headers
+#if defined (__APPLE__) || defined(MACOSX)
+    #include <OpenCL/opencl.h>
+#else
+    #include <CL/opencl.h>
+#endif 
+
+float *OpenClGaussianElimination(
+	cl_context context,
+	int timing);
+
+void printUsage();
+int parseCommandline(int argc, char *argv[], char* filename,
+                     int *q, int *t, int *p, int *d);
+                     
+void InitPerRun(int size,float *m);
+void ForwardSub(cl_context context, float *a, float *b, float *m, int size,int timing);
+void BackSub(float *a, float *b, float *finalVec, int size);
+void Fan1(float *m, float *a, int Size, int t);
+void Fan2(float *m, float *a, float *b,int Size, int j1, int t);
+//void Fan3(float *m, float *b, int Size, int t);
+void InitMat(FILE *fp, int size, float *ary, int nrow, int ncol);
+void InitAry(FILE *fp, float *ary, int ary_size);
+void PrintMat(float *ary, int size, int nrow, int ncolumn);
+void PrintAry(float *ary, int ary_size);
+float eventTime(cl_event event,cl_command_queue command_queue);
+#endif

benchmarks/new_opencl/guassian/gettimeofday.cpp

@@ -0,0 +1,74 @@
+#include "stdio.h"
+#include <time.h>
+#include <windows.h>
+#include <iostream> 
+//using namespace System;
+using namespace std;
+ 
+#if defined(_MSC_VER) || defined(_MSC_EXTENSIONS)
+  #define DELTA_EPOCH_IN_MICROSECS  11644473600000000Ui64
+#else
+  #define DELTA_EPOCH_IN_MICROSECS  11644473600000000ULL
+#endif
+ 
+struct timezone
+{
+  int  tz_minuteswest; /* minutes W of Greenwich */
+  int  tz_dsttime;     /* type of dst correction */
+};
+ 
+
+// Definition of a gettimeofday function
+ int gettimeofday(struct timeval *tv, struct timezone *tz)
+{
+// Define a structure to receive the current Windows filetime
+  FILETIME ft;
+ 
+// Initialize the present time to 0 and the timezone to UTC
+  unsigned __int64 tmpres = 0;
+  static int tzflag = 0;
+ 
+  if (NULL != tv)
+  {
+    GetSystemTimeAsFileTime(&ft);
+ 
+// The GetSystemTimeAsFileTime returns the number of 100 nanosecond 
+// intervals since Jan 1, 1601 in a structure. Copy the high bits to 
+// the 64 bit tmpres, shift it left by 32 then or in the low 32 bits.
+    tmpres |= ft.dwHighDateTime;
+    tmpres <<= 32;
+    tmpres |= ft.dwLowDateTime;
+ 
+// Convert to microseconds by dividing by 10
+    tmpres /= 10;
+ 
+// The Unix epoch starts on Jan 1 1970.  Need to subtract the difference 
+// in seconds from Jan 1 1601.
+    tmpres -= DELTA_EPOCH_IN_MICROSECS;
+ 
+// Finally change microseconds to seconds and place in the seconds value. 
+// The modulus picks up the microseconds.
+    tv->tv_sec = (long)(tmpres / 1000000UL);
+    tv->tv_usec = (long)(tmpres % 1000000UL);
+  }
+ 
+  if (NULL != tz)
+  {
+    if (!tzflag)
+    {
+      _tzset();
+      tzflag++;
+    }
+  
+// Adjust for the timezone west of Greenwich
+	long seconds_diff;
+	_get_timezone(&seconds_diff);
+      tz->tz_minuteswest = seconds_diff / 60;
+	int hours_offset;
+	_get_daylight(&hours_offset);
+    tz->tz_dsttime = hours_offset;
+  }
+ 
+  return 0;
+}
+    

benchmarks/new_opencl/guassian/gettimeofday.h

@@ -0,0 +1,17 @@
+
+#ifdef _WIN32
+#include <WinSock.h>
+/**
+Based on code seen at.
+
+http://www.winehq.org/pipermail/wine-devel/2003-June/018082.html
+
+http://msdn.microsoft.com/en-us/library/ms740560
+
+*/
+int gettimeofday(struct timeval *tv, struct timezone *tz);
+#else
+#include <sys/time.h>
+#endif
+
+

benchmarks/new_opencl/guassian/kernel.cl

@@ -0,0 +1,49 @@
+//#pragma OPENCL EXTENSION cl_khr_byte_addressable_store : enable
+
+typedef struct latLong
+    {
+        float lat;
+        float lng;
+    } LatLong;
+
+__kernel void Fan1(__global float *m_dev,
+                  __global float *a_dev,
+                  __global float *b_dev,
+                  const int size,
+                  const int t) {
+    int globalId = get_global_id(0);
+                              
+    if (globalId < size-1-t) {
+         *(m_dev + size * (globalId + t + 1)+t) = *(a_dev + size * (globalId + t + 1) + t) / *(a_dev + size * t + t);    
+    }
+}
+
+
+__kernel void Fan2(__global float *m_dev,
+                  __global float *a_dev,
+                  __global float *b_dev,
+                  const int size,
+                  const int t) {
+	 int globalId = get_global_id(0);
+	 
+	 int globalIdx = get_global_id(0);
+	 int globalIdy = get_global_id(1);
+      if (globalIdx < size-1-t && globalIdy < size-t) {
+         a_dev[size*(globalIdx+1+t)+(globalIdy+t)] -= m_dev[size*(globalIdx+1+t)+t] * a_dev[size*t+(globalIdy+t)];
+ 	 
+ 	    if(globalIdy == 0){
+ 		   b_dev[globalIdx+1+t] -= m_dev[size*(globalIdx+1+t)+(globalIdy+t)] * b_dev[t];
+ 	    }
+ 	 }
+//   One dimensional
+// 	 int globalIdx = globalId % size;
+// 	 int globalIdy = globalId / size;
+// 	 
+// 	 if (globalIdx < size-1-t && globalIdy < size-t) {
+//          a_dev[size*(globalIdx+1+t)+(globalIdy+t)] -= m_dev[size*(globalIdx+1+t)+t] * a_dev[size*t+(globalIdy+t)];
+// 	 }
+// 	 if(globalIdy == 0){
+//  		   b_dev[globalIdx+1+t] -= m_dev[size*(globalIdx+1+t)+(globalIdy+t)] * b_dev[t];
+//      }
+    
+}

benchmarks/new_opencl/guassian/main.cc

@@ -0,0 +1,411 @@
+#ifndef __GAUSSIAN_ELIMINATION__
+#define __GAUSSIAN_ELIMINATION__
+
+#include "gaussianElim.h"
+
+cl_context context = NULL;
+
+int main(int argc, char *argv[]) {
+  printf("enter demo main\n");
+  float *a = NULL, *b = NULL, *finalVec = NULL;
+  float *m = NULL;
+  int size;
+
+  FILE *fp;
+
+  // args
+  char filename[100];
+  int quiet = 0, timing = 0, platform = -1, device = -1;
+
+  // parse command line
+  if (parseCommandline(argc, argv, filename, &quiet, &timing, &platform,
+                       &device)) {
+    printUsage();
+    return 0;
+  }
+
+  context = cl_init_context(platform, device, quiet);
+
+  fp = fopen(filename, "r");
+  fscanf(fp, "%d", &size);
+
+  a = (float *)malloc(size * size * sizeof(float));
+
+  printf("OK\n");
+
+  InitMat(fp, size, a, size, size);
+  // printf("The input matrix a is:\n");
+  // PrintMat(a, size, size, size);
+  b = (float *)malloc(size * sizeof(float));
+
+  InitAry(fp, b, size);
+  // printf("The input array b is:\n");
+  // PrintAry(b, size);
+
+  // create the solution matrix
+  m = (float *)malloc(size * size * sizeof(float));
+
+  // create a new vector to hold the final answer
+  finalVec = (float *)malloc(size * sizeof(float));
+
+  InitPerRun(size, m);
+
+  // begin timing
+
+  // run kernels
+  ForwardSub(context, a, b, m, size, timing);
+
+  // end timing
+  if (!quiet) {
+    printf("The result of matrix m is: \n");
+
+    PrintMat(m, size, size, size);
+    printf("The result of matrix a is: \n");
+    PrintMat(a, size, size, size);
+    printf("The result of array b is: \n");
+    PrintAry(b, size);
+
+    BackSub(a, b, finalVec, size);
+    printf("The final solution is: \n");
+    PrintAry(finalVec, size);
+  }
+
+  fclose(fp);
+  free(m);
+  free(a);
+  free(b);
+  free(finalVec);
+  // OpenClGaussianElimination(context,timing);
+
+  return 0;
+}
+
+/*------------------------------------------------------
+ ** ForwardSub() -- Forward substitution of Gaussian
+ ** elimination.
+ **------------------------------------------------------
+ */
+void ForwardSub(cl_context context, float *a, float *b, float *m, int size,
+                int timing) {
+  // 1. set up kernels
+  cl_kernel fan1_kernel, fan2_kernel;
+  cl_int status = 0;
+  cl_program gaussianElim_program;
+  cl_event writeEvent, kernelEvent, readEvent;
+  float writeTime = 0, readTime = 0, kernelTime = 0;
+  float writeMB = 0, readMB = 0;
+  
+  gaussianElim_program = cl_compileProgram((char *)"gaussianElim_kernels.cl", NULL);
+  
+  fan1_kernel = clCreateKernel(gaussianElim_program, "Fan1", &status);
+  status = cl_errChk(status, (char *)"Error Creating Fan1 kernel", true);
+  if (status)
+    exit(1);
+
+  fan2_kernel = clCreateKernel(gaussianElim_program, "Fan2", &status);
+  status = cl_errChk(status, (char *)"Error Creating Fan2 kernel", true);
+  if (status)
+    exit(1);
+
+  // 2. set up memory on device and send ipts data to device
+
+  cl_mem a_dev, b_dev, m_dev;
+
+  cl_int error = 0;
+
+  a_dev = clCreateBuffer(context, CL_MEM_READ_WRITE,
+                         sizeof(float) * size * size, NULL, &error);
+
+  b_dev = clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(float) * size, NULL,
+                         &error);
+
+  m_dev = clCreateBuffer(context, CL_MEM_READ_WRITE,
+                         sizeof(float) * size * size, NULL, &error);
+
+  cl_command_queue command_queue = cl_getCommandQueue();
+
+  error = clEnqueueWriteBuffer(command_queue, a_dev,
+                               1, // change to 0 for nonblocking write
+                               0, // offset
+                               sizeof(float) * size * size, a, 0, NULL,
+                               &writeEvent);
+
+  if (timing)
+    writeTime += eventTime(writeEvent, command_queue);
+  clReleaseEvent(writeEvent);
+
+  error = clEnqueueWriteBuffer(command_queue, b_dev,
+                               1, // change to 0 for nonblocking write
+                               0, // offset
+                               sizeof(float) * size, b, 0, NULL, &writeEvent);
+  if (timing)
+    writeTime += eventTime(writeEvent, command_queue);
+  clReleaseEvent(writeEvent);
+
+  error = clEnqueueWriteBuffer(command_queue, m_dev,
+                               1, // change to 0 for nonblocking write
+                               0, // offset
+                               sizeof(float) * size * size, m, 0, NULL,
+                               &writeEvent);
+  if (timing)
+    writeTime += eventTime(writeEvent, command_queue);
+  clReleaseEvent(writeEvent);
+  writeMB = (float)(sizeof(float) * size * (size + size + 1) / 1e6);
+
+  // 3. Determine block sizes
+  size_t globalWorksizeFan1[1];
+  size_t globalWorksizeFan2[2];
+
+  globalWorksizeFan1[0] = size;
+  globalWorksizeFan2[0] = size;
+  globalWorksizeFan2[1] = size;
+
+  int t;
+  // 4. Setup and Run kernels
+  for (t = 0; t < (size - 1); t++) {
+    // kernel args
+    cl_int argchk;
+    argchk = clSetKernelArg(fan1_kernel, 0, sizeof(cl_mem), (void *)&m_dev);
+    argchk |= clSetKernelArg(fan1_kernel, 1, sizeof(cl_mem), (void *)&a_dev);
+    argchk |= clSetKernelArg(fan1_kernel, 2, sizeof(cl_mem), (void *)&b_dev);
+    argchk |= clSetKernelArg(fan1_kernel, 3, sizeof(int), (void *)&size);
+    argchk |= clSetKernelArg(fan1_kernel, 4, sizeof(int), (void *)&t);
+
+    cl_errChk(argchk, "ERROR in Setting Fan1 kernel args", true);
+
+    // launch kernel
+    error =
+        clEnqueueNDRangeKernel(command_queue, fan1_kernel, 1, 0,
+                               globalWorksizeFan1, NULL, 0, NULL, &kernelEvent);
+
+    cl_errChk(error, "ERROR in Executing Fan1 Kernel", true);
+    if (timing) {
+      //             printf("here1a\n");
+      kernelTime += eventTime(kernelEvent, command_queue);
+      //             printf("here1b\n");
+    }
+    clReleaseEvent(kernelEvent);
+    // Fan1<<<dimGrid,dimBlock>>>(m_cuda,a_cuda,Size,t);
+    // cudaThreadSynchronize();
+
+    // kernel args
+    argchk = clSetKernelArg(fan2_kernel, 0, sizeof(cl_mem), (void *)&m_dev);
+    argchk |= clSetKernelArg(fan2_kernel, 1, sizeof(cl_mem), (void *)&a_dev);
+    argchk |= clSetKernelArg(fan2_kernel, 2, sizeof(cl_mem), (void *)&b_dev);
+    argchk |= clSetKernelArg(fan2_kernel, 3, sizeof(int), (void *)&size);
+    argchk |= clSetKernelArg(fan2_kernel, 4, sizeof(int), (void *)&t);
+
+    cl_errChk(argchk, "ERROR in Setting Fan2 kernel args", true);
+
+    // launch kernel
+    error =
+        clEnqueueNDRangeKernel(command_queue, fan2_kernel, 2, 0,
+                               globalWorksizeFan2, NULL, 0, NULL, &kernelEvent);
+
+    cl_errChk(error, "ERROR in Executing Fan1 Kernel", true);
+    if (timing) {
+      //             printf("here2a\n");
+      kernelTime += eventTime(kernelEvent, command_queue);
+      //             printf("here2b\n");
+    }
+    clReleaseEvent(kernelEvent);
+    // Fan2<<<dimGridXY,dimBlockXY>>>(m_cuda,a_cuda,b_cuda,Size,Size-t,t);
+    // cudaThreadSynchronize();
+  }
+  // 5. transfer data off of device
+  error =
+      clEnqueueReadBuffer(command_queue, a_dev,
+                          1, // change to 0 for nonblocking write
+                          0, // offset
+                          sizeof(float) * size * size, a, 0, NULL, &readEvent);
+
+  cl_errChk(error, "ERROR with clEnqueueReadBuffer", true);
+  if (timing)
+    readTime += eventTime(readEvent, command_queue);
+  clReleaseEvent(readEvent);
+
+  error = clEnqueueReadBuffer(command_queue, b_dev,
+                              1, // change to 0 for nonblocking write
+                              0, // offset
+                              sizeof(float) * size, b, 0, NULL, &readEvent);
+  cl_errChk(error, "ERROR with clEnqueueReadBuffer", true);
+  if (timing)
+    readTime += eventTime(readEvent, command_queue);
+  clReleaseEvent(readEvent);
+
+  error =
+      clEnqueueReadBuffer(command_queue, m_dev,
+                          1, // change to 0 for nonblocking write
+                          0, // offset
+                          sizeof(float) * size * size, m, 0, NULL, &readEvent);
+
+  cl_errChk(error, "ERROR with clEnqueueReadBuffer", true);
+  if (timing)
+    readTime += eventTime(readEvent, command_queue);
+  clReleaseEvent(readEvent);
+  readMB = (float)(sizeof(float) * size * (size + size + 1) / 1e6);
+
+  if (timing) {
+    printf("Matrix Size\tWrite(s) [size]\t\tKernel(s)\tRead(s)  "
+           "[size]\t\tTotal(s)\n");
+    printf("%dx%d      \t", size, size);
+
+    printf("%f [%.2fMB]\t", writeTime, writeMB);
+
+    printf("%f\t", kernelTime);
+
+    printf("%f [%.2fMB]\t", readTime, readMB);
+
+    printf("%f\n\n", writeTime + kernelTime + readTime);
+  }
+}
+
+float eventTime(cl_event event, cl_command_queue command_queue) {
+  cl_int error = 0;
+  cl_ulong eventStart, eventEnd;
+  clFinish(command_queue);
+  error = clGetEventProfilingInfo(event, CL_PROFILING_COMMAND_START,
+                                  sizeof(cl_ulong), &eventStart, NULL);
+  cl_errChk(error, "ERROR in Event Profiling.", true);
+  error = clGetEventProfilingInfo(event, CL_PROFILING_COMMAND_END,
+                                  sizeof(cl_ulong), &eventEnd, NULL);
+  cl_errChk(error, "ERROR in Event Profiling.", true);
+
+  return (float)((eventEnd - eventStart) / 1e9);
+}
+
+int parseCommandline(int argc, char *argv[], char *filename, int *q, int *t,
+                     int *p, int *d) {
+  int i;
+  // if (argc < 2) return 1; // error
+  strncpy(filename, "matrix4.txt", 100);
+  char flag;
+
+  for (i = 1; i < argc; i++) {
+    if (argv[i][0] == '-') { // flag
+      flag = argv[i][1];
+      switch (flag) {
+      case 'h': // help
+        return 1;
+        break;
+      case 'q': // quiet
+        *q = 1;
+        break;
+      case 't': // timing
+        *t = 1;
+        break;
+      case 'p': // platform
+        i++;
+        *p = atoi(argv[i]);
+        break;
+      case 'd': // device
+        i++;
+        *d = atoi(argv[i]);
+        break;
+      }
+    }
+  }
+  if ((*d >= 0 && *p < 0) ||
+      (*p >= 0 &&
+       *d < 0)) // both p and d must be specified if either are specified
+    return 1;
+  return 0;
+}
+
+void printUsage() {
+  printf("Gaussian Elimination Usage\n");
+  printf("\n");
+  printf("gaussianElimination [filename] [-hqt] [-p [int] -d [int]]\n");
+  printf("\n");
+  printf("example:\n");
+  printf("$ ./gaussianElimination matrix4.txt\n");
+  printf("\n");
+  printf("filename     the filename that holds the matrix data\n");
+  printf("\n");
+  printf("-h           Display the help file\n");
+  printf("-q           Quiet mode. Suppress all text output.\n");
+  printf("-t           Print timing information.\n");
+  printf("\n");
+  printf("-p [int]     Choose the platform (must choose both platform and "
+         "device)\n");
+  printf("-d [int]     Choose the device (must choose both platform and "
+         "device)\n");
+  printf("\n");
+  printf("\n");
+  printf("Notes: 1. The filename is required as the first parameter.\n");
+  printf("       2. If you declare either the device or the platform,\n");
+  printf("          you must declare both.\n\n");
+}
+
+/*------------------------------------------------------
+ ** InitPerRun() -- Initialize the contents of the
+ ** multipier matrix **m
+ **------------------------------------------------------
+ */
+void InitPerRun(int size, float *m) {
+  int i;
+  for (i = 0; i < size * size; i++)
+    *(m + i) = 0.0;
+}
+void BackSub(float *a, float *b, float *finalVec, int size) {
+  // solve "bottom up"
+  int i, j;
+  for (i = 0; i < size; i++) {
+    finalVec[size - i - 1] = b[size - i - 1];
+    for (j = 0; j < i; j++) {
+      finalVec[size - i - 1] -= *(a + size * (size - i - 1) + (size - j - 1)) *
+                                finalVec[size - j - 1];
+    }
+    finalVec[size - i - 1] =
+        finalVec[size - i - 1] / *(a + size * (size - i - 1) + (size - i - 1));
+  }
+}
+void InitMat(FILE *fp, int size, float *ary, int nrow, int ncol) {
+  int i, j;
+
+  for (i = 0; i < nrow; i++) {
+    for (j = 0; j < ncol; j++) {
+      fscanf(fp, "%f", ary + size * i + j);
+    }
+  }
+}
+/*------------------------------------------------------
+ ** InitAry() -- Initialize the array (vector) by reading
+ ** data from the data file
+ **------------------------------------------------------
+ */
+void InitAry(FILE *fp, float *ary, int ary_size) {
+  int i;
+
+  for (i = 0; i < ary_size; i++) {
+    fscanf(fp, "%f", &ary[i]);
+  }
+}
+/*------------------------------------------------------
+ ** PrintMat() -- Print the contents of the matrix
+ **------------------------------------------------------
+ */
+void PrintMat(float *ary, int size, int nrow, int ncol) {
+  int i, j;
+
+  for (i = 0; i < nrow; i++) {
+    for (j = 0; j < ncol; j++) {
+      printf("%8.2f ", *(ary + size * i + j));
+    }
+    printf("\n");
+  }
+  printf("\n");
+}
+
+/*------------------------------------------------------
+ ** PrintAry() -- Print the contents of the array (vector)
+ **------------------------------------------------------
+ */
+void PrintAry(float *ary, int ary_size) {
+  int i;
+  for (i = 0; i < ary_size; i++) {
+    printf("%.2f ", ary[i]);
+  }
+  printf("\n\n");
+}
+#endif

benchmarks/new_opencl/guassian/matrix4.txt

@@ -0,0 +1,11 @@
+4
+
+-0.6	-0.5	0.7	0.3	
+-0.3	-0.9	0.3	0.7	
+-0.4	-0.5	-0.3	-0.8	
+0.0	-0.1	0.2	0.9	
+
+-0.85	-0.68	0.24	-0.53	
+
+0.7	0.0	-0.4	-0.5	
+

benchmarks/new_opencl/guassian/run

@@ -0,0 +1 @@
+./gaussian ../../data/gaussian/matrix4.txt

benchmarks/new_opencl/guassian/utils.cpp

@@ -0,0 +1,204 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#include <stdio.h>
+#include <sys/stat.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include "utils.h"
+
+static bool usingImages = true;
+
+//! A wrapper for malloc that checks the return value
+void* alloc(size_t size) {
+
+    void* ptr = NULL;
+    ptr = malloc(size);
+    if(ptr == NULL) {
+        perror("malloc");
+        exit(-1);
+    }
+
+    return ptr;
+}
+
+// This function checks to make sure a file exists before we open it
+void checkFile(char* filename) 
+{
+    
+    struct stat fileStatus;
+    if(stat(filename, &fileStatus) != 0) {
+        printf("Error opening file: %s\n", filename);
+        exit(-1);
+    }
+    else {
+        if(!(S_IFREG & fileStatus.st_mode)) {
+            printf("File %s is not a regular file\n", filename);
+            exit(-1);
+        }
+    }
+}
+
+
+// This function checks to make sure a directory exists 
+void checkDir(char* dirpath) 
+{
+    
+    struct stat fileStatus;
+    if(stat(dirpath, &fileStatus) != 0) {
+        printf("Directory does not exist: %s\n", dirpath);
+        exit(-1);
+    }
+    else {
+        if(!(S_IFDIR & fileStatus.st_mode)) {
+            printf("Directory was not provided: %s\n", dirpath);
+            exit(-1);
+        }
+    }
+}
+
+// Parse the command line arguments
+void parseArguments(int argc, char** argv, char** input, char** events, 
+    char** ipts, char* devicePref, bool* verifyResults) 
+{
+    
+    for(int i = 2; i < argc; i++) {
+        if(strcmp(argv[i], "-d") == 0) {   // Event dump found
+            if(i == argc-1) {
+                printf("Usage: -e Needs directory path\n");
+                exit(-1);
+            }
+            devicePref[0] = argv[i+1][0];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-e") == 0) {   // Event dump found
+            if(i == argc-1) {
+                printf("Usage: -e Needs directory path\n");
+                exit(-1);
+            }
+            *events = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-i") == 0) {   // Input found
+            if(i == argc-1) {
+                printf("Usage: -i Needs directory path\n");
+                exit(-1);
+            }
+            *input = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-l") == 0) {   // Ipts dump found
+            if(i == argc-1) {
+                printf("Usage: -l Needs directory path\n");
+                exit(-1);
+            }
+            *ipts = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-n") == 0) {   // Don't use OpenCL images
+            setUsingImages(false);
+            continue;
+        }
+        if(strcmp(argv[i], "-v") == 0) {   // Verify results
+            *verifyResults = true;
+            continue;
+        }
+    }
+}
+
+
+// This function that takes a positive integer 'value' and returns
+// the nearest multiple of 'multiple' (used for padding columns)
+unsigned int roundUp(unsigned int value, unsigned int multiple) {
+
+   unsigned int remainder = value % multiple;
+   
+   // Make the value a multiple of multiple
+   if(remainder != 0) {
+      value += (multiple-remainder);
+   }
+
+   return value;
+}
+
+
+// Concatenate two strings and return a pointer to the new string
+char* smartStrcat(char* str1, char* str2) 
+{
+    char* newStr = NULL;
+
+    newStr = (char*)alloc((strlen(str1)+strlen(str2)+1)*sizeof(char));
+
+    strcpy(newStr, str1);
+    strcat(newStr, str2);
+
+    return newStr;
+}
+
+
+// Set the value of using images to true if they are being
+// used, or false if they are not
+void setUsingImages(bool val) 
+{
+    usingImages = val;
+}
+
+
+// Return whether or not images are being used
+bool isUsingImages() 
+{
+    return usingImages;
+}

benchmarks/new_opencl/guassian/utils.h

@@ -0,0 +1,84 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#ifndef _UTILS_
+#define _UTILS_
+
+// Wrapper for malloc
+void* alloc(size_t size);
+
+// Checks for existence of directory
+void checkDir(char* dirpath);
+
+// Check for existence of file
+void checkFile(char* filename);
+
+// Parse the input command line options to the program
+void parseArguments(int argc, char** argv, char** input, char** events, 
+    char** ipts, char* devicePref, bool* verifyResults);
+
+
+// Print the program usage information
+void printUsage();
+
+// Rounds up size to the nearest multiple of multiple
+unsigned int roundUp(unsigned int value, unsigned int multiple);
+
+// Concatenate two strings, creating a new one
+char* smartStrcat(char* str1, char* str2);
+
+// Set the value of usingImages
+void setUsingImages(bool val);
+
+// Return whether or not images are being used
+bool isUsingImages(); 
+
+#endif

benchmarks/new_opencl/include/CL/cl_d3d10.h

@@ -0,0 +1,131 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2015 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+
+/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */
+
+#ifndef __OPENCL_CL_D3D10_H
+#define __OPENCL_CL_D3D10_H
+
+#include <d3d10.h>
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/******************************************************************************
+ * cl_khr_d3d10_sharing                                                       */
+#define cl_khr_d3d10_sharing 1
+
+typedef cl_uint cl_d3d10_device_source_khr;
+typedef cl_uint cl_d3d10_device_set_khr;
+
+/******************************************************************************/
+
+/* Error Codes */
+#define CL_INVALID_D3D10_DEVICE_KHR                  -1002
+#define CL_INVALID_D3D10_RESOURCE_KHR                -1003
+#define CL_D3D10_RESOURCE_ALREADY_ACQUIRED_KHR       -1004
+#define CL_D3D10_RESOURCE_NOT_ACQUIRED_KHR           -1005
+
+/* cl_d3d10_device_source_nv */
+#define CL_D3D10_DEVICE_KHR                          0x4010
+#define CL_D3D10_DXGI_ADAPTER_KHR                    0x4011
+
+/* cl_d3d10_device_set_nv */
+#define CL_PREFERRED_DEVICES_FOR_D3D10_KHR           0x4012
+#define CL_ALL_DEVICES_FOR_D3D10_KHR                 0x4013
+
+/* cl_context_info */
+#define CL_CONTEXT_D3D10_DEVICE_KHR                  0x4014
+#define CL_CONTEXT_D3D10_PREFER_SHARED_RESOURCES_KHR 0x402C
+
+/* cl_mem_info */
+#define CL_MEM_D3D10_RESOURCE_KHR                    0x4015
+
+/* cl_image_info */
+#define CL_IMAGE_D3D10_SUBRESOURCE_KHR               0x4016
+
+/* cl_command_type */
+#define CL_COMMAND_ACQUIRE_D3D10_OBJECTS_KHR         0x4017
+#define CL_COMMAND_RELEASE_D3D10_OBJECTS_KHR         0x4018
+
+/******************************************************************************/
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromD3D10KHR_fn)(
+    cl_platform_id             platform,
+    cl_d3d10_device_source_khr d3d_device_source,
+    void *                     d3d_object,
+    cl_d3d10_device_set_khr    d3d_device_set,
+    cl_uint                    num_entries,
+    cl_device_id *             devices,
+    cl_uint *                  num_devices) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10BufferKHR_fn)(
+    cl_context     context,
+    cl_mem_flags   flags,
+    ID3D10Buffer * resource,
+    cl_int *       errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10Texture2DKHR_fn)(
+    cl_context        context,
+    cl_mem_flags      flags,
+    ID3D10Texture2D * resource,
+    UINT              subresource,
+    cl_int *          errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10Texture3DKHR_fn)(
+    cl_context        context,
+    cl_mem_flags      flags,
+    ID3D10Texture3D * resource,
+    UINT              subresource,
+    cl_int *          errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireD3D10ObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseD3D10ObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_0;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_D3D10_H */
+

benchmarks/new_opencl/include/CL/cl_d3d11.h

@@ -0,0 +1,131 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2015 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+
+/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */
+
+#ifndef __OPENCL_CL_D3D11_H
+#define __OPENCL_CL_D3D11_H
+
+#include <d3d11.h>
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/******************************************************************************
+ * cl_khr_d3d11_sharing                                                       */
+#define cl_khr_d3d11_sharing 1
+
+typedef cl_uint cl_d3d11_device_source_khr;
+typedef cl_uint cl_d3d11_device_set_khr;
+
+/******************************************************************************/
+
+/* Error Codes */
+#define CL_INVALID_D3D11_DEVICE_KHR                  -1006
+#define CL_INVALID_D3D11_RESOURCE_KHR                -1007
+#define CL_D3D11_RESOURCE_ALREADY_ACQUIRED_KHR       -1008
+#define CL_D3D11_RESOURCE_NOT_ACQUIRED_KHR           -1009
+
+/* cl_d3d11_device_source */
+#define CL_D3D11_DEVICE_KHR                          0x4019
+#define CL_D3D11_DXGI_ADAPTER_KHR                    0x401A
+
+/* cl_d3d11_device_set */
+#define CL_PREFERRED_DEVICES_FOR_D3D11_KHR           0x401B
+#define CL_ALL_DEVICES_FOR_D3D11_KHR                 0x401C
+
+/* cl_context_info */
+#define CL_CONTEXT_D3D11_DEVICE_KHR                  0x401D
+#define CL_CONTEXT_D3D11_PREFER_SHARED_RESOURCES_KHR 0x402D
+
+/* cl_mem_info */
+#define CL_MEM_D3D11_RESOURCE_KHR                    0x401E
+
+/* cl_image_info */
+#define CL_IMAGE_D3D11_SUBRESOURCE_KHR               0x401F
+
+/* cl_command_type */
+#define CL_COMMAND_ACQUIRE_D3D11_OBJECTS_KHR         0x4020
+#define CL_COMMAND_RELEASE_D3D11_OBJECTS_KHR         0x4021
+
+/******************************************************************************/
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromD3D11KHR_fn)(
+    cl_platform_id             platform,
+    cl_d3d11_device_source_khr d3d_device_source,
+    void *                     d3d_object,
+    cl_d3d11_device_set_khr    d3d_device_set,
+    cl_uint                    num_entries,
+    cl_device_id *             devices,
+    cl_uint *                  num_devices) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11BufferKHR_fn)(
+    cl_context     context,
+    cl_mem_flags   flags,
+    ID3D11Buffer * resource,
+    cl_int *       errcode_ret) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11Texture2DKHR_fn)(
+    cl_context        context,
+    cl_mem_flags      flags,
+    ID3D11Texture2D * resource,
+    UINT              subresource,
+    cl_int *          errcode_ret) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11Texture3DKHR_fn)(
+    cl_context        context,
+    cl_mem_flags      flags,
+    ID3D11Texture3D * resource,
+    UINT              subresource,
+    cl_int *          errcode_ret) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireD3D11ObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseD3D11ObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_2;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_D3D11_H */
+

benchmarks/new_opencl/include/CL/cl_dx9_media_sharing.h

@@ -0,0 +1,132 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2015 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+
+/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */
+
+#ifndef __OPENCL_CL_DX9_MEDIA_SHARING_H
+#define __OPENCL_CL_DX9_MEDIA_SHARING_H
+
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/******************************************************************************/
+/* cl_khr_dx9_media_sharing                                                   */
+#define cl_khr_dx9_media_sharing 1
+
+typedef cl_uint             cl_dx9_media_adapter_type_khr;
+typedef cl_uint             cl_dx9_media_adapter_set_khr;
+    
+#if defined(_WIN32)
+#include <d3d9.h>
+typedef struct _cl_dx9_surface_info_khr
+{
+    IDirect3DSurface9 *resource;
+    HANDLE shared_handle;
+} cl_dx9_surface_info_khr;
+#endif
+
+
+/******************************************************************************/
+
+/* Error Codes */
+#define CL_INVALID_DX9_MEDIA_ADAPTER_KHR                -1010
+#define CL_INVALID_DX9_MEDIA_SURFACE_KHR                -1011
+#define CL_DX9_MEDIA_SURFACE_ALREADY_ACQUIRED_KHR       -1012
+#define CL_DX9_MEDIA_SURFACE_NOT_ACQUIRED_KHR           -1013
+
+/* cl_media_adapter_type_khr */
+#define CL_ADAPTER_D3D9_KHR                              0x2020
+#define CL_ADAPTER_D3D9EX_KHR                            0x2021
+#define CL_ADAPTER_DXVA_KHR                              0x2022
+
+/* cl_media_adapter_set_khr */
+#define CL_PREFERRED_DEVICES_FOR_DX9_MEDIA_ADAPTER_KHR   0x2023
+#define CL_ALL_DEVICES_FOR_DX9_MEDIA_ADAPTER_KHR         0x2024
+
+/* cl_context_info */
+#define CL_CONTEXT_ADAPTER_D3D9_KHR                      0x2025
+#define CL_CONTEXT_ADAPTER_D3D9EX_KHR                    0x2026
+#define CL_CONTEXT_ADAPTER_DXVA_KHR                      0x2027
+
+/* cl_mem_info */
+#define CL_MEM_DX9_MEDIA_ADAPTER_TYPE_KHR                0x2028
+#define CL_MEM_DX9_MEDIA_SURFACE_INFO_KHR                0x2029
+
+/* cl_image_info */
+#define CL_IMAGE_DX9_MEDIA_PLANE_KHR                     0x202A
+
+/* cl_command_type */
+#define CL_COMMAND_ACQUIRE_DX9_MEDIA_SURFACES_KHR        0x202B
+#define CL_COMMAND_RELEASE_DX9_MEDIA_SURFACES_KHR        0x202C
+
+/******************************************************************************/
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromDX9MediaAdapterKHR_fn)(
+    cl_platform_id                   platform,
+    cl_uint                          num_media_adapters,
+    cl_dx9_media_adapter_type_khr *  media_adapter_type,
+    void *                           media_adapters,
+    cl_dx9_media_adapter_set_khr     media_adapter_set,
+    cl_uint                          num_entries,
+    cl_device_id *                   devices,
+    cl_uint *                        num_devices) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromDX9MediaSurfaceKHR_fn)(
+    cl_context                    context,
+    cl_mem_flags                  flags,
+    cl_dx9_media_adapter_type_khr adapter_type,
+    void *                        surface_info,
+    cl_uint                       plane,                                                                          
+    cl_int *                      errcode_ret) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireDX9MediaSurfacesKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseDX9MediaSurfacesKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event) CL_API_SUFFIX__VERSION_1_2;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_DX9_MEDIA_SHARING_H */
+

benchmarks/new_opencl/include/CL/cl_dx9_media_sharing_intel.h

@@ -0,0 +1,182 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+/*****************************************************************************\
+
+Copyright (c) 2013-2019 Intel Corporation All Rights Reserved.
+
+THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
+CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
+NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
+MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+File Name: cl_dx9_media_sharing_intel.h
+
+Abstract:
+
+Notes:
+
+\*****************************************************************************/
+
+#ifndef __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H
+#define __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H
+
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+#include <d3d9.h>
+#include <dxvahd.h>
+#include <wtypes.h>
+#include <d3d9types.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/***************************************
+* cl_intel_dx9_media_sharing extension *
+****************************************/
+
+#define cl_intel_dx9_media_sharing 1
+
+typedef cl_uint cl_dx9_device_source_intel;
+typedef cl_uint cl_dx9_device_set_intel;
+
+/* error codes */
+#define CL_INVALID_DX9_DEVICE_INTEL                   -1010
+#define CL_INVALID_DX9_RESOURCE_INTEL                 -1011
+#define CL_DX9_RESOURCE_ALREADY_ACQUIRED_INTEL        -1012
+#define CL_DX9_RESOURCE_NOT_ACQUIRED_INTEL            -1013
+
+/* cl_dx9_device_source_intel */
+#define CL_D3D9_DEVICE_INTEL                          0x4022
+#define CL_D3D9EX_DEVICE_INTEL                        0x4070
+#define CL_DXVA_DEVICE_INTEL                          0x4071
+
+/* cl_dx9_device_set_intel */
+#define CL_PREFERRED_DEVICES_FOR_DX9_INTEL            0x4024
+#define CL_ALL_DEVICES_FOR_DX9_INTEL                  0x4025
+
+/* cl_context_info */
+#define CL_CONTEXT_D3D9_DEVICE_INTEL                  0x4026
+#define CL_CONTEXT_D3D9EX_DEVICE_INTEL                0x4072
+#define CL_CONTEXT_DXVA_DEVICE_INTEL                  0x4073
+
+/* cl_mem_info */
+#define CL_MEM_DX9_RESOURCE_INTEL                     0x4027
+#define CL_MEM_DX9_SHARED_HANDLE_INTEL                0x4074
+
+/* cl_image_info */
+#define CL_IMAGE_DX9_PLANE_INTEL                      0x4075
+
+/* cl_command_type */
+#define CL_COMMAND_ACQUIRE_DX9_OBJECTS_INTEL          0x402A
+#define CL_COMMAND_RELEASE_DX9_OBJECTS_INTEL          0x402B
+/******************************************************************************/
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetDeviceIDsFromDX9INTEL(
+    cl_platform_id              platform,
+    cl_dx9_device_source_intel  dx9_device_source,
+    void*                       dx9_object,
+    cl_dx9_device_set_intel     dx9_device_set,
+    cl_uint                     num_entries,
+    cl_device_id*               devices,
+    cl_uint*                    num_devices) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL* clGetDeviceIDsFromDX9INTEL_fn)(
+    cl_platform_id              platform,
+    cl_dx9_device_source_intel  dx9_device_source,
+    void*                       dx9_object,
+    cl_dx9_device_set_intel     dx9_device_set,
+    cl_uint                     num_entries,
+    cl_device_id*               devices,
+    cl_uint*                    num_devices) CL_EXT_SUFFIX__VERSION_1_1;
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromDX9MediaSurfaceINTEL(
+    cl_context                  context,
+    cl_mem_flags                flags,
+    IDirect3DSurface9*          resource,
+    HANDLE                      sharedHandle,
+    UINT                        plane,
+    cl_int*                     errcode_ret) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromDX9MediaSurfaceINTEL_fn)(
+    cl_context                  context,
+    cl_mem_flags                flags,
+    IDirect3DSurface9*          resource,
+    HANDLE                      sharedHandle,
+    UINT                        plane,
+    cl_int*                     errcode_ret) CL_EXT_SUFFIX__VERSION_1_1;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueAcquireDX9ObjectsINTEL(
+    cl_command_queue            command_queue,
+    cl_uint                     num_objects,
+    const cl_mem*               mem_objects,
+    cl_uint                     num_events_in_wait_list,
+    const cl_event*             event_wait_list,
+    cl_event*                   event) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireDX9ObjectsINTEL_fn)(
+    cl_command_queue            command_queue,
+    cl_uint                     num_objects,
+    const cl_mem*               mem_objects,
+    cl_uint                     num_events_in_wait_list,
+    const cl_event*             event_wait_list,
+    cl_event*                   event) CL_EXT_SUFFIX__VERSION_1_1;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueReleaseDX9ObjectsINTEL(
+    cl_command_queue            command_queue,
+    cl_uint                     num_objects,
+    cl_mem*                     mem_objects,
+    cl_uint                     num_events_in_wait_list,
+    const cl_event*             event_wait_list,
+    cl_event*                   event) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseDX9ObjectsINTEL_fn)(
+    cl_command_queue            command_queue,
+    cl_uint                     num_objects,
+    cl_mem*                     mem_objects,
+    cl_uint                     num_events_in_wait_list,
+    const cl_event*             event_wait_list,
+    cl_event*                   event) CL_EXT_SUFFIX__VERSION_1_1;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H */
+

benchmarks/new_opencl/include/CL/cl_egl.h

@@ -0,0 +1,132 @@
+/*******************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ ******************************************************************************/
+
+#ifndef __OPENCL_CL_EGL_H
+#define __OPENCL_CL_EGL_H
+
+#include <CL/cl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/* Command type for events created with clEnqueueAcquireEGLObjectsKHR */
+#define CL_COMMAND_EGL_FENCE_SYNC_OBJECT_KHR  0x202F
+#define CL_COMMAND_ACQUIRE_EGL_OBJECTS_KHR    0x202D
+#define CL_COMMAND_RELEASE_EGL_OBJECTS_KHR    0x202E
+
+/* Error type for clCreateFromEGLImageKHR */
+#define CL_INVALID_EGL_OBJECT_KHR             -1093
+#define CL_EGL_RESOURCE_NOT_ACQUIRED_KHR      -1092
+
+/* CLeglImageKHR is an opaque handle to an EGLImage */
+typedef void* CLeglImageKHR;
+
+/* CLeglDisplayKHR is an opaque handle to an EGLDisplay */
+typedef void* CLeglDisplayKHR;
+
+/* CLeglSyncKHR is an opaque handle to an EGLSync object */
+typedef void* CLeglSyncKHR;
+
+/* properties passed to clCreateFromEGLImageKHR */
+typedef intptr_t cl_egl_image_properties_khr;
+
+
+#define cl_khr_egl_image 1
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromEGLImageKHR(cl_context                  context,
+                        CLeglDisplayKHR             egldisplay,
+                        CLeglImageKHR               eglimage,
+                        cl_mem_flags                flags,
+                        const cl_egl_image_properties_khr * properties,
+                        cl_int *                    errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromEGLImageKHR_fn)(
+    cl_context                  context,
+    CLeglDisplayKHR             egldisplay,
+    CLeglImageKHR               eglimage,
+    cl_mem_flags                flags,
+    const cl_egl_image_properties_khr * properties,
+    cl_int *                    errcode_ret);
+
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueAcquireEGLObjectsKHR(cl_command_queue command_queue,
+                              cl_uint          num_objects,
+                              const cl_mem *   mem_objects,
+                              cl_uint          num_events_in_wait_list,
+                              const cl_event * event_wait_list,
+                              cl_event *       event) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireEGLObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event);
+
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueReleaseEGLObjectsKHR(cl_command_queue command_queue,
+                              cl_uint          num_objects,
+                              const cl_mem *   mem_objects,
+                              cl_uint          num_events_in_wait_list,
+                              const cl_event * event_wait_list,
+                              cl_event *       event) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseEGLObjectsKHR_fn)(
+    cl_command_queue command_queue,
+    cl_uint          num_objects,
+    const cl_mem *   mem_objects,
+    cl_uint          num_events_in_wait_list,
+    const cl_event * event_wait_list,
+    cl_event *       event);
+
+
+#define cl_khr_egl_event 1
+
+extern CL_API_ENTRY cl_event CL_API_CALL
+clCreateEventFromEGLSyncKHR(cl_context      context,
+                            CLeglSyncKHR    sync,
+                            CLeglDisplayKHR display,
+                            cl_int *        errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_event (CL_API_CALL *clCreateEventFromEGLSyncKHR_fn)(
+    cl_context      context,
+    CLeglSyncKHR    sync,
+    CLeglDisplayKHR display,
+    cl_int *        errcode_ret);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __OPENCL_CL_EGL_H */

benchmarks/new_opencl/include/CL/cl_ext.h

@@ -0,0 +1,762 @@
+/*******************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ ******************************************************************************/
+
+/* cl_ext.h contains OpenCL extensions which don't have external */
+/* (OpenGL, D3D) dependencies.                                   */
+
+#ifndef __CL_EXT_H
+#define __CL_EXT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <CL/cl.h>
+
+/* cl_khr_fp64 extension - no extension #define since it has no functions  */
+/* CL_DEVICE_DOUBLE_FP_CONFIG is defined in CL.h for OpenCL >= 120 */
+
+#if CL_TARGET_OPENCL_VERSION <= 110
+#define CL_DEVICE_DOUBLE_FP_CONFIG                       0x1032
+#endif
+
+/* cl_khr_fp16 extension - no extension #define since it has no functions  */
+#define CL_DEVICE_HALF_FP_CONFIG                    0x1033
+
+/* Memory object destruction
+ *
+ * Apple extension for use to manage externally allocated buffers used with cl_mem objects with CL_MEM_USE_HOST_PTR
+ *
+ * Registers a user callback function that will be called when the memory object is deleted and its resources
+ * freed. Each call to clSetMemObjectCallbackFn registers the specified user callback function on a callback
+ * stack associated with memobj. The registered user callback functions are called in the reverse order in
+ * which they were registered. The user callback functions are called and then the memory object is deleted
+ * and its resources freed. This provides a mechanism for the application (and libraries) using memobj to be
+ * notified when the memory referenced by host_ptr, specified when the memory object is created and used as
+ * the storage bits for the memory object, can be reused or freed.
+ *
+ * The application may not call CL api's with the cl_mem object passed to the pfn_notify.
+ *
+ * Please check for the "cl_APPLE_SetMemObjectDestructor" extension using clGetDeviceInfo(CL_DEVICE_EXTENSIONS)
+ * before using.
+ */
+#define cl_APPLE_SetMemObjectDestructor 1
+cl_int  CL_API_ENTRY clSetMemObjectDestructorAPPLE(  cl_mem memobj,
+                                        void (* pfn_notify)(cl_mem memobj, void * user_data),
+                                        void * user_data)             CL_EXT_SUFFIX__VERSION_1_0;
+
+
+/* Context Logging Functions
+ *
+ * The next three convenience functions are intended to be used as the pfn_notify parameter to clCreateContext().
+ * Please check for the "cl_APPLE_ContextLoggingFunctions" extension using clGetDeviceInfo(CL_DEVICE_EXTENSIONS)
+ * before using.
+ *
+ * clLogMessagesToSystemLog forwards on all log messages to the Apple System Logger
+ */
+#define cl_APPLE_ContextLoggingFunctions 1
+extern void CL_API_ENTRY clLogMessagesToSystemLogAPPLE(  const char * errstr,
+                                            const void * private_info,
+                                            size_t       cb,
+                                            void *       user_data)  CL_EXT_SUFFIX__VERSION_1_0;
+
+/* clLogMessagesToStdout sends all log messages to the file descriptor stdout */
+extern void CL_API_ENTRY clLogMessagesToStdoutAPPLE(   const char * errstr,
+                                          const void * private_info,
+                                          size_t       cb,
+                                          void *       user_data)    CL_EXT_SUFFIX__VERSION_1_0;
+
+/* clLogMessagesToStderr sends all log messages to the file descriptor stderr */
+extern void CL_API_ENTRY clLogMessagesToStderrAPPLE(   const char * errstr,
+                                          const void * private_info,
+                                          size_t       cb,
+                                          void *       user_data)    CL_EXT_SUFFIX__VERSION_1_0;
+
+
+/************************
+* cl_khr_icd extension *
+************************/
+#define cl_khr_icd 1
+
+/* cl_platform_info                                                        */
+#define CL_PLATFORM_ICD_SUFFIX_KHR                  0x0920
+
+/* Additional Error Codes                                                  */
+#define CL_PLATFORM_NOT_FOUND_KHR                   -1001
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clIcdGetPlatformIDsKHR(cl_uint          num_entries,
+                       cl_platform_id * platforms,
+                       cl_uint *        num_platforms);
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL *clIcdGetPlatformIDsKHR_fn)(cl_uint          num_entries,
+                                         cl_platform_id * platforms,
+                                         cl_uint *        num_platforms);
+
+
+/*******************************
+ * cl_khr_il_program extension *
+ *******************************/
+#define cl_khr_il_program 1
+
+/* New property to clGetDeviceInfo for retrieving supported intermediate
+ * languages
+ */
+#define CL_DEVICE_IL_VERSION_KHR                    0x105B
+
+/* New property to clGetProgramInfo for retrieving for retrieving the IL of a
+ * program
+ */
+#define CL_PROGRAM_IL_KHR                           0x1169
+
+extern CL_API_ENTRY cl_program CL_API_CALL
+clCreateProgramWithILKHR(cl_context   context,
+                         const void * il,
+                         size_t       length,
+                         cl_int *     errcode_ret);
+
+typedef CL_API_ENTRY cl_program
+(CL_API_CALL *clCreateProgramWithILKHR_fn)(cl_context   context,
+                                           const void * il,
+                                           size_t       length,
+                                           cl_int *     errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+/* Extension: cl_khr_image2d_from_buffer
+ *
+ * This extension allows a 2D image to be created from a cl_mem buffer without
+ * a copy. The type associated with a 2D image created from a buffer in an
+ * OpenCL program is image2d_t. Both the sampler and sampler-less read_image
+ * built-in functions are supported for 2D images and 2D images created from
+ * a buffer.  Similarly, the write_image built-ins are also supported for 2D
+ * images created from a buffer.
+ *
+ * When the 2D image from buffer is created, the client must specify the
+ * width, height, image format (i.e. channel order and channel data type)
+ * and optionally the row pitch.
+ *
+ * The pitch specified must be a multiple of
+ * CL_DEVICE_IMAGE_PITCH_ALIGNMENT_KHR pixels.
+ * The base address of the buffer must be aligned to
+ * CL_DEVICE_IMAGE_BASE_ADDRESS_ALIGNMENT_KHR pixels.
+ */
+
+#define CL_DEVICE_IMAGE_PITCH_ALIGNMENT_KHR              0x104A
+#define CL_DEVICE_IMAGE_BASE_ADDRESS_ALIGNMENT_KHR       0x104B
+
+
+/**************************************
+ * cl_khr_initialize_memory extension *
+ **************************************/
+
+#define CL_CONTEXT_MEMORY_INITIALIZE_KHR            0x2030
+
+
+/**************************************
+ * cl_khr_terminate_context extension *
+ **************************************/
+
+#define CL_DEVICE_TERMINATE_CAPABILITY_KHR          0x2031
+#define CL_CONTEXT_TERMINATE_KHR                    0x2032
+
+#define cl_khr_terminate_context 1
+extern CL_API_ENTRY cl_int CL_API_CALL
+clTerminateContextKHR(cl_context context) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL *clTerminateContextKHR_fn)(cl_context context) CL_EXT_SUFFIX__VERSION_1_2;
+
+
+/*
+ * Extension: cl_khr_spir
+ *
+ * This extension adds support to create an OpenCL program object from a
+ * Standard Portable Intermediate Representation (SPIR) instance
+ */
+
+#define CL_DEVICE_SPIR_VERSIONS                     0x40E0
+#define CL_PROGRAM_BINARY_TYPE_INTERMEDIATE         0x40E1
+
+
+/*****************************************
+ * cl_khr_create_command_queue extension *
+ *****************************************/
+#define cl_khr_create_command_queue 1
+
+typedef cl_bitfield cl_queue_properties_khr;
+
+extern CL_API_ENTRY cl_command_queue CL_API_CALL
+clCreateCommandQueueWithPropertiesKHR(cl_context context,
+                                      cl_device_id device,
+                                      const cl_queue_properties_khr* properties,
+                                      cl_int* errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_command_queue
+(CL_API_CALL *clCreateCommandQueueWithPropertiesKHR_fn)(cl_context context,
+                                                        cl_device_id device,
+                                                        const cl_queue_properties_khr* properties,
+                                                        cl_int* errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+
+/******************************************
+* cl_nv_device_attribute_query extension *
+******************************************/
+
+/* cl_nv_device_attribute_query extension - no extension #define since it has no functions */
+#define CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV       0x4000
+#define CL_DEVICE_COMPUTE_CAPABILITY_MINOR_NV       0x4001
+#define CL_DEVICE_REGISTERS_PER_BLOCK_NV            0x4002
+#define CL_DEVICE_WARP_SIZE_NV                      0x4003
+#define CL_DEVICE_GPU_OVERLAP_NV                    0x4004
+#define CL_DEVICE_KERNEL_EXEC_TIMEOUT_NV            0x4005
+#define CL_DEVICE_INTEGRATED_MEMORY_NV              0x4006
+
+
+/*********************************
+* cl_amd_device_attribute_query *
+*********************************/
+
+#define CL_DEVICE_PROFILING_TIMER_OFFSET_AMD        0x4036
+
+
+/*********************************
+* cl_arm_printf extension
+*********************************/
+
+#define CL_PRINTF_CALLBACK_ARM                      0x40B0
+#define CL_PRINTF_BUFFERSIZE_ARM                    0x40B1
+
+
+/***********************************
+* cl_ext_device_fission extension
+***********************************/
+#define cl_ext_device_fission   1
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clReleaseDeviceEXT(cl_device_id device) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL *clReleaseDeviceEXT_fn)(cl_device_id device) CL_EXT_SUFFIX__VERSION_1_1;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clRetainDeviceEXT(cl_device_id device) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL *clRetainDeviceEXT_fn)(cl_device_id device) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef cl_ulong  cl_device_partition_property_ext;
+extern CL_API_ENTRY cl_int CL_API_CALL
+clCreateSubDevicesEXT(cl_device_id   in_device,
+                      const cl_device_partition_property_ext * properties,
+                      cl_uint        num_entries,
+                      cl_device_id * out_devices,
+                      cl_uint *      num_devices) CL_EXT_SUFFIX__VERSION_1_1;
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL * clCreateSubDevicesEXT_fn)(cl_device_id   in_device,
+                                         const cl_device_partition_property_ext * properties,
+                                         cl_uint        num_entries,
+                                         cl_device_id * out_devices,
+                                         cl_uint *      num_devices) CL_EXT_SUFFIX__VERSION_1_1;
+
+/* cl_device_partition_property_ext */
+#define CL_DEVICE_PARTITION_EQUALLY_EXT             0x4050
+#define CL_DEVICE_PARTITION_BY_COUNTS_EXT           0x4051
+#define CL_DEVICE_PARTITION_BY_NAMES_EXT            0x4052
+#define CL_DEVICE_PARTITION_BY_AFFINITY_DOMAIN_EXT  0x4053
+
+/* clDeviceGetInfo selectors */
+#define CL_DEVICE_PARENT_DEVICE_EXT                 0x4054
+#define CL_DEVICE_PARTITION_TYPES_EXT               0x4055
+#define CL_DEVICE_AFFINITY_DOMAINS_EXT              0x4056
+#define CL_DEVICE_REFERENCE_COUNT_EXT               0x4057
+#define CL_DEVICE_PARTITION_STYLE_EXT               0x4058
+
+/* error codes */
+#define CL_DEVICE_PARTITION_FAILED_EXT              -1057
+#define CL_INVALID_PARTITION_COUNT_EXT              -1058
+#define CL_INVALID_PARTITION_NAME_EXT               -1059
+
+/* CL_AFFINITY_DOMAINs */
+#define CL_AFFINITY_DOMAIN_L1_CACHE_EXT             0x1
+#define CL_AFFINITY_DOMAIN_L2_CACHE_EXT             0x2
+#define CL_AFFINITY_DOMAIN_L3_CACHE_EXT             0x3
+#define CL_AFFINITY_DOMAIN_L4_CACHE_EXT             0x4
+#define CL_AFFINITY_DOMAIN_NUMA_EXT                 0x10
+#define CL_AFFINITY_DOMAIN_NEXT_FISSIONABLE_EXT     0x100
+
+/* cl_device_partition_property_ext list terminators */
+#define CL_PROPERTIES_LIST_END_EXT                  ((cl_device_partition_property_ext) 0)
+#define CL_PARTITION_BY_COUNTS_LIST_END_EXT         ((cl_device_partition_property_ext) 0)
+#define CL_PARTITION_BY_NAMES_LIST_END_EXT          ((cl_device_partition_property_ext) 0 - 1)
+
+
+/***********************************
+ * cl_ext_migrate_memobject extension definitions
+ ***********************************/
+#define cl_ext_migrate_memobject 1
+
+typedef cl_bitfield cl_mem_migration_flags_ext;
+
+#define CL_MIGRATE_MEM_OBJECT_HOST_EXT              0x1
+
+#define CL_COMMAND_MIGRATE_MEM_OBJECT_EXT           0x4040
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueMigrateMemObjectEXT(cl_command_queue command_queue,
+                             cl_uint          num_mem_objects,
+                             const cl_mem *   mem_objects,
+                             cl_mem_migration_flags_ext flags,
+                             cl_uint          num_events_in_wait_list,
+                             const cl_event * event_wait_list,
+                             cl_event *       event);
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL *clEnqueueMigrateMemObjectEXT_fn)(cl_command_queue command_queue,
+                                               cl_uint          num_mem_objects,
+                                               const cl_mem *   mem_objects,
+                                               cl_mem_migration_flags_ext flags,
+                                               cl_uint          num_events_in_wait_list,
+                                               const cl_event * event_wait_list,
+                                               cl_event *       event);
+
+
+/*********************************
+* cl_qcom_ext_host_ptr extension
+*********************************/
+#define cl_qcom_ext_host_ptr 1
+
+#define CL_MEM_EXT_HOST_PTR_QCOM                  (1 << 29)
+
+#define CL_DEVICE_EXT_MEM_PADDING_IN_BYTES_QCOM   0x40A0
+#define CL_DEVICE_PAGE_SIZE_QCOM                  0x40A1
+#define CL_IMAGE_ROW_ALIGNMENT_QCOM               0x40A2
+#define CL_IMAGE_SLICE_ALIGNMENT_QCOM             0x40A3
+#define CL_MEM_HOST_UNCACHED_QCOM                 0x40A4
+#define CL_MEM_HOST_WRITEBACK_QCOM                0x40A5
+#define CL_MEM_HOST_WRITETHROUGH_QCOM             0x40A6
+#define CL_MEM_HOST_WRITE_COMBINING_QCOM          0x40A7
+
+typedef cl_uint                                   cl_image_pitch_info_qcom;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetDeviceImageInfoQCOM(cl_device_id             device,
+                         size_t                   image_width,
+                         size_t                   image_height,
+                         const cl_image_format   *image_format,
+                         cl_image_pitch_info_qcom param_name,
+                         size_t                   param_value_size,
+                         void                    *param_value,
+                         size_t                  *param_value_size_ret);
+
+typedef struct _cl_mem_ext_host_ptr
+{
+    /* Type of external memory allocation. */
+    /* Legal values will be defined in layered extensions. */
+    cl_uint  allocation_type;
+
+    /* Host cache policy for this external memory allocation. */
+    cl_uint  host_cache_policy;
+
+} cl_mem_ext_host_ptr;
+
+
+/*******************************************
+* cl_qcom_ext_host_ptr_iocoherent extension
+********************************************/
+
+/* Cache policy specifying io-coherence */
+#define CL_MEM_HOST_IOCOHERENT_QCOM               0x40A9
+
+
+/*********************************
+* cl_qcom_ion_host_ptr extension
+*********************************/
+
+#define CL_MEM_ION_HOST_PTR_QCOM                  0x40A8
+
+typedef struct _cl_mem_ion_host_ptr
+{
+    /* Type of external memory allocation. */
+    /* Must be CL_MEM_ION_HOST_PTR_QCOM for ION allocations. */
+    cl_mem_ext_host_ptr  ext_host_ptr;
+
+    /* ION file descriptor */
+    int                  ion_filedesc;
+
+    /* Host pointer to the ION allocated memory */
+    void*                ion_hostptr;
+
+} cl_mem_ion_host_ptr;
+
+
+/*********************************
+* cl_qcom_android_native_buffer_host_ptr extension
+*********************************/
+
+#define CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM                  0x40C6
+
+typedef struct _cl_mem_android_native_buffer_host_ptr
+{
+    /* Type of external memory allocation. */
+    /* Must be CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM for Android native buffers. */
+    cl_mem_ext_host_ptr  ext_host_ptr;
+
+    /* Virtual pointer to the android native buffer */
+    void*                anb_ptr;
+
+} cl_mem_android_native_buffer_host_ptr;
+
+
+/******************************************
+ * cl_img_yuv_image extension *
+ ******************************************/
+
+/* Image formats used in clCreateImage */
+#define CL_NV21_IMG                                 0x40D0
+#define CL_YV12_IMG                                 0x40D1
+
+
+/******************************************
+ * cl_img_cached_allocations extension *
+ ******************************************/
+
+/* Flag values used by clCreateBuffer */
+#define CL_MEM_USE_UNCACHED_CPU_MEMORY_IMG          (1 << 26)
+#define CL_MEM_USE_CACHED_CPU_MEMORY_IMG            (1 << 27)
+
+
+/******************************************
+ * cl_img_use_gralloc_ptr extension *
+ ******************************************/
+#define cl_img_use_gralloc_ptr 1
+
+/* Flag values used by clCreateBuffer */
+#define CL_MEM_USE_GRALLOC_PTR_IMG                  (1 << 28)
+
+/* To be used by clGetEventInfo: */
+#define CL_COMMAND_ACQUIRE_GRALLOC_OBJECTS_IMG      0x40D2
+#define CL_COMMAND_RELEASE_GRALLOC_OBJECTS_IMG      0x40D3
+
+/* Error code from clEnqueueReleaseGrallocObjectsIMG */
+#define CL_GRALLOC_RESOURCE_NOT_ACQUIRED_IMG        0x40D4
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueAcquireGrallocObjectsIMG(cl_command_queue      command_queue,
+                                  cl_uint               num_objects,
+                                  const cl_mem *        mem_objects,
+                                  cl_uint               num_events_in_wait_list,
+                                  const cl_event *      event_wait_list,
+                                  cl_event *            event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueReleaseGrallocObjectsIMG(cl_command_queue      command_queue,
+                                  cl_uint               num_objects,
+                                  const cl_mem *        mem_objects,
+                                  cl_uint               num_events_in_wait_list,
+                                  const cl_event *      event_wait_list,
+                                  cl_event *            event) CL_EXT_SUFFIX__VERSION_1_2;
+
+
+/*********************************
+* cl_khr_subgroups extension
+*********************************/
+#define cl_khr_subgroups 1
+
+#if !defined(CL_VERSION_2_1)
+/* For OpenCL 2.1 and newer, cl_kernel_sub_group_info is declared in CL.h.
+   In hindsight, there should have been a khr suffix on this type for
+   the extension, but keeping it un-suffixed to maintain backwards
+   compatibility. */
+typedef cl_uint             cl_kernel_sub_group_info;
+#endif
+
+/* cl_kernel_sub_group_info */
+#define CL_KERNEL_MAX_SUB_GROUP_SIZE_FOR_NDRANGE_KHR    0x2033
+#define CL_KERNEL_SUB_GROUP_COUNT_FOR_NDRANGE_KHR       0x2034
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetKernelSubGroupInfoKHR(cl_kernel    in_kernel,
+                           cl_device_id in_device,
+                           cl_kernel_sub_group_info param_name,
+                           size_t       input_value_size,
+                           const void * input_value,
+                           size_t       param_value_size,
+                           void *       param_value,
+                           size_t *     param_value_size_ret) CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED;
+
+typedef CL_API_ENTRY cl_int
+(CL_API_CALL * clGetKernelSubGroupInfoKHR_fn)(cl_kernel    in_kernel,
+                                              cl_device_id in_device,
+                                              cl_kernel_sub_group_info param_name,
+                                              size_t       input_value_size,
+                                              const void * input_value,
+                                              size_t       param_value_size,
+                                              void *       param_value,
+                                              size_t *     param_value_size_ret) CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED;
+
+
+/*********************************
+* cl_khr_mipmap_image extension
+*********************************/
+
+/* cl_sampler_properties */
+#define CL_SAMPLER_MIP_FILTER_MODE_KHR              0x1155
+#define CL_SAMPLER_LOD_MIN_KHR                      0x1156
+#define CL_SAMPLER_LOD_MAX_KHR                      0x1157
+
+
+/*********************************
+* cl_khr_priority_hints extension
+*********************************/
+/* This extension define is for backwards compatibility.
+   It shouldn't be required since this extension has no new functions. */
+#define cl_khr_priority_hints 1
+
+typedef cl_uint  cl_queue_priority_khr;
+
+/* cl_command_queue_properties */
+#define CL_QUEUE_PRIORITY_KHR 0x1096
+
+/* cl_queue_priority_khr */
+#define CL_QUEUE_PRIORITY_HIGH_KHR (1<<0)
+#define CL_QUEUE_PRIORITY_MED_KHR (1<<1)
+#define CL_QUEUE_PRIORITY_LOW_KHR (1<<2)
+
+
+/*********************************
+* cl_khr_throttle_hints extension
+*********************************/
+/* This extension define is for backwards compatibility.
+   It shouldn't be required since this extension has no new functions. */
+#define cl_khr_throttle_hints 1
+
+typedef cl_uint  cl_queue_throttle_khr;
+
+/* cl_command_queue_properties */
+#define CL_QUEUE_THROTTLE_KHR 0x1097
+
+/* cl_queue_throttle_khr */
+#define CL_QUEUE_THROTTLE_HIGH_KHR (1<<0)
+#define CL_QUEUE_THROTTLE_MED_KHR (1<<1)
+#define CL_QUEUE_THROTTLE_LOW_KHR (1<<2)
+
+
+/*********************************
+* cl_khr_subgroup_named_barrier
+*********************************/
+/* This extension define is for backwards compatibility.
+   It shouldn't be required since this extension has no new functions. */
+#define cl_khr_subgroup_named_barrier 1
+
+/* cl_device_info */
+#define CL_DEVICE_MAX_NAMED_BARRIER_COUNT_KHR       0x2035
+
+
+/**********************************
+ * cl_arm_import_memory extension *
+ **********************************/
+#define cl_arm_import_memory 1
+
+typedef intptr_t cl_import_properties_arm;
+
+/* Default and valid proporties name for cl_arm_import_memory */
+#define CL_IMPORT_TYPE_ARM                        0x40B2
+
+/* Host process memory type default value for CL_IMPORT_TYPE_ARM property */
+#define CL_IMPORT_TYPE_HOST_ARM                   0x40B3
+
+/* DMA BUF memory type value for CL_IMPORT_TYPE_ARM property */
+#define CL_IMPORT_TYPE_DMA_BUF_ARM                0x40B4
+
+/* Protected DMA BUF memory type value for CL_IMPORT_TYPE_ARM property */
+#define CL_IMPORT_TYPE_PROTECTED_ARM              0x40B5
+
+/* This extension adds a new function that allows for direct memory import into
+ * OpenCL via the clImportMemoryARM function.
+ *
+ * Memory imported through this interface will be mapped into the device's page
+ * tables directly, providing zero copy access. It will never fall back to copy
+ * operations and aliased buffers.
+ *
+ * Types of memory supported for import are specified as additional extension
+ * strings.
+ *
+ * This extension produces cl_mem allocations which are compatible with all other
+ * users of cl_mem in the standard API.
+ *
+ * This extension maps pages with the same properties as the normal buffer creation
+ * function clCreateBuffer.
+ */
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clImportMemoryARM( cl_context context,
+                   cl_mem_flags flags,
+                   const cl_import_properties_arm *properties,
+                   void *memory,
+                   size_t size,
+                   cl_int *errcode_ret) CL_EXT_SUFFIX__VERSION_1_0;
+
+
+/******************************************
+ * cl_arm_shared_virtual_memory extension *
+ ******************************************/
+#define cl_arm_shared_virtual_memory 1
+
+/* Used by clGetDeviceInfo */
+#define CL_DEVICE_SVM_CAPABILITIES_ARM                  0x40B6
+
+/* Used by clGetMemObjectInfo */
+#define CL_MEM_USES_SVM_POINTER_ARM                     0x40B7
+
+/* Used by clSetKernelExecInfoARM: */
+#define CL_KERNEL_EXEC_INFO_SVM_PTRS_ARM                0x40B8
+#define CL_KERNEL_EXEC_INFO_SVM_FINE_GRAIN_SYSTEM_ARM   0x40B9
+
+/* To be used by clGetEventInfo: */
+#define CL_COMMAND_SVM_FREE_ARM                         0x40BA
+#define CL_COMMAND_SVM_MEMCPY_ARM                       0x40BB
+#define CL_COMMAND_SVM_MEMFILL_ARM                      0x40BC
+#define CL_COMMAND_SVM_MAP_ARM                          0x40BD
+#define CL_COMMAND_SVM_UNMAP_ARM                        0x40BE
+
+/* Flag values returned by clGetDeviceInfo with CL_DEVICE_SVM_CAPABILITIES_ARM as the param_name. */
+#define CL_DEVICE_SVM_COARSE_GRAIN_BUFFER_ARM           (1 << 0)
+#define CL_DEVICE_SVM_FINE_GRAIN_BUFFER_ARM             (1 << 1)
+#define CL_DEVICE_SVM_FINE_GRAIN_SYSTEM_ARM             (1 << 2)
+#define CL_DEVICE_SVM_ATOMICS_ARM                       (1 << 3)
+
+/* Flag values used by clSVMAllocARM: */
+#define CL_MEM_SVM_FINE_GRAIN_BUFFER_ARM                (1 << 10)
+#define CL_MEM_SVM_ATOMICS_ARM                          (1 << 11)
+
+typedef cl_bitfield cl_svm_mem_flags_arm;
+typedef cl_uint     cl_kernel_exec_info_arm;
+typedef cl_bitfield cl_device_svm_capabilities_arm;
+
+extern CL_API_ENTRY void * CL_API_CALL
+clSVMAllocARM(cl_context       context,
+              cl_svm_mem_flags_arm flags,
+              size_t           size,
+              cl_uint          alignment) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY void CL_API_CALL
+clSVMFreeARM(cl_context        context,
+             void *            svm_pointer) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueSVMFreeARM(cl_command_queue  command_queue,
+                    cl_uint           num_svm_pointers,
+                    void *            svm_pointers[],
+                    void (CL_CALLBACK * pfn_free_func)(cl_command_queue queue,
+                                                       cl_uint          num_svm_pointers,
+                                                       void *           svm_pointers[],
+                                                       void *           user_data),
+                    void *            user_data,
+                    cl_uint           num_events_in_wait_list,
+                    const cl_event *  event_wait_list,
+                    cl_event *        event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueSVMMemcpyARM(cl_command_queue  command_queue,
+                      cl_bool           blocking_copy,
+                      void *            dst_ptr,
+                      const void *      src_ptr,
+                      size_t            size,
+                      cl_uint           num_events_in_wait_list,
+                      const cl_event *  event_wait_list,
+                      cl_event *        event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueSVMMemFillARM(cl_command_queue  command_queue,
+                       void *            svm_ptr,
+                       const void *      pattern,
+                       size_t            pattern_size,
+                       size_t            size,
+                       cl_uint           num_events_in_wait_list,
+                       const cl_event *  event_wait_list,
+                       cl_event *        event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueSVMMapARM(cl_command_queue  command_queue,
+                   cl_bool           blocking_map,
+                   cl_map_flags      flags,
+                   void *            svm_ptr,
+                   size_t            size,
+                   cl_uint           num_events_in_wait_list,
+                   const cl_event *  event_wait_list,
+                   cl_event *        event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueSVMUnmapARM(cl_command_queue  command_queue,
+                     void *            svm_ptr,
+                     cl_uint           num_events_in_wait_list,
+                     const cl_event *  event_wait_list,
+                     cl_event *        event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clSetKernelArgSVMPointerARM(cl_kernel    kernel,
+                            cl_uint      arg_index,
+                            const void * arg_value) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clSetKernelExecInfoARM(cl_kernel            kernel,
+                       cl_kernel_exec_info_arm  param_name,
+                       size_t               param_value_size,
+                       const void *         param_value) CL_EXT_SUFFIX__VERSION_1_2;
+
+/********************************
+ * cl_arm_get_core_id extension *
+ ********************************/
+
+#ifdef CL_VERSION_1_2
+
+#define cl_arm_get_core_id 1
+
+/* Device info property for bitfield of cores present */
+#define CL_DEVICE_COMPUTE_UNITS_BITFIELD_ARM      0x40BF
+
+#endif  /* CL_VERSION_1_2 */
+
+/*********************************
+* cl_arm_job_slot_selection
+*********************************/
+
+#define cl_arm_job_slot_selection 1
+
+/* cl_device_info */
+#define CL_DEVICE_JOB_SLOTS_ARM                   0x41E0
+
+/* cl_command_queue_properties */
+#define CL_QUEUE_JOB_SLOT_ARM                     0x41E1
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif /* __CL_EXT_H */

benchmarks/new_opencl/include/CL/cl_ext_intel.h

@@ -0,0 +1,423 @@
+/*******************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ ******************************************************************************/
+/*****************************************************************************\
+
+Copyright (c) 2013-2019 Intel Corporation All Rights Reserved.
+
+THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
+CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
+NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
+MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+File Name: cl_ext_intel.h
+
+Abstract:
+
+Notes:
+
+\*****************************************************************************/
+
+#ifndef __CL_EXT_INTEL_H
+#define __CL_EXT_INTEL_H
+
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/***************************************
+* cl_intel_thread_local_exec extension *
+****************************************/
+
+#define cl_intel_thread_local_exec 1
+
+#define CL_QUEUE_THREAD_LOCAL_EXEC_ENABLE_INTEL      (((cl_bitfield)1) << 31)
+
+/***********************************************
+* cl_intel_device_partition_by_names extension *
+************************************************/
+
+#define cl_intel_device_partition_by_names 1
+
+#define CL_DEVICE_PARTITION_BY_NAMES_INTEL          0x4052
+#define CL_PARTITION_BY_NAMES_LIST_END_INTEL        -1
+
+/************************************************
+* cl_intel_accelerator extension                *
+* cl_intel_motion_estimation extension          *
+* cl_intel_advanced_motion_estimation extension *
+*************************************************/
+
+#define cl_intel_accelerator 1
+#define cl_intel_motion_estimation 1
+#define cl_intel_advanced_motion_estimation 1
+
+typedef struct _cl_accelerator_intel* cl_accelerator_intel;
+typedef cl_uint cl_accelerator_type_intel;
+typedef cl_uint cl_accelerator_info_intel;
+
+typedef struct _cl_motion_estimation_desc_intel {
+    cl_uint mb_block_type;
+    cl_uint subpixel_mode;
+    cl_uint sad_adjust_mode;
+    cl_uint search_path_type;
+} cl_motion_estimation_desc_intel;
+
+/* error codes */
+#define CL_INVALID_ACCELERATOR_INTEL                              -1094
+#define CL_INVALID_ACCELERATOR_TYPE_INTEL                         -1095
+#define CL_INVALID_ACCELERATOR_DESCRIPTOR_INTEL                   -1096
+#define CL_ACCELERATOR_TYPE_NOT_SUPPORTED_INTEL                   -1097
+
+/* cl_accelerator_type_intel */
+#define CL_ACCELERATOR_TYPE_MOTION_ESTIMATION_INTEL               0x0
+
+/* cl_accelerator_info_intel */
+#define CL_ACCELERATOR_DESCRIPTOR_INTEL                           0x4090
+#define CL_ACCELERATOR_REFERENCE_COUNT_INTEL                      0x4091
+#define CL_ACCELERATOR_CONTEXT_INTEL                              0x4092
+#define CL_ACCELERATOR_TYPE_INTEL                                 0x4093
+
+/* cl_motion_detect_desc_intel flags */
+#define CL_ME_MB_TYPE_16x16_INTEL                                 0x0
+#define CL_ME_MB_TYPE_8x8_INTEL                                   0x1
+#define CL_ME_MB_TYPE_4x4_INTEL                                   0x2
+
+#define CL_ME_SUBPIXEL_MODE_INTEGER_INTEL                         0x0
+#define CL_ME_SUBPIXEL_MODE_HPEL_INTEL                            0x1
+#define CL_ME_SUBPIXEL_MODE_QPEL_INTEL                            0x2
+
+#define CL_ME_SAD_ADJUST_MODE_NONE_INTEL                          0x0
+#define CL_ME_SAD_ADJUST_MODE_HAAR_INTEL                          0x1
+
+#define CL_ME_SEARCH_PATH_RADIUS_2_2_INTEL                        0x0
+#define CL_ME_SEARCH_PATH_RADIUS_4_4_INTEL                        0x1
+#define CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL                      0x5
+
+#define CL_ME_SKIP_BLOCK_TYPE_16x16_INTEL                         0x0
+#define CL_ME_CHROMA_INTRA_PREDICT_ENABLED_INTEL                  0x1
+#define CL_ME_LUMA_INTRA_PREDICT_ENABLED_INTEL                    0x2
+#define CL_ME_SKIP_BLOCK_TYPE_8x8_INTEL                           0x4
+
+#define CL_ME_FORWARD_INPUT_MODE_INTEL                            0x1
+#define CL_ME_BACKWARD_INPUT_MODE_INTEL                           0x2
+#define CL_ME_BIDIRECTION_INPUT_MODE_INTEL                        0x3
+
+#define CL_ME_BIDIR_WEIGHT_QUARTER_INTEL                          16
+#define CL_ME_BIDIR_WEIGHT_THIRD_INTEL                            21
+#define CL_ME_BIDIR_WEIGHT_HALF_INTEL                             32
+#define CL_ME_BIDIR_WEIGHT_TWO_THIRD_INTEL                        43
+#define CL_ME_BIDIR_WEIGHT_THREE_QUARTER_INTEL                    48
+
+#define CL_ME_COST_PENALTY_NONE_INTEL                             0x0
+#define CL_ME_COST_PENALTY_LOW_INTEL                              0x1
+#define CL_ME_COST_PENALTY_NORMAL_INTEL                           0x2
+#define CL_ME_COST_PENALTY_HIGH_INTEL                             0x3
+
+#define CL_ME_COST_PRECISION_QPEL_INTEL                           0x0
+#define CL_ME_COST_PRECISION_HPEL_INTEL                           0x1
+#define CL_ME_COST_PRECISION_PEL_INTEL                            0x2
+#define CL_ME_COST_PRECISION_DPEL_INTEL                           0x3
+
+#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_INTEL                  0x0
+#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_INTEL                0x1
+#define CL_ME_LUMA_PREDICTOR_MODE_DC_INTEL                        0x2
+#define CL_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_LEFT_INTEL        0x3
+
+#define CL_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_RIGHT_INTEL       0x4
+#define CL_ME_LUMA_PREDICTOR_MODE_PLANE_INTEL                     0x4
+#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_RIGHT_INTEL            0x5
+#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_DOWN_INTEL           0x6
+#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_LEFT_INTEL             0x7
+#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_UP_INTEL             0x8
+
+#define CL_ME_CHROMA_PREDICTOR_MODE_DC_INTEL                      0x0
+#define CL_ME_CHROMA_PREDICTOR_MODE_HORIZONTAL_INTEL              0x1
+#define CL_ME_CHROMA_PREDICTOR_MODE_VERTICAL_INTEL                0x2
+#define CL_ME_CHROMA_PREDICTOR_MODE_PLANE_INTEL                   0x3
+
+/* cl_device_info */
+#define CL_DEVICE_ME_VERSION_INTEL                                0x407E
+
+#define CL_ME_VERSION_LEGACY_INTEL                                0x0
+#define CL_ME_VERSION_ADVANCED_VER_1_INTEL                        0x1
+#define CL_ME_VERSION_ADVANCED_VER_2_INTEL                        0x2
+
+extern CL_API_ENTRY cl_accelerator_intel CL_API_CALL
+clCreateAcceleratorINTEL(
+    cl_context                   context,
+    cl_accelerator_type_intel    accelerator_type,
+    size_t                       descriptor_size,
+    const void*                  descriptor,
+    cl_int*                      errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_accelerator_intel (CL_API_CALL *clCreateAcceleratorINTEL_fn)(
+    cl_context                   context,
+    cl_accelerator_type_intel    accelerator_type,
+    size_t                       descriptor_size,
+    const void*                  descriptor,
+    cl_int*                      errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetAcceleratorInfoINTEL(
+    cl_accelerator_intel         accelerator,
+    cl_accelerator_info_intel    param_name,
+    size_t                       param_value_size,
+    void*                        param_value,
+    size_t*                      param_value_size_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetAcceleratorInfoINTEL_fn)(
+    cl_accelerator_intel         accelerator,
+    cl_accelerator_info_intel    param_name,
+    size_t                       param_value_size,
+    void*                        param_value,
+    size_t*                      param_value_size_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clRetainAcceleratorINTEL(
+    cl_accelerator_intel         accelerator) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clRetainAcceleratorINTEL_fn)(
+    cl_accelerator_intel         accelerator) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clReleaseAcceleratorINTEL(
+    cl_accelerator_intel         accelerator) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clReleaseAcceleratorINTEL_fn)(
+    cl_accelerator_intel         accelerator) CL_EXT_SUFFIX__VERSION_1_2;
+
+/******************************************
+* cl_intel_simultaneous_sharing extension *
+*******************************************/
+
+#define cl_intel_simultaneous_sharing 1
+
+#define CL_DEVICE_SIMULTANEOUS_INTEROPS_INTEL            0x4104
+#define CL_DEVICE_NUM_SIMULTANEOUS_INTEROPS_INTEL        0x4105
+
+/***********************************
+* cl_intel_egl_image_yuv extension *
+************************************/
+
+#define cl_intel_egl_image_yuv 1
+
+#define CL_EGL_YUV_PLANE_INTEL                           0x4107
+
+/********************************
+* cl_intel_packed_yuv extension *
+*********************************/
+
+#define cl_intel_packed_yuv 1
+
+#define CL_YUYV_INTEL                                    0x4076
+#define CL_UYVY_INTEL                                    0x4077
+#define CL_YVYU_INTEL                                    0x4078
+#define CL_VYUY_INTEL                                    0x4079
+
+/********************************************
+* cl_intel_required_subgroup_size extension *
+*********************************************/
+
+#define cl_intel_required_subgroup_size 1
+
+#define CL_DEVICE_SUB_GROUP_SIZES_INTEL                  0x4108
+#define CL_KERNEL_SPILL_MEM_SIZE_INTEL                   0x4109
+#define CL_KERNEL_COMPILE_SUB_GROUP_SIZE_INTEL           0x410A
+
+/****************************************
+* cl_intel_driver_diagnostics extension *
+*****************************************/
+
+#define cl_intel_driver_diagnostics 1
+
+typedef cl_uint cl_diagnostics_verbose_level;
+
+#define CL_CONTEXT_SHOW_DIAGNOSTICS_INTEL                0x4106
+
+#define CL_CONTEXT_DIAGNOSTICS_LEVEL_ALL_INTEL           ( 0xff )
+#define CL_CONTEXT_DIAGNOSTICS_LEVEL_GOOD_INTEL          ( 1 )
+#define CL_CONTEXT_DIAGNOSTICS_LEVEL_BAD_INTEL           ( 1 << 1 )
+#define CL_CONTEXT_DIAGNOSTICS_LEVEL_NEUTRAL_INTEL       ( 1 << 2 )
+
+/********************************
+* cl_intel_planar_yuv extension *
+*********************************/
+
+#define CL_NV12_INTEL                                       0x410E
+
+#define CL_MEM_NO_ACCESS_INTEL                              ( 1 << 24 )
+#define CL_MEM_ACCESS_FLAGS_UNRESTRICTED_INTEL              ( 1 << 25 )
+
+#define CL_DEVICE_PLANAR_YUV_MAX_WIDTH_INTEL                0x417E
+#define CL_DEVICE_PLANAR_YUV_MAX_HEIGHT_INTEL               0x417F
+
+/*******************************************************
+* cl_intel_device_side_avc_motion_estimation extension *
+********************************************************/
+
+#define CL_DEVICE_AVC_ME_VERSION_INTEL                      0x410B
+#define CL_DEVICE_AVC_ME_SUPPORTS_TEXTURE_SAMPLER_USE_INTEL 0x410C
+#define CL_DEVICE_AVC_ME_SUPPORTS_PREEMPTION_INTEL          0x410D
+
+#define CL_AVC_ME_VERSION_0_INTEL                           0x0;  // No support.
+#define CL_AVC_ME_VERSION_1_INTEL                           0x1;  // First supported version.
+
+#define CL_AVC_ME_MAJOR_16x16_INTEL                         0x0
+#define CL_AVC_ME_MAJOR_16x8_INTEL                          0x1
+#define CL_AVC_ME_MAJOR_8x16_INTEL                          0x2
+#define CL_AVC_ME_MAJOR_8x8_INTEL                           0x3
+
+#define CL_AVC_ME_MINOR_8x8_INTEL                           0x0
+#define CL_AVC_ME_MINOR_8x4_INTEL                           0x1
+#define CL_AVC_ME_MINOR_4x8_INTEL                           0x2
+#define CL_AVC_ME_MINOR_4x4_INTEL                           0x3
+
+#define CL_AVC_ME_MAJOR_FORWARD_INTEL                       0x0
+#define CL_AVC_ME_MAJOR_BACKWARD_INTEL                      0x1
+#define CL_AVC_ME_MAJOR_BIDIRECTIONAL_INTEL                 0x2
+
+#define CL_AVC_ME_PARTITION_MASK_ALL_INTEL                  0x0
+#define CL_AVC_ME_PARTITION_MASK_16x16_INTEL                0x7E
+#define CL_AVC_ME_PARTITION_MASK_16x8_INTEL                 0x7D
+#define CL_AVC_ME_PARTITION_MASK_8x16_INTEL                 0x7B
+#define CL_AVC_ME_PARTITION_MASK_8x8_INTEL                  0x77
+#define CL_AVC_ME_PARTITION_MASK_8x4_INTEL                  0x6F
+#define CL_AVC_ME_PARTITION_MASK_4x8_INTEL                  0x5F
+#define CL_AVC_ME_PARTITION_MASK_4x4_INTEL                  0x3F
+
+#define CL_AVC_ME_SEARCH_WINDOW_EXHAUSTIVE_INTEL            0x0
+#define CL_AVC_ME_SEARCH_WINDOW_SMALL_INTEL                 0x1
+#define CL_AVC_ME_SEARCH_WINDOW_TINY_INTEL                  0x2
+#define CL_AVC_ME_SEARCH_WINDOW_EXTRA_TINY_INTEL            0x3
+#define CL_AVC_ME_SEARCH_WINDOW_DIAMOND_INTEL               0x4
+#define CL_AVC_ME_SEARCH_WINDOW_LARGE_DIAMOND_INTEL         0x5
+#define CL_AVC_ME_SEARCH_WINDOW_RESERVED0_INTEL             0x6
+#define CL_AVC_ME_SEARCH_WINDOW_RESERVED1_INTEL             0x7
+#define CL_AVC_ME_SEARCH_WINDOW_CUSTOM_INTEL                0x8
+#define CL_AVC_ME_SEARCH_WINDOW_16x12_RADIUS_INTEL          0x9
+#define CL_AVC_ME_SEARCH_WINDOW_4x4_RADIUS_INTEL            0x2
+#define CL_AVC_ME_SEARCH_WINDOW_2x2_RADIUS_INTEL            0xa
+
+#define CL_AVC_ME_SAD_ADJUST_MODE_NONE_INTEL                0x0
+#define CL_AVC_ME_SAD_ADJUST_MODE_HAAR_INTEL                0x2
+
+#define CL_AVC_ME_SUBPIXEL_MODE_INTEGER_INTEL               0x0
+#define CL_AVC_ME_SUBPIXEL_MODE_HPEL_INTEL                  0x1
+#define CL_AVC_ME_SUBPIXEL_MODE_QPEL_INTEL                  0x3
+
+#define CL_AVC_ME_COST_PRECISION_QPEL_INTEL                 0x0
+#define CL_AVC_ME_COST_PRECISION_HPEL_INTEL                 0x1
+#define CL_AVC_ME_COST_PRECISION_PEL_INTEL                  0x2
+#define CL_AVC_ME_COST_PRECISION_DPEL_INTEL                 0x3
+
+#define CL_AVC_ME_BIDIR_WEIGHT_QUARTER_INTEL                0x10
+#define CL_AVC_ME_BIDIR_WEIGHT_THIRD_INTEL                  0x15
+#define CL_AVC_ME_BIDIR_WEIGHT_HALF_INTEL                   0x20
+#define CL_AVC_ME_BIDIR_WEIGHT_TWO_THIRD_INTEL              0x2B
+#define CL_AVC_ME_BIDIR_WEIGHT_THREE_QUARTER_INTEL          0x30
+
+#define CL_AVC_ME_BORDER_REACHED_LEFT_INTEL                 0x0
+#define CL_AVC_ME_BORDER_REACHED_RIGHT_INTEL                0x2
+#define CL_AVC_ME_BORDER_REACHED_TOP_INTEL                  0x4
+#define CL_AVC_ME_BORDER_REACHED_BOTTOM_INTEL               0x8
+
+#define CL_AVC_ME_SKIP_BLOCK_PARTITION_16x16_INTEL          0x0
+#define CL_AVC_ME_SKIP_BLOCK_PARTITION_8x8_INTEL            0x4000
+
+#define CL_AVC_ME_SKIP_BLOCK_16x16_FORWARD_ENABLE_INTEL     ( 0x1 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_16x16_BACKWARD_ENABLE_INTEL    ( 0x2 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_16x16_DUAL_ENABLE_INTEL        ( 0x3 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_FORWARD_ENABLE_INTEL       ( 0x55 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_BACKWARD_ENABLE_INTEL      ( 0xAA << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_DUAL_ENABLE_INTEL          ( 0xFF << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_0_FORWARD_ENABLE_INTEL     ( 0x1 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_0_BACKWARD_ENABLE_INTEL    ( 0x2 << 24 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_1_FORWARD_ENABLE_INTEL     ( 0x1 << 26 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_1_BACKWARD_ENABLE_INTEL    ( 0x2 << 26 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_2_FORWARD_ENABLE_INTEL     ( 0x1 << 28 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_2_BACKWARD_ENABLE_INTEL    ( 0x2 << 28 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_3_FORWARD_ENABLE_INTEL     ( 0x1 << 30 )
+#define CL_AVC_ME_SKIP_BLOCK_8x8_3_BACKWARD_ENABLE_INTEL    ( 0x2 << 30 )
+
+#define CL_AVC_ME_BLOCK_BASED_SKIP_4x4_INTEL                0x00
+#define CL_AVC_ME_BLOCK_BASED_SKIP_8x8_INTEL                0x80
+
+#define CL_AVC_ME_INTRA_16x16_INTEL                         0x0
+#define CL_AVC_ME_INTRA_8x8_INTEL                           0x1
+#define CL_AVC_ME_INTRA_4x4_INTEL                           0x2
+
+#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_16x16_INTEL     0x6
+#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_8x8_INTEL       0x5
+#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_4x4_INTEL       0x3
+
+#define CL_AVC_ME_INTRA_NEIGHBOR_LEFT_MASK_ENABLE_INTEL         0x60
+#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_MASK_ENABLE_INTEL        0x10
+#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_RIGHT_MASK_ENABLE_INTEL  0x8
+#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_LEFT_MASK_ENABLE_INTEL   0x4
+
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_INTEL            0x0
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_INTEL          0x1
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DC_INTEL                  0x2
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_LEFT_INTEL  0x3
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_RIGHT_INTEL 0x4
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_PLANE_INTEL               0x4
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_RIGHT_INTEL      0x5
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_DOWN_INTEL     0x6
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_LEFT_INTEL       0x7
+#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_UP_INTEL       0x8
+#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_DC_INTEL                0x0
+#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_HORIZONTAL_INTEL        0x1
+#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_VERTICAL_INTEL          0x2
+#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_PLANE_INTEL             0x3
+
+#define CL_AVC_ME_FRAME_FORWARD_INTEL                       0x1
+#define CL_AVC_ME_FRAME_BACKWARD_INTEL                      0x2
+#define CL_AVC_ME_FRAME_DUAL_INTEL                          0x3
+
+#define CL_AVC_ME_SLICE_TYPE_PRED_INTEL                     0x0
+#define CL_AVC_ME_SLICE_TYPE_BPRED_INTEL                    0x1
+#define CL_AVC_ME_SLICE_TYPE_INTRA_INTEL                    0x2
+
+#define CL_AVC_ME_INTERLACED_SCAN_TOP_FIELD_INTEL           0x0
+#define CL_AVC_ME_INTERLACED_SCAN_BOTTOM_FIELD_INTEL        0x1
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __CL_EXT_INTEL_H */

benchmarks/new_opencl/include/CL/cl_gl.h

@@ -0,0 +1,171 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+
+#ifndef __OPENCL_CL_GL_H
+#define __OPENCL_CL_GL_H
+
+#include <CL/cl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef cl_uint     cl_gl_object_type;
+typedef cl_uint     cl_gl_texture_info;
+typedef cl_uint     cl_gl_platform_info;
+typedef struct __GLsync *cl_GLsync;
+
+/* cl_gl_object_type = 0x2000 - 0x200F enum values are currently taken           */
+#define CL_GL_OBJECT_BUFFER                     0x2000
+#define CL_GL_OBJECT_TEXTURE2D                  0x2001
+#define CL_GL_OBJECT_TEXTURE3D                  0x2002
+#define CL_GL_OBJECT_RENDERBUFFER               0x2003
+#ifdef CL_VERSION_1_2
+#define CL_GL_OBJECT_TEXTURE2D_ARRAY            0x200E
+#define CL_GL_OBJECT_TEXTURE1D                  0x200F
+#define CL_GL_OBJECT_TEXTURE1D_ARRAY            0x2010
+#define CL_GL_OBJECT_TEXTURE_BUFFER             0x2011
+#endif
+
+/* cl_gl_texture_info           */
+#define CL_GL_TEXTURE_TARGET                    0x2004
+#define CL_GL_MIPMAP_LEVEL                      0x2005
+#ifdef CL_VERSION_1_2
+#define CL_GL_NUM_SAMPLES                       0x2012
+#endif
+
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromGLBuffer(cl_context     context,
+                     cl_mem_flags   flags,
+                     cl_GLuint      bufobj,
+                     cl_int *       errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+#ifdef CL_VERSION_1_2
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromGLTexture(cl_context      context,
+                      cl_mem_flags    flags,
+                      cl_GLenum       target,
+                      cl_GLint        miplevel,
+                      cl_GLuint       texture,
+                      cl_int *        errcode_ret) CL_API_SUFFIX__VERSION_1_2;
+
+#endif
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromGLRenderbuffer(cl_context   context,
+                           cl_mem_flags flags,
+                           cl_GLuint    renderbuffer,
+                           cl_int *     errcode_ret) CL_API_SUFFIX__VERSION_1_0;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetGLObjectInfo(cl_mem                memobj,
+                  cl_gl_object_type *   gl_object_type,
+                  cl_GLuint *           gl_object_name) CL_API_SUFFIX__VERSION_1_0;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetGLTextureInfo(cl_mem               memobj,
+                   cl_gl_texture_info   param_name,
+                   size_t               param_value_size,
+                   void *               param_value,
+                   size_t *             param_value_size_ret) CL_API_SUFFIX__VERSION_1_0;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueAcquireGLObjects(cl_command_queue      command_queue,
+                          cl_uint               num_objects,
+                          const cl_mem *        mem_objects,
+                          cl_uint               num_events_in_wait_list,
+                          const cl_event *      event_wait_list,
+                          cl_event *            event) CL_API_SUFFIX__VERSION_1_0;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueReleaseGLObjects(cl_command_queue      command_queue,
+                          cl_uint               num_objects,
+                          const cl_mem *        mem_objects,
+                          cl_uint               num_events_in_wait_list,
+                          const cl_event *      event_wait_list,
+                          cl_event *            event) CL_API_SUFFIX__VERSION_1_0;
+
+
+/* Deprecated OpenCL 1.1 APIs */
+extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL
+clCreateFromGLTexture2D(cl_context      context,
+                        cl_mem_flags    flags,
+                        cl_GLenum       target,
+                        cl_GLint        miplevel,
+                        cl_GLuint       texture,
+                        cl_int *        errcode_ret) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED;
+
+extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL
+clCreateFromGLTexture3D(cl_context      context,
+                        cl_mem_flags    flags,
+                        cl_GLenum       target,
+                        cl_GLint        miplevel,
+                        cl_GLuint       texture,
+                        cl_int *        errcode_ret) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED;
+
+/* cl_khr_gl_sharing extension  */
+
+#define cl_khr_gl_sharing 1
+
+typedef cl_uint     cl_gl_context_info;
+
+/* Additional Error Codes  */
+#define CL_INVALID_GL_SHAREGROUP_REFERENCE_KHR  -1000
+
+/* cl_gl_context_info  */
+#define CL_CURRENT_DEVICE_FOR_GL_CONTEXT_KHR    0x2006
+#define CL_DEVICES_FOR_GL_CONTEXT_KHR           0x2007
+
+/* Additional cl_context_properties  */
+#define CL_GL_CONTEXT_KHR                       0x2008
+#define CL_EGL_DISPLAY_KHR                      0x2009
+#define CL_GLX_DISPLAY_KHR                      0x200A
+#define CL_WGL_HDC_KHR                          0x200B
+#define CL_CGL_SHAREGROUP_KHR                   0x200C
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetGLContextInfoKHR(const cl_context_properties * properties,
+                      cl_gl_context_info            param_name,
+                      size_t                        param_value_size,
+                      void *                        param_value,
+                      size_t *                      param_value_size_ret) CL_API_SUFFIX__VERSION_1_0;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetGLContextInfoKHR_fn)(
+    const cl_context_properties * properties,
+    cl_gl_context_info            param_name,
+    size_t                        param_value_size,
+    void *                        param_value,
+    size_t *                      param_value_size_ret);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_GL_H */

benchmarks/new_opencl/include/CL/cl_gl_ext.h

@@ -0,0 +1,52 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+
+#ifndef __OPENCL_CL_GL_EXT_H
+#define __OPENCL_CL_GL_EXT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <CL/cl_gl.h>
+
+/* 
+ *  cl_khr_gl_event extension
+ */
+#define CL_COMMAND_GL_FENCE_SYNC_OBJECT_KHR     0x200D
+
+extern CL_API_ENTRY cl_event CL_API_CALL
+clCreateEventFromGLsyncKHR(cl_context context,
+                           cl_GLsync  cl_GLsync,
+                           cl_int *   errcode_ret) CL_EXT_SUFFIX__VERSION_1_1;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif	/* __OPENCL_CL_GL_EXT_H  */

benchmarks/new_opencl/include/CL/cl_va_api_media_sharing_intel.h

@@ -0,0 +1,172 @@
+/**********************************************************************************
+ * Copyright (c) 2008-2019 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ **********************************************************************************/
+/*****************************************************************************\
+
+Copyright (c) 2013-2019 Intel Corporation All Rights Reserved.
+
+THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
+CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
+NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
+MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+File Name: cl_va_api_media_sharing_intel.h
+
+Abstract:
+
+Notes:
+
+\*****************************************************************************/
+
+
+#ifndef __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H
+#define __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H
+
+#include <CL/cl.h>
+#include <CL/cl_platform.h>
+#include <va/va.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/******************************************
+* cl_intel_va_api_media_sharing extension *
+*******************************************/
+
+#define cl_intel_va_api_media_sharing 1
+
+/* error codes */
+#define CL_INVALID_VA_API_MEDIA_ADAPTER_INTEL               -1098
+#define CL_INVALID_VA_API_MEDIA_SURFACE_INTEL               -1099
+#define CL_VA_API_MEDIA_SURFACE_ALREADY_ACQUIRED_INTEL      -1100
+#define CL_VA_API_MEDIA_SURFACE_NOT_ACQUIRED_INTEL          -1101
+
+/* cl_va_api_device_source_intel */
+#define CL_VA_API_DISPLAY_INTEL                             0x4094
+
+/* cl_va_api_device_set_intel */
+#define CL_PREFERRED_DEVICES_FOR_VA_API_INTEL               0x4095
+#define CL_ALL_DEVICES_FOR_VA_API_INTEL                     0x4096
+
+/* cl_context_info */
+#define CL_CONTEXT_VA_API_DISPLAY_INTEL                     0x4097
+
+/* cl_mem_info */
+#define CL_MEM_VA_API_MEDIA_SURFACE_INTEL                   0x4098
+
+/* cl_image_info */
+#define CL_IMAGE_VA_API_PLANE_INTEL                         0x4099
+
+/* cl_command_type */
+#define CL_COMMAND_ACQUIRE_VA_API_MEDIA_SURFACES_INTEL      0x409A
+#define CL_COMMAND_RELEASE_VA_API_MEDIA_SURFACES_INTEL      0x409B
+
+typedef cl_uint cl_va_api_device_source_intel;
+typedef cl_uint cl_va_api_device_set_intel;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clGetDeviceIDsFromVA_APIMediaAdapterINTEL(
+    cl_platform_id                platform,
+    cl_va_api_device_source_intel media_adapter_type,
+    void*                         media_adapter,
+    cl_va_api_device_set_intel    media_adapter_set,
+    cl_uint                       num_entries,
+    cl_device_id*                 devices,
+    cl_uint*                      num_devices) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL * clGetDeviceIDsFromVA_APIMediaAdapterINTEL_fn)(
+    cl_platform_id                platform,
+    cl_va_api_device_source_intel media_adapter_type,
+    void*                         media_adapter,
+    cl_va_api_device_set_intel    media_adapter_set,
+    cl_uint                       num_entries,
+    cl_device_id*                 devices,
+    cl_uint*                      num_devices) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateFromVA_APIMediaSurfaceINTEL(
+    cl_context                    context,
+    cl_mem_flags                  flags,
+    VASurfaceID*                  surface,
+    cl_uint                       plane,
+    cl_int*                       errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL * clCreateFromVA_APIMediaSurfaceINTEL_fn)(
+    cl_context                    context,
+    cl_mem_flags                  flags,
+    VASurfaceID*                  surface,
+    cl_uint                       plane,
+    cl_int*                       errcode_ret) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueAcquireVA_APIMediaSurfacesINTEL(
+    cl_command_queue              command_queue,
+    cl_uint                       num_objects,
+    const cl_mem*                 mem_objects,
+    cl_uint                       num_events_in_wait_list,
+    const cl_event*               event_wait_list,
+    cl_event*                     event) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireVA_APIMediaSurfacesINTEL_fn)(
+    cl_command_queue              command_queue,
+    cl_uint                       num_objects,
+    const cl_mem*                 mem_objects,
+    cl_uint                       num_events_in_wait_list,
+    const cl_event*               event_wait_list,
+    cl_event*                     event) CL_EXT_SUFFIX__VERSION_1_2;
+
+extern CL_API_ENTRY cl_int CL_API_CALL
+clEnqueueReleaseVA_APIMediaSurfacesINTEL(
+    cl_command_queue              command_queue,
+    cl_uint                       num_objects,
+    const cl_mem*                 mem_objects,
+    cl_uint                       num_events_in_wait_list,
+    const cl_event*               event_wait_list,
+    cl_event*                     event) CL_EXT_SUFFIX__VERSION_1_2;
+
+typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseVA_APIMediaSurfacesINTEL_fn)(
+    cl_command_queue              command_queue,
+    cl_uint                       num_objects,
+    const cl_mem*                 mem_objects,
+    cl_uint                       num_events_in_wait_list,
+    const cl_event*               event_wait_list,
+    cl_event*                     event) CL_EXT_SUFFIX__VERSION_1_2;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H */
+

benchmarks/new_opencl/include/CL/cl_version.h

@@ -0,0 +1,86 @@
+/*******************************************************************************
+ * Copyright (c) 2018 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ ******************************************************************************/
+
+#ifndef __CL_VERSION_H
+#define __CL_VERSION_H
+
+/* Detect which version to target */
+#if !defined(CL_TARGET_OPENCL_VERSION)
+#pragma message("cl_version.h: CL_TARGET_OPENCL_VERSION is not defined. Defaulting to 220 (OpenCL 2.2)")
+#define CL_TARGET_OPENCL_VERSION 220
+#endif
+#if CL_TARGET_OPENCL_VERSION != 100 && \
+    CL_TARGET_OPENCL_VERSION != 110 && \
+    CL_TARGET_OPENCL_VERSION != 120 && \
+    CL_TARGET_OPENCL_VERSION != 200 && \
+    CL_TARGET_OPENCL_VERSION != 210 && \
+    CL_TARGET_OPENCL_VERSION != 220
+#pragma message("cl_version: CL_TARGET_OPENCL_VERSION is not a valid value (100, 110, 120, 200, 210, 220). Defaulting to 220 (OpenCL 2.2)")
+#undef CL_TARGET_OPENCL_VERSION
+#define CL_TARGET_OPENCL_VERSION 220
+#endif
+
+
+/* OpenCL Version */
+#if CL_TARGET_OPENCL_VERSION >= 220 && !defined(CL_VERSION_2_2)
+#define CL_VERSION_2_2  1
+#endif
+#if CL_TARGET_OPENCL_VERSION >= 210 && !defined(CL_VERSION_2_1)
+#define CL_VERSION_2_1  1
+#endif
+#if CL_TARGET_OPENCL_VERSION >= 200 && !defined(CL_VERSION_2_0)
+#define CL_VERSION_2_0  1
+#endif
+#if CL_TARGET_OPENCL_VERSION >= 120 && !defined(CL_VERSION_1_2)
+#define CL_VERSION_1_2  1
+#endif
+#if CL_TARGET_OPENCL_VERSION >= 110 && !defined(CL_VERSION_1_1)
+#define CL_VERSION_1_1  1
+#endif
+#if CL_TARGET_OPENCL_VERSION >= 100 && !defined(CL_VERSION_1_0)
+#define CL_VERSION_1_0  1
+#endif
+
+/* Allow deprecated APIs for older OpenCL versions. */
+#if CL_TARGET_OPENCL_VERSION <= 210 && !defined(CL_USE_DEPRECATED_OPENCL_2_1_APIS)
+#define CL_USE_DEPRECATED_OPENCL_2_1_APIS
+#endif
+#if CL_TARGET_OPENCL_VERSION <= 200 && !defined(CL_USE_DEPRECATED_OPENCL_2_0_APIS)
+#define CL_USE_DEPRECATED_OPENCL_2_0_APIS
+#endif
+#if CL_TARGET_OPENCL_VERSION <= 120 && !defined(CL_USE_DEPRECATED_OPENCL_1_2_APIS)
+#define CL_USE_DEPRECATED_OPENCL_1_2_APIS
+#endif
+#if CL_TARGET_OPENCL_VERSION <= 110 && !defined(CL_USE_DEPRECATED_OPENCL_1_1_APIS)
+#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
+#endif
+#if CL_TARGET_OPENCL_VERSION <= 100 && !defined(CL_USE_DEPRECATED_OPENCL_1_0_APIS)
+#define CL_USE_DEPRECATED_OPENCL_1_0_APIS
+#endif
+
+#endif  /* __CL_VERSION_H */

benchmarks/new_opencl/include/CL/opencl.h

@@ -0,0 +1,47 @@
+/*******************************************************************************
+ * Copyright (c) 2008-2015 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and/or associated documentation files (the
+ * "Materials"), to deal in the Materials without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Materials, and to
+ * permit persons to whom the Materials are 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 Materials.
+ *
+ * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
+ * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
+ * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
+ *    https://www.khronos.org/registry/
+ *
+ * THE MATERIALS ARE 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
+ * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ ******************************************************************************/
+
+/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */
+
+#ifndef __OPENCL_H
+#define __OPENCL_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <CL/cl.h>
+#include <CL/cl_gl.h>
+#include <CL/cl_gl_ext.h>
+#include <CL/cl_ext.h>
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* __OPENCL_H   */

benchmarks/new_opencl/nearn/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = nearn
+
+SRCS = main.cc clutils.cpp utils.cpp
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/nearn/README.txt

@@ -0,0 +1,33 @@
+The Nearest Neighbor application computes the nearest location to a specific 
+latitude and longitude for a number of hurricanes (data from: http://weather.unisys.com/hurricane/).
+
+The Makefile may need to be adjusted for different machines, but it was written for Mac OS X and
+Linux with either NVIDIA or AMD OpenCL SDKs.
+
+The hurricane data is located in a number of data files that are copied into the working
+directory by the Makefile.  A separate text file lists the names of the data files that
+will be used, and it is this text file that should be passed to the application (see usage, below).
+
+Nearest Neighbor Usage
+
+nearestNeighbor [filename] -r [int] -lat [float] -lng [float] [-hqt] [-p [int] -d [int]]
+
+example:
+$ ./nearestNeighbor filelist.txt -r 5 -lat 30 -lng 90
+
+filename     the filename that lists the data input files
+-r [int]     the number of records to return (default: 10)
+-lat [float] the latitude for nearest neighbors (default: 0)
+-lng [float] the longitude for nearest neighbors (default: 0)
+
+-h, --help   Display the help file
+-q           Quiet mode. Suppress all text output.
+-t           Print timing information.
+
+-p [int]     Choose the platform (must choose both platform and device)
+-d [int]     Choose the device (must choose both platform and device)
+
+
+Notes: 1. The filename is required as the first parameter.
+       2. If you declare either the device or the platform,
+          you must declare both.

benchmarks/new_opencl/nearn/clutils.h

@@ -0,0 +1,281 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#ifndef __CL_UTILS_H__
+#define __CL_UTILS_H__
+
+#include <CL/cl.h>
+
+// The cl_time type is OS specific
+#ifdef _WIN32
+#include <tchar.h>
+#include <Windows.h>
+typedef __int64 cl_time; 
+#else
+#include <sys/time.h>
+typedef double cl_time;
+#endif
+
+//-------------------------------------------------------
+// Initialization and Cleanup
+//-------------------------------------------------------
+
+// Detects platforms and devices, creates context and command queue
+cl_context cl_init(char devicePreference='\0');
+
+// Creates a context given a platform and a device
+cl_context cl_init_context(int platform,int dev,int quiet=0);
+
+// Releases resources used by clutils
+void    cl_cleanup();
+
+// Releases a kernel object
+void    cl_freeKernel(cl_kernel kernel);
+
+// Releases a memory object
+void    cl_freeMem(cl_mem mem);
+
+// Releases a program object
+void    cl_freeProgram(cl_program program);
+
+// Returns the global command queue
+cl_command_queue cl_getCommandQueue();
+
+
+//-------------------------------------------------------
+// Synchronization functions
+//-------------------------------------------------------
+
+// Performs a clFinish on the command queue
+void    cl_sync();
+
+
+//-------------------------------------------------------
+// Memory allocation
+//-------------------------------------------------------
+
+// Allocates a regular buffer on the device
+cl_mem  cl_allocBuffer(size_t mem_size, 
+            cl_mem_flags flags = CL_MEM_READ_WRITE);
+
+// XXX I don't think this does exactly what we want it to do
+// Allocates a read-only buffer and transfers the data
+cl_mem  cl_allocBufferConst(size_t mem_size, void* host_ptr);
+
+// Allocates pinned memory on the host
+cl_mem  cl_allocBufferPinned(size_t mem_size);
+
+// Allocates an image on the device
+cl_mem  cl_allocImage(size_t height, size_t width, char type, 
+            cl_mem_flags flags = CL_MEM_READ_WRITE);
+
+
+
+//-------------------------------------------------------
+// Data transfers
+//-------------------------------------------------------
+
+// Copies a buffer from the device to pinned memory on the host and 
+// maps it so it can be read
+void*   cl_copyAndMapBuffer(cl_mem dst, cl_mem src, size_t size); 
+
+// Copies from one buffer to another
+void    cl_copyBufferToBuffer(cl_mem dst, cl_mem src, size_t size);
+
+// Copies data to a buffer on the device
+void    cl_copyBufferToDevice(cl_mem dst, void *src, size_t mem_size, 
+            cl_bool blocking = CL_TRUE);
+
+// Copies data to an image on the device
+void cl_copyImageToDevice(cl_mem dst, void* src, size_t height, size_t width);
+
+// Copies an image from the device to the host
+void    cl_copyImageToHost(void* dst, cl_mem src, size_t height, size_t width);
+
+// Copies data from a device buffer to the host
+void    cl_copyBufferToHost(void *dst, cl_mem src, size_t mem_size, 
+            cl_bool blocking = CL_TRUE);
+
+// Copies data from a buffer on the device to an image on the device
+void    cl_copyBufferToImage(cl_mem src, cl_mem dst, int height, int width);
+
+// Maps a buffer
+void*   cl_mapBuffer(cl_mem mem, size_t mem_size, cl_mem_flags flags);
+
+// Unmaps a buffer
+void    cl_unmapBuffer(cl_mem mem, void *ptr);
+
+// Writes data to a zero-copy buffer on the device
+void    cl_writeToZCBuffer(cl_mem mem, void* data, size_t size);
+
+//-------------------------------------------------------
+// Program and kernels
+//-------------------------------------------------------
+
+// Compiles a program
+cl_program  cl_compileProgram(char* kernelPath, char* compileoptions, 
+                bool verboseoptions = 0);
+
+// Creates a kernel
+cl_kernel   cl_createKernel(cl_program program, const char* kernelName);
+
+
+// Sets a kernel argument
+void        cl_setKernelArg(cl_kernel kernel, unsigned int index, size_t size, 
+                void* data);
+
+
+//-------------------------------------------------------
+// Profiling/events
+//-------------------------------------------------------
+
+// Computes the execution time (start to end) for an event
+double  cl_computeExecTime(cl_event);
+
+// Compute the elapsed time between two CPU timer values
+double  cl_computeTime(cl_time start, cl_time end); 
+
+// Creates an event from CPU timers
+void    cl_createUserEvent(cl_time start, cl_time end, char* desc);
+
+// Disable logging of events
+void    cl_disableEvents();
+
+// Enable logging of events
+void    cl_enableEvents();
+
+// Query the current system time
+void    cl_getTime(cl_time* time); 
+
+// Calls a function which prints events to the terminal
+void    cl_printEvents();
+
+// Calls a function which writes the events to a file
+void    cl_writeEventsToFile(char* path);
+
+
+//-------------------------------------------------------
+// Error handling
+//-------------------------------------------------------
+
+// Compare a status value to CL_SUCCESS and optionally exit on error
+int     cl_errChk(const cl_int status, const char *msg, bool exitOnErr);
+
+// Queries the supported image formats for the device and prints
+// them to the screen
+void    printSupportedImageFormats();
+
+//-------------------------------------------------------
+// Platform and device information
+//-------------------------------------------------------
+
+bool    cl_deviceIsAMD(cl_device_id dev=NULL);
+bool    cl_deviceIsNVIDIA(cl_device_id dev=NULL);
+bool    cl_platformIsNVIDIA(cl_platform_id plat=NULL);
+char*   cl_getDeviceDriverVersion(cl_device_id dev=NULL);
+char*   cl_getDeviceName(cl_device_id dev=NULL);
+char*   cl_getDeviceVendor(cl_device_id dev=NULL);
+char*   cl_getDeviceVersion(cl_device_id dev=NULL);
+char*   cl_getPlatformName(cl_platform_id platform);
+char*   cl_getPlatformVendor(cl_platform_id platform);
+
+//-------------------------------------------------------
+// Utility functions
+//-------------------------------------------------------
+
+char* catStringWithInt(const char* str, int integer);
+
+char* itoa_portable(int value, char* result, int base);
+
+//-------------------------------------------------------
+// Data types
+//-------------------------------------------------------
+typedef struct{
+    int x;
+    int y;
+} int2;
+
+typedef struct{
+    float x;
+    float y;
+}float2;
+
+typedef struct{
+    float x;
+    float y;
+    float z;
+    float w;
+}float4;
+
+//-------------------------------------------------------
+// Defines
+//-------------------------------------------------------
+
+#define MAX_ERR_VAL 64
+
+#define NUM_PROGRAMS 7
+
+#define NUM_KERNELS 13
+#define KERNEL_INIT_DET 0 
+#define KERNEL_BUILD_DET 1 
+#define KERNEL_SURF_DESC 2
+#define KERNEL_NORM_DESC 3
+#define KERNEL_NON_MAX_SUP 4
+#define KERNEL_GET_ORIENT1 5
+#define KERNEL_GET_ORIENT2 6
+#define KERNEL_NN 7
+#define KERNEL_SCAN 8
+#define KERNEL_SCAN4 9
+#define KERNEL_TRANSPOSE 10
+#define KERNEL_SCANIMAGE 11
+#define KERNEL_TRANSPOSEIMAGE 12
+
+#endif

benchmarks/new_opencl/nearn/filelist.txt

@@ -0,0 +1,4 @@
+cane4_0.db
+cane4_1.db
+cane4_2.db
+cane4_3.db

benchmarks/new_opencl/nearn/ipoint.h

@@ -0,0 +1,29 @@
+/*********************************************************** 
+*  --- OpenSURF ---                                        *
+*  This library is distributed under the GNU GPL. Please   *
+*  contact chris.evans@irisys.co.uk for more information.  *
+*                                                          *
+*  C. Evans, Research Into Robust Visual Features,         *
+*  MSc University of Bristol, 2008.                        *
+*                                                          *
+************************************************************/
+
+#ifndef IPOINT_H
+#define IPOINT_H
+
+#include <vector>
+#include <math.h>
+
+
+
+//-------------------------------------------------------
+typedef struct{
+        int x;
+        int y;
+		float descriptor[64];
+	} Ipoint;
+
+//-------------------------------------------------------
+
+  typedef std::vector<Ipoint> IpVec;
+#endif

benchmarks/new_opencl/nearn/kernel.cl

@@ -0,0 +1,22 @@
+//#pragma OPENCL EXTENSION cl_khr_byte_addressable_store : enable
+
+typedef struct latLong
+    {
+        float lat;
+        float lng;
+    } LatLong;
+
+__kernel void NearestNeighbor(__global LatLong *d_locations,
+							  __global float *d_distances,
+							  const int numRecords,
+							  const float lat,
+							  const float lng) {
+	 int globalId = get_global_id(0);
+							  
+     if (globalId < numRecords) {
+         __global LatLong *latLong = d_locations+globalId;
+    
+         __global float *dist=d_distances+globalId;
+         *dist = (float)sqrt((lat-latLong->lat)*(lat-latLong->lat)+(lng-latLong->lng)*(lng-latLong->lng));
+	 }
+}

benchmarks/new_opencl/nearn/main.cc

@@ -0,0 +1,346 @@
+#ifndef __NEAREST_NEIGHBOR__
+#define __NEAREST_NEIGHBOR__
+
+#include "nearestNeighbor.h"
+
+cl_context context = NULL;
+
+int main(int argc, char *argv[]) {
+  std::vector<Record> records;
+  float *recordDistances;
+  // LatLong locations[REC_WINDOW];
+  std::vector<LatLong> locations;
+  int i;
+  // args
+  char filename[100];
+  int resultsCount = 5, quiet = 0, timing = 0, platform = -1, device = -1;
+  float lat = 30, lng = 90;
+
+  // parse command line
+  if (parseCommandline(argc, argv, filename, &resultsCount, &lat, &lng, &quiet,
+                       &timing, &platform, &device)) {
+    printUsage();
+    return 0;
+  }
+
+  int numRecords = loadData(filename, records, locations);
+
+  // for(i=0;i<numRecords;i++)
+  //    printf("%s, %f,
+  //    %f\n",(records[i].recString),locations[i].lat,locations[i].lng);
+
+  printf("Number of records: %d\n", numRecords);
+  printf("Finding the %d closest neighbors.\n", resultsCount);
+
+  if (resultsCount > numRecords)
+    resultsCount = numRecords;
+
+  context = cl_init_context(platform, device, quiet);
+  
+  recordDistances = OpenClFindNearestNeighbors(context, numRecords, locations,
+                                               lat, lng, timing);
+
+  // find the resultsCount least distances
+  findLowest(records, recordDistances, numRecords, resultsCount);
+
+  // print out results
+  if (!quiet)
+    for (i = 0; i < resultsCount; i++) {
+      printf("%s --> Distance=%f\n", records[i].recString, records[i].distance);
+    }
+  free(recordDistances);
+  return 0;
+}
+
+float *OpenClFindNearestNeighbors(cl_context context, int numRecords,
+                                  std::vector<LatLong> &locations, float lat,
+                                  float lng, int timing) {
+
+  // 1. set up kernel
+  cl_kernel NN_kernel;
+  cl_int status;
+  cl_program cl_NN_program;
+  cl_NN_program = cl_compileProgram((char *)"nearestNeighbor_kernel.cl", NULL);
+
+  NN_kernel = clCreateKernel(cl_NN_program, "NearestNeighbor", &status);
+  status =
+      cl_errChk(status, (char *)"Error Creating Nearest Neighbor kernel", true);
+  if (status)
+    exit(1);
+  // 2. set up memory on device and send ipts data to device
+  // copy ipts(1,2) to device
+  // also need to alloate memory for the distancePoints
+  cl_mem d_locations;
+  cl_mem d_distances;
+
+  cl_int error = 0;
+
+  d_locations = clCreateBuffer(context, CL_MEM_READ_ONLY,
+                               sizeof(LatLong) * numRecords, NULL, &error);
+
+  d_distances = clCreateBuffer(context, CL_MEM_READ_WRITE,
+                               sizeof(float) * numRecords, NULL, &error);
+
+  cl_command_queue command_queue = cl_getCommandQueue();
+  cl_event writeEvent, kernelEvent, readEvent;
+  error = clEnqueueWriteBuffer(command_queue, d_locations,
+                               1, // change to 0 for nonblocking write
+                               0, // offset
+                               sizeof(LatLong) * numRecords, &locations[0], 0,
+                               NULL, &writeEvent);
+
+  // 3. send arguments to device
+  cl_int argchk;
+  argchk = clSetKernelArg(NN_kernel, 0, sizeof(cl_mem), (void *)&d_locations);
+  argchk |= clSetKernelArg(NN_kernel, 1, sizeof(cl_mem), (void *)&d_distances);
+  argchk |= clSetKernelArg(NN_kernel, 2, sizeof(int), (void *)&numRecords);
+  argchk |= clSetKernelArg(NN_kernel, 3, sizeof(float), (void *)&lat);
+  argchk |= clSetKernelArg(NN_kernel, 4, sizeof(float), (void *)&lng);
+
+  cl_errChk(argchk, "ERROR in Setting Nearest Neighbor kernel args", true);
+
+  // 4. enqueue kernel
+  size_t globalWorkSize[1];
+  globalWorkSize[0] = numRecords;
+  if (numRecords % 64)
+    globalWorkSize[0] += 64 - (numRecords % 64);
+  // printf("Global Work Size: %zu\n",globalWorkSize[0]);
+
+  error = clEnqueueNDRangeKernel(command_queue, NN_kernel, 1, 0, globalWorkSize,
+                                 NULL, 0, NULL, &kernelEvent);
+
+  cl_errChk(error, "ERROR in Executing Kernel NearestNeighbor", true);
+
+  // 5. transfer data off of device
+
+  // create distances std::vector
+  float *distances = (float *)malloc(sizeof(float) * numRecords);
+
+  error = clEnqueueReadBuffer(command_queue, d_distances,
+                              1, // change to 0 for nonblocking write
+                              0, // offset
+                              sizeof(float) * numRecords, distances, 0, NULL,
+                              &readEvent);
+
+  cl_errChk(error, "ERROR with clEnqueueReadBuffer", true);
+  if (timing) {
+    clFinish(command_queue);
+    cl_ulong eventStart, eventEnd, totalTime = 0;
+    printf("# Records\tWrite(s) [size]\t\tKernel(s)\tRead(s)  "
+           "[size]\t\tTotal(s)\n");
+    printf("%d        \t", numRecords);
+    // Write Buffer
+    error = clGetEventProfilingInfo(writeEvent, CL_PROFILING_COMMAND_START,
+                                    sizeof(cl_ulong), &eventStart, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Write Start)", true);
+    error = clGetEventProfilingInfo(writeEvent, CL_PROFILING_COMMAND_END,
+                                    sizeof(cl_ulong), &eventEnd, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Write End)", true);
+
+    printf("%f [%.2fMB]\t", (float)((eventEnd - eventStart) / 1e9),
+           (float)((sizeof(LatLong) * numRecords) / 1e6));
+    totalTime += eventEnd - eventStart;
+    // Kernel
+    error = clGetEventProfilingInfo(kernelEvent, CL_PROFILING_COMMAND_START,
+                                    sizeof(cl_ulong), &eventStart, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Kernel Start)", true);
+    error = clGetEventProfilingInfo(kernelEvent, CL_PROFILING_COMMAND_END,
+                                    sizeof(cl_ulong), &eventEnd, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Kernel End)", true);
+
+    printf("%f\t", (float)((eventEnd - eventStart) / 1e9));
+    totalTime += eventEnd - eventStart;
+    // Read Buffer
+    error = clGetEventProfilingInfo(readEvent, CL_PROFILING_COMMAND_START,
+                                    sizeof(cl_ulong), &eventStart, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Read Start)", true);
+    error = clGetEventProfilingInfo(readEvent, CL_PROFILING_COMMAND_END,
+                                    sizeof(cl_ulong), &eventEnd, NULL);
+    cl_errChk(error, "ERROR in Event Profiling (Read End)", true);
+
+    printf("%f [%.2fMB]\t", (float)((eventEnd - eventStart) / 1e9),
+           (float)((sizeof(float) * numRecords) / 1e6));
+    totalTime += eventEnd - eventStart;
+
+    printf("%f\n\n", (float)(totalTime / 1e9));
+  }
+  // 6. return finalized data and release buffers
+  clReleaseMemObject(d_locations);
+  clReleaseMemObject(d_distances);
+  return distances;
+}
+
+int loadData(char *filename, std::vector<Record> &records,
+             std::vector<LatLong> &locations) {
+  FILE *flist, *fp;
+  int i = 0;
+  char dbname[64];
+  int recNum = 0;
+
+  /**Main processing **/
+
+  int q = 0;
+
+  flist = fopen(filename, "r");
+  while (!feof(flist)) {
+    /**
+    * Read in REC_WINDOW records of length REC_LENGTH
+    * If this is the last file in the filelist, then done
+    * else open next file to be read next iteration
+    */       
+    if (fscanf(flist, "%s\n", dbname) != 1) {
+      printf("error reading filelist\n");
+      exit(0);
+    }
+    printf("loading db: %s\n", dbname);
+    fp = fopen(dbname, "r");
+    if (!fp) {
+      printf("error opening a db\n");
+      exit(1);
+    }    
+    // read each record
+    while (!feof(fp)) {
+      Record record;
+      LatLong latLong;
+      fgets(record.recString, 49, fp);
+      fgetc(fp); // newline
+      if (feof(fp))
+        break;
+
+      // parse for lat and long
+      char substr[6];
+
+      for (i = 0; i < 5; i++)
+        substr[i] = *(record.recString + i + 28);
+      substr[5] = '\0';
+      latLong.lat = atof(substr);
+
+      for (i = 0; i < 5; i++)
+        substr[i] = *(record.recString + i + 33);
+      substr[5] = '\0';
+      latLong.lng = atof(substr);
+
+      locations.push_back(latLong);
+      records.push_back(record);
+      recNum++;
+      if (0 == (recNum % 500))
+        break;
+    }
+    
+    if (++q == 3)
+        break;
+    fclose(fp);
+  }
+  fclose(flist);
+  return recNum;
+}
+
+void findLowest(std::vector<Record> &records, float *distances, int numRecords,
+                int topN) {
+  int i, j;
+  float val;
+  int minLoc;
+  Record *tempRec;
+  float tempDist;
+
+  for (i = 0; i < topN; i++) {
+    minLoc = i;
+    for (j = i; j < numRecords; j++) {
+      val = distances[j];
+      if (val < distances[minLoc])
+        minLoc = j;
+    }
+    // swap locations and distances
+    tempRec = &records[i];
+    records[i] = records[minLoc];
+    records[minLoc] = *tempRec;
+
+    tempDist = distances[i];
+    distances[i] = distances[minLoc];
+    distances[minLoc] = tempDist;
+
+    // add distance to the min we just found
+    records[i].distance = distances[i];
+  }
+}
+
+int parseCommandline(int argc, char *argv[], char *filename, int *r, float *lat,
+                     float *lng, int *q, int *t, int *p, int *d) {
+  int i;
+  // if (argc < 2) return 1; // error
+  strncpy(filename, "filelist.txt", 100);
+  char flag;
+
+  for (i = 1; i < argc; i++) {
+    if (argv[i][0] == '-') { // flag
+      flag = argv[i][1];
+      switch (flag) {
+      case 'r': // number of results
+        i++;
+        *r = atoi(argv[i]);
+        break;
+      case 'l':                  // lat or lng
+        if (argv[i][2] == 'a') { // lat
+          *lat = atof(argv[i + 1]);
+        } else { // lng
+          *lng = atof(argv[i + 1]);
+        }
+        i++;
+        break;
+      case 'h': // help
+        return 1;
+        break;
+      case 'q': // quiet
+        *q = 1;
+        break;
+      case 't': // timing
+        *t = 1;
+        break;
+      case 'p': // platform
+        i++;
+        *p = atoi(argv[i]);
+        break;
+      case 'd': // device
+        i++;
+        *d = atoi(argv[i]);
+        break;
+      }
+    }
+  }
+  if ((*d >= 0 && *p < 0) ||
+      (*p >= 0 &&
+       *d < 0)) // both p and d must be specified if either are specified
+    return 1;
+  return 0;
+}
+
+void printUsage() {
+  printf("Nearest Neighbor Usage\n");
+  printf("\n");
+  printf("nearestNeighbor [filename] -r [int] -lat [float] -lng [float] [-hqt] "
+         "[-p [int] -d [int]]\n");
+  printf("\n");
+  printf("example:\n");
+  printf("$ ./nearestNeighbor filelist.txt -r 5 -lat 30 -lng 90\n");
+  printf("\n");
+  printf("filename     the filename that lists the data input files\n");
+  printf("-r [int]     the number of records to return (default: 10)\n");
+  printf("-lat [float] the latitude for nearest neighbors (default: 0)\n");
+  printf("-lng [float] the longitude for nearest neighbors (default: 0)\n");
+  printf("\n");
+  printf("-h, --help   Display the help file\n");
+  printf("-q           Quiet mode. Suppress all text output.\n");
+  printf("-t           Print timing information.\n");
+  printf("\n");
+  printf("-p [int]     Choose the platform (must choose both platform and "
+         "device)\n");
+  printf("-d [int]     Choose the device (must choose both platform and "
+         "device)\n");
+  printf("\n");
+  printf("\n");
+  printf("Notes: 1. The filename is required as the first parameter.\n");
+  printf("       2. If you declare either the device or the platform,\n");
+  printf("          you must declare both.\n\n");
+}
+
+#endif

benchmarks/new_opencl/nearn/nearestNeighbor.h

@@ -0,0 +1,50 @@
+#ifndef _NEARESTNEIGHBOR
+#define _NEARESTNEIGHBOR
+
+#include <iostream>
+#include <vector>
+#include <float.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+// All OpenCL headers
+#if defined (__APPLE__) || defined(MACOSX)
+    #include <OpenCL/opencl.h>
+#else
+    #include <CL/opencl.h>
+#endif 
+
+#include "clutils.h"
+//#include "utils.h"
+
+#include <algorithm>
+
+
+	
+#define REC_LENGTH 49 // size of a record in db
+
+typedef struct latLong
+{
+  float lat;
+  float lng;
+} LatLong;
+
+typedef struct record
+{
+  char recString[REC_LENGTH];
+  float distance;
+} Record;
+
+float *OpenClFindNearestNeighbors(
+	cl_context context,
+	int numRecords,
+	std::vector<LatLong> &locations,float lat,float lng,
+	int timing);
+
+int loadData(char *filename,std::vector<Record> &records,std::vector<LatLong> &locations);
+void findLowest(std::vector<Record> &records,float *distances,int numRecords,int topN);
+void printUsage();
+int parseCommandline(int argc, char *argv[], char* filename,int *r,float *lat,float *lng,
+                     int *q, int *t, int *p, int *d);
+#endif

benchmarks/new_opencl/nearn/run

@@ -0,0 +1 @@
+./nn filelist.txt -r 5 -lat 30 -lng 90

benchmarks/new_opencl/nearn/utils.cpp

@@ -0,0 +1,204 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#include <stdio.h>
+#include <sys/stat.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include "utils.h"
+
+static bool usingImages = true;
+
+//! A wrapper for malloc that checks the return value
+void* alloc(size_t size) {
+
+    void* ptr = NULL;
+    ptr = malloc(size);
+    if(ptr == NULL) {
+        perror("malloc");
+        exit(-1);
+    }
+
+    return ptr;
+}
+
+// This function checks to make sure a file exists before we open it
+void checkFile(char* filename) 
+{
+    
+    struct stat fileStatus;
+    if(stat(filename, &fileStatus) != 0) {
+        printf("Error opening file: %s\n", filename);
+        exit(-1);
+    }
+    else {
+        if(!(S_IFREG & fileStatus.st_mode)) {
+            printf("File %s is not a regular file\n", filename);
+            exit(-1);
+        }
+    }
+}
+
+
+// This function checks to make sure a directory exists 
+void checkDir(char* dirpath) 
+{
+    
+    struct stat fileStatus;
+    if(stat(dirpath, &fileStatus) != 0) {
+        printf("Directory does not exist: %s\n", dirpath);
+        exit(-1);
+    }
+    else {
+        if(!(S_IFDIR & fileStatus.st_mode)) {
+            printf("Directory was not provided: %s\n", dirpath);
+            exit(-1);
+        }
+    }
+}
+
+// Parse the command line arguments
+void parseArguments(int argc, char** argv, char** input, char** events, 
+    char** ipts, char* devicePref, bool* verifyResults) 
+{
+    
+    for(int i = 2; i < argc; i++) {
+        if(strcmp(argv[i], "-d") == 0) {   // Event dump found
+            if(i == argc-1) {
+                printf("Usage: -e Needs directory path\n");
+                exit(-1);
+            }
+            devicePref[0] = argv[i+1][0];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-e") == 0) {   // Event dump found
+            if(i == argc-1) {
+                printf("Usage: -e Needs directory path\n");
+                exit(-1);
+            }
+            *events = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-i") == 0) {   // Input found
+            if(i == argc-1) {
+                printf("Usage: -i Needs directory path\n");
+                exit(-1);
+            }
+            *input = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-l") == 0) {   // Ipts dump found
+            if(i == argc-1) {
+                printf("Usage: -l Needs directory path\n");
+                exit(-1);
+            }
+            *ipts = argv[i+1];
+            i++;
+            continue;
+        }
+        if(strcmp(argv[i], "-n") == 0) {   // Don't use OpenCL images
+            setUsingImages(false);
+            continue;
+        }
+        if(strcmp(argv[i], "-v") == 0) {   // Verify results
+            *verifyResults = true;
+            continue;
+        }
+    }
+}
+
+
+// This function that takes a positive integer 'value' and returns
+// the nearest multiple of 'multiple' (used for padding columns)
+unsigned int roundUp(unsigned int value, unsigned int multiple) {
+
+   unsigned int remainder = value % multiple;
+   
+   // Make the value a multiple of multiple
+   if(remainder != 0) {
+      value += (multiple-remainder);
+   }
+
+   return value;
+}
+
+
+// Concatenate two strings and return a pointer to the new string
+char* smartStrcat(char* str1, char* str2) 
+{
+    char* newStr = NULL;
+
+    newStr = (char*)alloc((strlen(str1)+strlen(str2)+1)*sizeof(char));
+
+    strcpy(newStr, str1);
+    strcat(newStr, str2);
+
+    return newStr;
+}
+
+
+// Set the value of using images to true if they are being
+// used, or false if they are not
+void setUsingImages(bool val) 
+{
+    usingImages = val;
+}
+
+
+// Return whether or not images are being used
+bool isUsingImages() 
+{
+    return usingImages;
+}

benchmarks/new_opencl/nearn/utils.h

@@ -0,0 +1,84 @@
+/****************************************************************************\ 
+ * Copyright (c) 2011, Advanced Micro Devices, Inc.                           *
+ * All rights reserved.                                                       *
+ *                                                                            *
+ * Redistribution and use in source and binary forms, with or without         *
+ * modification, are permitted provided that the following conditions         *
+ * are met:                                                                   *
+ *                                                                            *
+ * Redistributions of source code must retain the above copyright notice,     *
+ * this list of conditions and the following disclaimer.                      *
+ *                                                                            *
+ * Redistributions in binary form must reproduce the above copyright notice,  *
+ * this list of conditions and the following disclaimer in the documentation  *
+ * and/or other materials provided with the distribution.                     *
+ *                                                                            *
+ * Neither the name of the copyright holder nor the names of its contributors *
+ * may be used to endorse or promote products derived from this software      *
+ * without specific prior written permission.                                 *
+ *                                                                            *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS        *
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  *
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR          *
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,      *
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,        *
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR         *
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF     *
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING       *
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS         *
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.               *
+ *                                                                            *
+ * If you use the software (in whole or in part), you shall adhere to all     *
+ * applicable U.S., European, and other export laws, including but not        *
+ * limited to the U.S. Export Administration Regulations (“EAR”), (15 C.F.R.  *
+ * Sections 730 through 774), and E.U. Council Regulation (EC) No 1334/2000   *
+ * of 22 June 2000.  Further, pursuant to Section 740.6 of the EAR, you       *
+ * hereby certify that, except pursuant to a license granted by the United    *
+ * States Department of Commerce Bureau of Industry and Security or as        *
+ * otherwise permitted pursuant to a License Exception under the U.S. Export  *
+ * Administration Regulations ("EAR"), you will not (1) export, re-export or  *
+ * release to a national of a country in Country Groups D:1, E:1 or E:2 any   *
+ * restricted technology, software, or source code you receive hereunder,     *
+ * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such *
+ * technology or software, if such foreign produced direct product is subject *
+ * to national security controls as identified on the Commerce Control List   *
+ *(currently found in Supplement 1 to Part 774 of EAR).  For the most current *
+ * Country Group listings, or for additional information about the EAR or     *
+ * your obligations under those regulations, please refer to the U.S. Bureau  *
+ * of Industry and Security’s website at http://www.bis.doc.gov/.             *
+ \****************************************************************************/
+
+#ifndef _UTILS_
+#define _UTILS_
+
+// Wrapper for malloc
+void* alloc(size_t size);
+
+// Checks for existence of directory
+void checkDir(char* dirpath);
+
+// Check for existence of file
+void checkFile(char* filename);
+
+// Parse the input command line options to the program
+void parseArguments(int argc, char** argv, char** input, char** events, 
+    char** ipts, char* devicePref, bool* verifyResults);
+
+
+// Print the program usage information
+void printUsage();
+
+// Rounds up size to the nearest multiple of multiple
+unsigned int roundUp(unsigned int value, unsigned int multiple);
+
+// Concatenate two strings, creating a new one
+char* smartStrcat(char* str1, char* str2);
+
+// Set the value of usingImages
+void setUsingImages(bool val);
+
+// Return whether or not images are being used
+bool isUsingImages(); 
+
+#endif

benchmarks/new_opencl/results.txt

@@ -0,0 +1,7 @@
+# Dynamic Instructions: -1
+# of total cycles:      2519
+# of forwarding stalls: 0
+# of branch stalls:     0
+# CPI:                  -2519
+# time to simulate:     4.94066e-323 milliseconds
+# GRADE:                Failed on test: 0

benchmarks/new_opencl/saxpy/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = saxpy
+
+SRCS = main.cc
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/saxpy/kernel.cl

@@ -0,0 +1,5 @@
+__kernel void saxpy(__global float *src, __global float *dst, float factor)
+{
+  long i = get_global_id(0);
+  dst[i] += src[i] * factor;
+}

benchmarks/new_opencl/saxpy/main.cc

@@ -0,0 +1,221 @@
+/*
+ *  Simple OpenCL demo program
+ *
+ *  Copyright (C) 2009  Clifford Wolf <clifford@clifford.at>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ *  gcc -o cldemo -std=gnu99 -Wall -I/usr/include/nvidia-current cldemo.c
+ * -lOpenCL
+ *
+ */
+
+#include <CL/cl.h>
+#include <fstream>
+#include <iostream>
+#include <sstream>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+//#define NUM_DATA 65536
+#define NUM_DATA 4096
+
+#define CL_CHECK(_expr)                                                        \
+  do {                                                                         \
+    cl_int _err = _expr;                                                       \
+    if (_err == CL_SUCCESS)                                                    \
+      break;                                                                   \
+    fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err);   \
+    abort();                                                                   \
+  } while (0)
+
+#define CL_CHECK_ERR(_expr)                                                    \
+  ({                                                                           \
+    cl_int _err = CL_INVALID_VALUE;                                            \
+    decltype(_expr) _ret = _expr;                                                \
+    if (_err != CL_SUCCESS) {                                                  \
+      fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err); \
+      abort();                                                                 \
+    }                                                                          \
+    _ret;                                                                      \
+  })
+
+void pfn_notify(const char *errinfo, const void *private_info, size_t cb,
+                void *user_data) {
+  fprintf(stderr, "OpenCL Error (via pfn_notify): %s\n", errinfo);
+}
+
+static int read_kernel_file(const char* filename, uint8_t** data, size_t* size) {
+  if (nullptr == filename || nullptr == data || 0 == size)
+    return -1;
+
+  FILE* fp = fopen(filename, "r");
+  if (NULL == fp) {
+    fprintf(stderr, "Failed to load kernel.");
+    return -1;
+  }
+  fseek(fp , 0 , SEEK_END);
+  long fsize = ftell(fp);
+  rewind(fp);
+
+  *data = (uint8_t*)malloc(fsize);
+  *size = fread(*data, 1, fsize, fp);
+  
+  fclose(fp);
+  
+  return 0;
+}
+
+uint8_t *kernel_bin = NULL;
+
+///
+//  Cleanup any created OpenCL resources
+//
+void Cleanup(cl_context context, cl_command_queue commandQueue,
+             cl_program program, cl_kernel kernel, cl_mem memObjects[3]) {
+  for (int i = 0; i < 3; i++) {
+    if (memObjects[i] != 0)
+      clReleaseMemObject(memObjects[i]);
+  }
+  if (commandQueue != 0)
+    clReleaseCommandQueue(commandQueue);
+
+  if (kernel != 0)
+    clReleaseKernel(kernel);
+
+  if (program != 0)
+    clReleaseProgram(program);
+
+  if (context != 0)
+    clReleaseContext(context);
+
+  if (kernel_bin) free(kernel_bin);
+}
+
+int main(int argc, char **argv) {
+  printf("enter demo main\n");
+  
+  cl_platform_id platform_id;
+  cl_device_id device_id;
+  size_t kernel_size;
+  cl_int binary_status = 0;
+  int i;
+
+  // read kernel binary from file  
+  if (0 != read_kernel_file("kernel.pocl", &kernel_bin, &kernel_size))
+    return -1;
+
+  // Getting platform and device information
+  CL_CHECK(clGetPlatformIDs(1, &platform_id, NULL));
+  CL_CHECK(clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, NULL));
+
+  cl_context context;
+  context = CL_CHECK_ERR(clCreateContext(NULL, 1, &device_id, &pfn_notify, NULL, &_err));
+
+  cl_command_queue queue;
+  queue = CL_CHECK_ERR(clCreateCommandQueue(context, device_id, CL_QUEUE_PROFILING_ENABLE, &_err));
+
+  cl_kernel kernel = 0;
+  cl_mem memObjects[2] = {0, 0};
+
+  // Create OpenCL program - first attempt to load cached binary.
+  //  If that is not available, then create the program from source
+  //  and store the binary for future use.
+  std::cout << "Attempting to create program from binary..." << std::endl;
+  cl_program program = CL_CHECK_ERR(clCreateProgramWithBinary(
+    context, 1, &device_id, &kernel_size, &kernel_bin, &binary_status, &_err));
+  if (program == NULL) {
+    std::cerr << "Failed to write program binary" << std::endl;
+    Cleanup(context, queue, program, kernel, memObjects);
+    return 1;
+  } else {
+    std::cout << "Read program from binary." << std::endl;
+  }
+
+  // Build program
+  CL_CHECK(clBuildProgram(program, 1, &device_id, NULL, NULL, NULL));
+
+  printf("attempting to create input buffer\n");
+  fflush(stdout);
+  cl_mem input_buffer;
+  input_buffer = CL_CHECK_ERR(clCreateBuffer(
+      context, CL_MEM_READ_ONLY, sizeof(float) * NUM_DATA, NULL, &_err));
+
+  printf("attempting to create output buffer\n");
+  fflush(stdout);
+  cl_mem output_buffer;
+  output_buffer = CL_CHECK_ERR(clCreateBuffer(
+      context, CL_MEM_WRITE_ONLY, sizeof(float) * NUM_DATA, NULL, &_err));
+
+  memObjects[0] = input_buffer;
+  memObjects[1] = output_buffer;
+
+  float factor = ((float)rand() / (float)(RAND_MAX)) * 100.0;
+
+  printf("attempting to create kernel\n");
+  fflush(stdout);
+  kernel = CL_CHECK_ERR(clCreateKernel(program, "saxpy", &_err));
+  printf("setting up kernel args cl_mem:%lx \n", input_buffer);
+  fflush(stdout);
+  CL_CHECK(clSetKernelArg(kernel, 0, sizeof(input_buffer), &input_buffer));
+  CL_CHECK(clSetKernelArg(kernel, 1, sizeof(output_buffer), &output_buffer));
+  CL_CHECK(clSetKernelArg(kernel, 2, sizeof(factor), &factor));
+
+  printf("attempting to enqueue write buffer\n");
+  fflush(stdout);
+  for (int i = 0; i < NUM_DATA; i++) {
+    float in = ((float)rand() / (float)(RAND_MAX)) * 100.0;
+    CL_CHECK(clEnqueueWriteBuffer(queue, input_buffer, CL_TRUE,
+                                  i * sizeof(float), 4, &in, 0, NULL, NULL));
+  }
+
+  cl_event kernel_completion;
+  size_t global_work_size[] = {NUM_DATA/2,NUM_DATA/2};
+  printf("attempting to enqueue kernel\n");
+  fflush(stdout);
+  CL_CHECK(clEnqueueNDRangeKernel(queue, kernel, 1, NULL, global_work_size,
+                                  NULL, 0, NULL, &kernel_completion));
+  printf("Enqueue'd kerenel\n");
+  fflush(stdout);
+  cl_ulong time_start, time_end;
+  CL_CHECK(clWaitForEvents(1, &kernel_completion));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion,
+                                   CL_PROFILING_COMMAND_START,
+                                   sizeof(time_start), &time_start, NULL));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion, CL_PROFILING_COMMAND_END,
+                                   sizeof(time_end), &time_end, NULL));
+  double elapsed = time_end - time_start;
+  printf("time(ns):%lg\n", elapsed);
+  CL_CHECK(clReleaseEvent(kernel_completion));
+
+  printf("Result:");
+  for (int i = 0; i < NUM_DATA; i++) {
+    float data;
+    CL_CHECK(clEnqueueReadBuffer(queue, output_buffer, CL_TRUE,
+                                 i * sizeof(float), 4, &data, 0, NULL, NULL));
+    // printf(" %f", data);
+  }
+  printf("\n");
+
+  CL_CHECK(clReleaseMemObject(memObjects[0]));
+  CL_CHECK(clReleaseMemObject(memObjects[1]));
+
+  CL_CHECK(clReleaseKernel(kernel));
+  CL_CHECK(clReleaseProgram(program));
+  CL_CHECK(clReleaseContext(context));
+
+  return 0;
+}

benchmarks/new_opencl/sfilter/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = sfilter
+
+SRCS = main.cc
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/sfilter/kernel.cl

@@ -0,0 +1,21 @@
+// m0 m1 m2
+// m3 m4 m5
+// m6 m7 m8
+__kernel void sfilter(__global float *src, __global float *dst, long ldc,
+  float m0, float m1, float m2, float m3, float m4, float m5, float m6, float m7, float m8)
+{
+  long x = get_global_id(0);
+  long y = get_global_id(1);
+  
+  float i0 = src[(x-1)+(y-1)*ldc]*m0;
+  float i1 = src[(x)  +(y-1)*ldc]*m1;
+  float i2 = src[(x+1)+(y-1)*ldc]*m2;
+  float i3 = src[(x-1)+(y)  *ldc]*m3;
+  float i4 = src[(x)  + y  * ldc]*m4;
+  float i5 = src[(x+1)+(y)  *ldc]*m5;
+  float i6 = src[(x-1)+(y+1)*ldc]*m6;
+  float i7 = src[(x)  +(y+1)*ldc]*m7;
+  float i8 = src[(x+1)+(y+1)*ldc]*m8;
+
+  dst[x+y*ldc] = i0 + i1 + i2 + i3 + i4 + i5 + i6 + i7 + i8;
+}

benchmarks/new_opencl/sfilter/main.cc

@@ -0,0 +1,319 @@
+/*
+ *  Simple OpenCL demo program
+ *
+ *  Copyright (C) 2009  Clifford Wolf <clifford@clifford.at>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ *  gcc -o cldemo -std=gnu99 -Wall -I/usr/include/nvidia-current cldemo.c
+ * -lOpenCL
+ *
+ */
+
+#include <CL/cl.h>
+
+#include <errno.h>
+#include <fstream>
+#include <iostream>
+#include <math.h>
+#include <sstream>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#define NUM_DATA 66
+
+#define CL_CHECK(_expr)                                                        \
+  do {                                                                         \
+    cl_int _err = _expr;                                                       \
+    if (_err == CL_SUCCESS)                                                    \
+      break;                                                                   \
+    fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err);   \
+    abort();                                                                   \
+  } while (0)
+
+#define CL_CHECK_ERR(_expr)                                                    \
+  ({                                                                           \
+    cl_int _err = CL_INVALID_VALUE;                                            \
+    decltype(_expr) _ret = _expr;                                                \
+    if (_err != CL_SUCCESS) {                                                  \
+      fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err); \
+      abort();                                                                 \
+    }                                                                          \
+    _ret;                                                                      \
+  })
+
+void pfn_notify(const char *errinfo, const void *private_info, size_t cb,
+                void *user_data) {
+  fprintf(stderr, "OpenCL Error (via pfn_notify): %s\n", errinfo);
+}
+
+static int read_kernel_file(const char* filename, uint8_t** data, size_t* size) {
+  if (nullptr == filename || nullptr == data || 0 == size)
+    return -1;
+
+  FILE* fp = fopen(filename, "r");
+  if (NULL == fp) {
+    fprintf(stderr, "Failed to load kernel.");
+    return -1;
+  }
+  fseek(fp , 0 , SEEK_END);
+  long fsize = ftell(fp);
+  rewind(fp);
+
+  *data = (uint8_t*)malloc(fsize);
+  *size = fread(*data, 1, fsize, fp);
+  
+  fclose(fp);
+  
+  return 0;
+}
+
+uint8_t *kernel_bin = NULL;
+
+// inlcude pocl float to half conversions
+typedef union {
+  int32_t i;
+  float f;
+} FloatConvUnion;
+cl_half poclu_float_to_cl_half(float value) {
+  FloatConvUnion u;
+  u.f = value;
+  cl_half half = (u.i >> 16) & 0x8000; // sign
+  cl_half fraction =
+      (u.i >> 12) & 0x007ff;             // fraction with extra bit for rounding
+  cl_half exponent = (u.i >> 23) & 0xff; // exponent
+
+  if (exponent < 0x0067) // Return signed zero if zero or value is too small for
+                         // denormal half
+    return half;
+
+  if (exponent > 0x008e) { // value was NaN or Inf
+    half |= 0x7c00u;       // Make into inf
+    half |= exponent == 255 &&
+            (u.i & 0x007fffffu); // If value was NaN make this into NaN
+    return half;
+  }
+
+  if (exponent < 0x0071) { // Denormal
+    fraction |= 0x0800u;
+
+    // rounding
+    half |= (fraction >> (0x0072 - exponent)) +
+            ((fraction >> (0x0071 - exponent)) & 1);
+    return half;
+  }
+
+  half |= ((exponent - 0x0070) << 10) | (fraction >> 1);
+  half += fraction & 1; // rounding
+  return half;
+}
+#ifndef INFINITY
+#define INFINITY 1.0 / 0.0
+#endif
+
+#ifndef NAN
+#define NAN 0.0 / 0.0
+#endif
+
+float poclu_cl_half_to_float(cl_half value) {
+  if (value == 0xFC00) {
+    return -INFINITY;
+  }
+  if (value == 0x7C00) {
+    return INFINITY;
+  }
+
+  int sgn = ((value & 0x8000) >> 15);
+  int exp = (value & 0x7C00) >> 10;
+  int mant = value & 0x03FF;
+
+  if (exp == 0x1F && mant != 0) {
+    return NAN;
+  }
+
+  float v = (exp == 0) ? mant : mant | 0x0400; // 1.x if not denormal
+  v /= 0x400;
+  float mul = exp2((float)exp - 15);
+  v *= mul;
+  if (sgn) {
+    v *= -1;
+  }
+  return v;
+}
+
+///
+//  Cleanup any created OpenCL resources
+//
+void Cleanup(cl_context context, cl_command_queue commandQueue,
+             cl_program program, cl_kernel kernel, cl_mem memObjects[3]) {
+  for (int i = 0; i < 3; i++) {
+    if (memObjects[i] != 0)
+      clReleaseMemObject(memObjects[i]);
+  }
+  if (commandQueue != 0)
+    clReleaseCommandQueue(commandQueue);
+
+  if (kernel != 0)
+    clReleaseKernel(kernel);
+
+  if (program != 0)
+    clReleaseProgram(program);
+
+  if (context != 0)
+    clReleaseContext(context);
+
+  if (kernel_bin) free(kernel_bin);
+}
+
+int main(int argc, char **argv) {
+  printf("enter demo main\n");
+
+  cl_platform_id platform_id;
+  cl_device_id device_id;
+  size_t kernel_size;
+  cl_int binary_status = 0;
+  int i;
+
+  // read kernel binary from file  
+  if (0 != read_kernel_file("kernel.pocl", &kernel_bin, &kernel_size))
+    return -1;
+
+  // Getting platform and device information
+  CL_CHECK(clGetPlatformIDs(1, &platform_id, NULL));
+  CL_CHECK(clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, NULL));
+
+  cl_context context;
+  context = CL_CHECK_ERR(clCreateContext(NULL, 1, &device_id, &pfn_notify, NULL, &_err));
+
+  cl_command_queue queue;
+  queue = CL_CHECK_ERR(clCreateCommandQueue(context, device_id, CL_QUEUE_PROFILING_ENABLE, &_err));
+
+  cl_kernel kernel = 0;
+  cl_mem memObjects[2] = {0, 0};
+
+  // Create OpenCL program - first attempt to load cached binary.
+  //  If that is not available, then create the program from source
+  //  and store the binary for future use.
+  std::cout << "Attempting to create program from binary..." << std::endl;
+  cl_program program = CL_CHECK_ERR(clCreateProgramWithBinary(
+    context, 1, &device_id, &kernel_size, &kernel_bin, &binary_status, &_err));
+  if (program == NULL) {
+    std::cerr << "Failed to write program binary" << std::endl;
+    Cleanup(context, queue, program, kernel, memObjects);
+    return 1;
+  } else {
+    std::cout << "Read program from binary." << std::endl;
+  }
+
+  // Build program
+  CL_CHECK(clBuildProgram(program, 1, &device_id, NULL, NULL, NULL));
+
+  printf("attempting to create input buffer\n");
+  fflush(stdout);
+  cl_mem input_buffer;
+  input_buffer = CL_CHECK_ERR(
+      clCreateBuffer(context, CL_MEM_READ_ONLY,
+                     sizeof(float) * NUM_DATA * NUM_DATA, NULL, &_err));
+
+  printf("attempting to create output buffer\n");
+  fflush(stdout);
+  cl_mem output_buffer;
+  output_buffer = CL_CHECK_ERR(
+      clCreateBuffer(context, CL_MEM_WRITE_ONLY,
+                     sizeof(float) * NUM_DATA * NUM_DATA, NULL, &_err));
+
+  memObjects[0] = input_buffer;
+  memObjects[1] = output_buffer;
+
+  long long ldc = NUM_DATA;
+
+  float m0 = 1.0;
+  float m1 = 1.0;
+  float m2 = 1.0;
+  float m3 = 1.0;
+  float m4 = 1.0;
+  float m5 = 1.0;
+  float m6 = 1.0;
+  float m7 = 1.0;
+  float m8 = 1.0;
+
+  printf("attempting to create kernel\n");
+  fflush(stdout);
+  kernel = CL_CHECK_ERR(clCreateKernel(program, "sfilter", &_err));
+  printf("setting up kernel args cl_mem:%lx \n", input_buffer);
+  fflush(stdout);
+  CL_CHECK(clSetKernelArg(kernel, 0, sizeof(input_buffer), &input_buffer));
+  CL_CHECK(clSetKernelArg(kernel, 1, sizeof(output_buffer), &output_buffer));
+  CL_CHECK(clSetKernelArg(kernel, 2, sizeof(ldc), (&ldc)));
+  CL_CHECK(clSetKernelArg(kernel, 3, sizeof(m0), (&m0)));
+  CL_CHECK(clSetKernelArg(kernel, 4, sizeof(m1), (&m1)));
+  CL_CHECK(clSetKernelArg(kernel, 5, sizeof(m2), (&m2)));
+  CL_CHECK(clSetKernelArg(kernel, 6, sizeof(m3), (&m3)));
+  CL_CHECK(clSetKernelArg(kernel, 7, sizeof(m4), (&m4)));
+  CL_CHECK(clSetKernelArg(kernel, 8, sizeof(m5), (&m5)));
+  CL_CHECK(clSetKernelArg(kernel, 9, sizeof(m6), (&m6)));
+  CL_CHECK(clSetKernelArg(kernel, 10, sizeof(m7), (&m7)));
+  CL_CHECK(clSetKernelArg(kernel, 11, sizeof(m8), (&m8)));
+
+  printf("attempting to enqueue write buffer\n");
+  fflush(stdout);
+  for (int i = 0; i < NUM_DATA * NUM_DATA; i++) {
+    float in = ((float)rand() / (float)(RAND_MAX)) * 100.0;
+    CL_CHECK(clEnqueueWriteBuffer(queue, input_buffer, CL_TRUE,
+                                  i * sizeof(float), 4, &in, 0, NULL, NULL));
+  }
+
+  cl_event kernel_completion;
+  size_t global_offset[2] = {1, 1};
+  size_t global_work_size[2] = {NUM_DATA - 2, NUM_DATA - 2}; // avoid the edges
+  const size_t local_work_size[2] = {64, 1};
+  printf("attempting to enqueue kernel\n");
+  fflush(stdout);
+  CL_CHECK(clEnqueueNDRangeKernel(queue, kernel, 2, global_offset,
+                                  global_work_size, local_work_size, 0, NULL,
+                                  &kernel_completion));
+  printf("Enqueue'd kerenel\n");
+  fflush(stdout);
+  cl_ulong time_start, time_end;
+  CL_CHECK(clWaitForEvents(1, &kernel_completion));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion,
+                                   CL_PROFILING_COMMAND_START,
+                                   sizeof(time_start), &time_start, NULL));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion, CL_PROFILING_COMMAND_END,
+                                   sizeof(time_end), &time_end, NULL));
+  double elapsed = time_end - time_start;
+  printf("time(ns):%lg\n", elapsed);
+  CL_CHECK(clReleaseEvent(kernel_completion));
+
+  printf("Result:");
+  for (int i = 0; i < NUM_DATA * NUM_DATA; i++) {
+    float data;
+    CL_CHECK(clEnqueueReadBuffer(queue, output_buffer, CL_TRUE,
+                                 i * sizeof(float), 4, &data, 0, NULL, NULL));
+    // printf(" %f", data);
+  }
+  printf("\n");
+
+  CL_CHECK(clReleaseMemObject(memObjects[0]));
+  CL_CHECK(clReleaseMemObject(memObjects[1]));
+
+  CL_CHECK(clReleaseKernel(kernel));
+  CL_CHECK(clReleaseProgram(program));
+  CL_CHECK(clReleaseContext(context));
+
+  return 0;
+}

benchmarks/new_opencl/sgemm/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = sgemm
+
+SRCS = main.cc
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/sgemm/kernel.cl

@@ -0,0 +1,9 @@
+__kernel void sgemm(__global float *A, __global float *B, __global float *C, int ldc)
+{
+  long i = get_global_id(0);
+  long m = get_global_id(1);
+  long n = get_global_id(2);
+  float a = A[m+n*ldc];
+  float b = B[m*ldc+i];
+  C[i+n*ldc] = C[i+n*ldc] + a * b;
+}

benchmarks/new_opencl/sgemm/main.cc

@@ -0,0 +1,243 @@
+/*
+ *  Simple OpenCL demo program
+ *
+ *  Copyright (C) 2009  Clifford Wolf <clifford@clifford.at>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ *  gcc -o cldemo -std=gnu99 -Wall -I/usr/include/nvidia-current cldemo.c
+ * -lOpenCL
+ *
+ */
+
+#include <CL/cl.h>
+#include <errno.h>
+#include <fstream>
+#include <iostream>
+#include <sstream>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#define NUM_DATA 64
+
+#define CL_CHECK(_expr)                                                        \
+  do {                                                                         \
+    cl_int _err = _expr;                                                       \
+    if (_err == CL_SUCCESS)                                                    \
+      break;                                                                   \
+    fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err);   \
+    abort();                                                                   \
+  } while (0)
+
+#define CL_CHECK_ERR(_expr)                                                    \
+  ({                                                                           \
+    cl_int _err = CL_INVALID_VALUE;                                            \
+    decltype(_expr) _ret = _expr;                                              \
+    if (_err != CL_SUCCESS) {                                                  \
+      fprintf(stderr, "OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err); \
+      abort();                                                                 \
+    }                                                                          \
+    _ret;                                                                      \
+  })
+
+void pfn_notify(const char *errinfo, const void *private_info, size_t cb,
+                void *user_data) {
+  fprintf(stderr, "OpenCL Error (via pfn_notify): %s\n", errinfo);
+}
+
+static int read_kernel_file(const char* filename, uint8_t** data, size_t* size) {
+  if (nullptr == filename || nullptr == data || 0 == size)
+    return -1;
+
+  FILE* fp = fopen(filename, "r");
+  if (NULL == fp) {
+    fprintf(stderr, "Failed to load kernel.");
+    return -1;
+  }
+  fseek(fp , 0 , SEEK_END);
+  long fsize = ftell(fp);
+  rewind(fp);
+
+  *data = (uint8_t*)malloc(fsize);
+  *size = fread(*data, 1, fsize, fp);
+  
+  fclose(fp);
+  
+  return 0;
+}
+
+uint8_t *kernel_bin = NULL;
+
+///
+//  Cleanup any created OpenCL resources
+//
+void Cleanup(cl_context context, cl_command_queue commandQueue,
+             cl_program program, cl_kernel kernel, cl_mem memObjects[3]) {
+  for (int i = 0; i < 3; i++) {
+    if (memObjects[i] != 0)
+      clReleaseMemObject(memObjects[i]);
+  }
+  if (commandQueue != 0)
+    clReleaseCommandQueue(commandQueue);
+
+  if (kernel != 0)
+    clReleaseKernel(kernel);
+
+  if (program != 0)
+    clReleaseProgram(program);
+
+  if (context != 0)
+    clReleaseContext(context);
+
+  if (kernel_bin) free(kernel_bin);
+}
+
+int main(int argc, char **argv) {
+  printf("enter demo main\n");
+  
+  cl_platform_id platform_id;
+  cl_device_id device_id;
+  size_t kernel_size;
+  cl_int binary_status = 0;
+  int i;
+
+  // read kernel binary from file  
+  if (0 != read_kernel_file("kernel.pocl", &kernel_bin, &kernel_size))
+    return -1;
+
+  // Getting platform and device information
+  CL_CHECK(clGetPlatformIDs(1, &platform_id, NULL));
+  CL_CHECK(clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, NULL));
+
+  cl_context context;
+  context = CL_CHECK_ERR(
+      clCreateContext(NULL, 1, &device_id, &pfn_notify, NULL, &_err));
+
+  cl_command_queue queue;
+  queue = CL_CHECK_ERR(clCreateCommandQueue(context, device_id,
+                                            CL_QUEUE_PROFILING_ENABLE, &_err));
+
+  cl_kernel kernel = 0;
+  cl_mem memObjects[3] = {0, 0, 0};
+
+  // Create OpenCL program - first attempt to load cached binary.
+  //  If that is not available, then create the program from source
+  //  and store the binary for future use.
+  std::cout << "Attempting to create program from binary..." << std::endl;
+  // cl_program program = CreateProgramFromBinary(context, device_id,
+  // "kernel.cl.bin");
+  cl_program program = CL_CHECK_ERR(clCreateProgramWithBinary(
+    context, 1, &device_id, &kernel_size, &kernel_bin, &binary_status, &_err));
+  if (program == NULL) {
+    std::cerr << "Failed to write program binary" << std::endl;
+    Cleanup(context, queue, program, kernel, memObjects);
+    return 1;
+  } else {
+    std::cout << "Read program from binary." << std::endl;
+  }
+
+  // Build program
+  CL_CHECK(clBuildProgram(program, 1, &device_id, NULL, NULL, NULL));
+
+  printf("attempting to create input buffer\n");
+  fflush(stdout);
+  cl_mem input_bufferA;
+  input_bufferA = CL_CHECK_ERR(
+      clCreateBuffer(context, CL_MEM_READ_ONLY,
+                     sizeof(float) * NUM_DATA * NUM_DATA, NULL, &_err));
+
+  cl_mem input_bufferB;
+  input_bufferB = CL_CHECK_ERR(
+      clCreateBuffer(context, CL_MEM_READ_ONLY,
+                     sizeof(float) * NUM_DATA * NUM_DATA, NULL, &_err));
+
+  printf("attempting to create output buffer\n");
+  fflush(stdout);
+  cl_mem output_buffer;
+  output_buffer = CL_CHECK_ERR(
+      clCreateBuffer(context, CL_MEM_WRITE_ONLY,
+                     sizeof(float) * NUM_DATA * NUM_DATA, NULL, &_err));
+
+  memObjects[0] = input_bufferA;
+  memObjects[1] = input_bufferB;
+  memObjects[2] = output_buffer;
+
+  int width = NUM_DATA;
+
+  printf("attempting to create kernel\n");
+  fflush(stdout);
+  kernel = CL_CHECK_ERR(clCreateKernel(program, "sgemm", &_err));
+  CL_CHECK(clSetKernelArg(kernel, 0, sizeof(input_bufferA), &input_bufferA));
+  CL_CHECK(clSetKernelArg(kernel, 1, sizeof(input_bufferB), &input_bufferB));
+  CL_CHECK(clSetKernelArg(kernel, 2, sizeof(output_buffer), &output_buffer));
+  CL_CHECK(clSetKernelArg(kernel, 3, sizeof(width), &width));
+  
+  printf("attempting to enqueue write buffer\n");
+  fflush(stdout);
+  for (int i = 0; i < NUM_DATA * NUM_DATA; i++) {
+
+    float in = ((float)rand() / (float)(RAND_MAX)) * 100.0;
+    CL_CHECK(clEnqueueWriteBuffer(queue, input_bufferA, CL_TRUE,
+                                  i * sizeof(float), 4, &in, 0, NULL, NULL));
+    in = ((float)rand() / (float)(RAND_MAX)) * 100.0;
+    CL_CHECK(clEnqueueWriteBuffer(queue, input_bufferB, CL_TRUE,
+                                  i * sizeof(float), 4, &in, 0, NULL, NULL));
+  }
+
+  printf("Done enqueueing\n");
+
+  cl_event kernel_completion;
+  const size_t local_work_size[3] = {1, 1, 1};
+  //                             a_offset
+  size_t global_work_size[3] = {NUM_DATA, NUM_DATA, NUM_DATA};
+  printf("attempting to enqueue kernel\n");
+  fflush(stdout);
+  CL_CHECK(clEnqueueNDRangeKernel(queue, kernel, 3, NULL, global_work_size,
+                                  local_work_size, 0, NULL,
+                                  &kernel_completion));
+  printf("Enqueue'd kerenel\n");
+  fflush(stdout);
+  cl_ulong time_start, time_end;
+  CL_CHECK(clWaitForEvents(1, &kernel_completion));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion,
+                                   CL_PROFILING_COMMAND_START,
+                                   sizeof(time_start), &time_start, NULL));
+  CL_CHECK(clGetEventProfilingInfo(kernel_completion, CL_PROFILING_COMMAND_END,
+                                   sizeof(time_end), &time_end, NULL));
+  double elapsed = time_end - time_start;
+  printf("time(ns):%lg\n", elapsed);
+  CL_CHECK(clReleaseEvent(kernel_completion));
+
+  printf("Result:");
+  for (int i = 0; i < NUM_DATA * NUM_DATA; i++) {
+    float data;
+    CL_CHECK(clEnqueueReadBuffer(queue, output_buffer, CL_TRUE,
+                                 i * sizeof(float), 4, &data, 0, NULL, NULL));
+    // printf(" %f", data);
+  }
+  printf("\n");
+
+  CL_CHECK(clReleaseMemObject(memObjects[0]));
+  CL_CHECK(clReleaseMemObject(memObjects[1]));
+  CL_CHECK(clReleaseMemObject(memObjects[2]));
+
+  CL_CHECK(clReleaseKernel(kernel));
+  CL_CHECK(clReleaseProgram(program));
+  CL_CHECK(clReleaseContext(context));
+
+  return 0;
+}

benchmarks/new_opencl/vecadd/Makefile

@@ -0,0 +1,44 @@
+RISCV_TOOL_PATH ?= $(wildcard ~/dev/riscv-gnu-toolchain/drops)
+POCLCC_PATH ?= $(wildcard ~/dev/pocl/drops_vortex_cc)
+POCLRT_PATH ?= $(wildcard ..)
+DRIVER_PATH ?= $(wildcard ../../../driver/sw)
+
+CXXFLAGS += -std=c++11 -O0 -g -fpermissive -Wall -Wextra -pedantic -Wfatal-errors
+
+CXXFLAGS += -I$(POCLRT_PATH)/include
+
+LDFLAGS += -L$(POCLRT_PATH)/lib -L$(DRIVER_PATH)/simx -lOpenCL -lvortex
+
+PROJECT = vecadd
+
+SRCS = main.cc
+
+all: $(PROJECT)
+
+kernel.pocl: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCLCC_PATH)/lib:$(DRIVER_PATH)/simx $(POCLCC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
+ 
+$(PROJECT): $(SRCS)
+	$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
+
+run-fpga: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-ase: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/opae/ase:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-simx: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+run-rtlsim: $(PROJECT) kernel.pocl   
+	LD_LIBRARY_PATH=$(POCLRT_PATH)/lib:$(DRIVER_PATH)/rtlsim:$(LD_LIBRARY_PATH) ./$(PROJECT)
+
+.depend: $(SRCS)
+	$(CXX) $(CXXFLAGS) -MM $^ > .depend;
+
+clean:
+	rm -rf $(PROJECT) *.o *.dump .depend 
+
+ifneq ($(MAKECMDGOALS),clean)
+    -include .depend
+endif

benchmarks/new_opencl/vecadd/kernel.cl

@@ -0,0 +1,8 @@
+kernel void
+vecadd (__global const int *a,
+	__global const int *b,
+	__global int *c)
+{
+  int gid = get_global_id(0);
+  c[gid] = a[gid] + b[gid];
+}

benchmarks/new_opencl/vecadd/main.cc

@@ -0,0 +1,187 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <assert.h>
+#include <CL/opencl.h>
+#include <string.h>
+
+#define SIZE 4
+#define NUM_WORK_GROUPS 2
+#define KERNEL_NAME "vecadd"
+
+#define CL_CHECK(_expr)                                                \
+   do {                                                                \
+     cl_int _err = _expr;                                              \
+     if (_err == CL_SUCCESS)                                           \
+       break;                                                          \
+     printf("OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err);   \
+	 cleanup();			                                                     \
+     exit(-1);                                                         \
+   } while (0)
+
+#define CL_CHECK2(_expr)                                               \
+   ({                                                                  \
+     cl_int _err = CL_INVALID_VALUE;                                   \
+     decltype(_expr) _ret = _expr;                                     \
+     if (_err != CL_SUCCESS) {                                         \
+       printf("OpenCL Error: '%s' returned %d!\n", #_expr, (int)_err); \
+	   cleanup();			                                                   \
+       exit(-1);                                                       \
+     }                                                                 \
+     _ret;                                                             \
+   })
+
+int exitcode = 0;
+cl_context context = NULL;
+cl_command_queue commandQueue = NULL;
+cl_program program = NULL;
+cl_kernel kernel = NULL;
+cl_mem a_memobj = NULL;
+cl_mem b_memobj = NULL;
+cl_mem c_memobj = NULL;  
+cl_int *A = NULL;
+cl_int *B = NULL;
+cl_int *C = NULL;
+uint8_t *kernel_bin = NULL;
+
+static int read_kernel_file(const char* filename, uint8_t** data, size_t* size) {
+  if (nullptr == filename || nullptr == data || 0 == size)
+    return -1;
+
+  FILE* fp = fopen(filename, "r");
+  if (NULL == fp) {
+    fprintf(stderr, "Failed to load kernel.");
+    return -1;
+  }
+  fseek(fp , 0 , SEEK_END);
+  long fsize = ftell(fp);
+  rewind(fp);
+
+  *data = (uint8_t*)malloc(fsize);
+  *size = fread(*data, 1, fsize, fp);
+  
+  fclose(fp);
+  
+  return 0;
+}
+
+static void cleanup() {
+  if (commandQueue) clReleaseCommandQueue(commandQueue);
+  if (kernel) clReleaseKernel(kernel);
+  if (program) clReleaseProgram(program);
+  if (a_memobj) clReleaseMemObject(a_memobj);
+  if (b_memobj) clReleaseMemObject(b_memobj);
+  if (c_memobj) clReleaseMemObject(c_memobj);  
+  if (context) clReleaseContext(context);
+  if (kernel_bin) free(kernel_bin);
+  if (A) free(A);
+  if (B) free(B);
+  if (C) free(C);
+}
+
+static int find_device(char* name, cl_platform_id platform_id, cl_device_id *device_id) {
+  cl_device_id device_ids[64];
+  cl_uint num_devices = 0;
+
+  CL_CHECK(clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_ALL, 64, device_ids, &num_devices));
+
+  for (int i=0; i<num_devices; i++) 	{
+		char buffer[1024];
+		cl_uint buf_uint;
+		cl_ulong buf_ulong;
+
+		CL_CHECK(clGetDeviceInfo(device_ids[i], CL_DEVICE_NAME, sizeof(buffer), buffer, NULL));
+		
+    if (0 == strncmp(buffer, name, strlen(name))) {
+      *device_id = device_ids[i];
+      return 0;
+    }
+	}
+
+  return 1;
+}
+
+int main (int argc, char **argv) {
+  printf("enter demo main\n");
+
+  cl_platform_id platform_id;
+  cl_device_id device_id;
+  size_t kernel_size;
+  cl_int binary_status = 0;
+  int i;
+
+  // read kernel binary from file  
+  if (0 != read_kernel_file("kernel.pocl", &kernel_bin, &kernel_size))
+    return -1;
+  
+  // Getting platform and device information
+  CL_CHECK(clGetPlatformIDs(1, &platform_id, NULL));
+  CL_CHECK(clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, NULL));
+
+  // Creating context.
+  context = CL_CHECK2(clCreateContext(NULL, 1, &device_id, NULL, NULL,  &_err));
+
+  // Memory buffers for each array
+  a_memobj = CL_CHECK2(clCreateBuffer(context, CL_MEM_READ_ONLY, SIZE * sizeof(cl_int), NULL, &_err));
+  b_memobj = CL_CHECK2(clCreateBuffer(context, CL_MEM_READ_ONLY, SIZE * sizeof(cl_int), NULL, &_err));
+  c_memobj = CL_CHECK2(clCreateBuffer(context, CL_MEM_WRITE_ONLY, SIZE * sizeof(cl_int), NULL, &_err));
+
+  // Allocate memories for input arrays and output arrays.  
+  A = (cl_int*)malloc(sizeof(cl_int)*SIZE);
+  B = (cl_int*)malloc(sizeof(cl_int)*SIZE);
+  C = (cl_int*)malloc(sizeof(cl_int)*SIZE);	
+	
+  // Initialize values for array members.  
+  for (i=0; i<SIZE; ++i) {
+    A[i] = i*2+0;
+    B[i] = i*2+1;
+  }
+
+  // Create program from kernel source
+  program = CL_CHECK2(clCreateProgramWithBinary(
+    context, 1, &device_id, &kernel_size, &kernel_bin, &binary_status, &_err));
+
+  // Build program
+  CL_CHECK(clBuildProgram(program, 1, &device_id, NULL, NULL, NULL));
+  
+  // Create kernel
+  kernel = CL_CHECK2(clCreateKernel(program, KERNEL_NAME, &_err));
+
+  // Set arguments for kernel
+  CL_CHECK(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&a_memobj));	
+  CL_CHECK(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&b_memobj));	
+  CL_CHECK(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&c_memobj));
+
+  // Creating command queue
+  commandQueue = CL_CHECK2(clCreateCommandQueue(context, device_id, 0, &_err));
+
+	// Copy lists to memory buffers
+  CL_CHECK(clEnqueueWriteBuffer(commandQueue, a_memobj, CL_TRUE, 0, SIZE * sizeof(float), A, 0, NULL, NULL));
+  CL_CHECK(clEnqueueWriteBuffer(commandQueue, b_memobj, CL_TRUE, 0, SIZE * sizeof(float), B, 0, NULL, NULL));
+
+  // Execute the kernel
+  size_t globalItemSize = SIZE;
+  size_t localItemSize = SIZE/NUM_WORK_GROUPS;
+  CL_CHECK(clEnqueueNDRangeKernel(commandQueue, kernel, 1, NULL, &globalItemSize, &localItemSize, 0, NULL, NULL));
+  CL_CHECK(clFinish(commandQueue));
+
+  // Read from device back to host.
+  CL_CHECK(clEnqueueReadBuffer(commandQueue, c_memobj, CL_TRUE, 0, SIZE * sizeof(float), C, 0, NULL, NULL));
+
+  // Test if correct answer
+  int exitcode = 0;
+  for (i=0; i<SIZE; ++i) {
+    if (C[i] != (A[i] + B[i])) {
+      printf("Failed!\n");
+      exitcode = 1;
+      break;
+    }
+  }
+  if (i == SIZE) {
+    printf("Ok!\n");
+  }
+
+  // Clean up		
+  cleanup();  
+
+  return exitcode;
+}

benchmarks/opencl/BlackScholes/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT=BlackScholes
 

benchmarks/opencl/DotProduct/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT=DotProduct
 

benchmarks/opencl/VectorHypot/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT=VectorHypot
 

benchmarks/opencl/bfs/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH)
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = bfs
 

benchmarks/opencl/convolution/Makefile

@@ -0,0 +1,68 @@
+RISCV_TOOL_PATH ?= $(wildcard ../../../../riscv-gnu-toolchain/drops)
+POCL_CC_PATH ?= $(wildcard ../../../../pocl/drops_riscv_cc)
+POCL_INC_PATH ?= $(wildcard ../include)
+POCL_LIB_PATH ?= $(wildcard ../lib)
+VX_RT_PATH ?= $(wildcard ../../../runtime)
+VX_SIMX_PATH ?= $(wildcard ../../../simX/obj_dir)
+
+CC  = $(RISCV_TOOL_PATH)/bin/riscv32-unknown-elf-gcc
+CXX = $(RISCV_TOOL_PATH)/bin/riscv32-unknown-elf-g++
+DMP = $(RISCV_TOOL_PATH)/bin/riscv32-unknown-elf-objdump
+HEX = $(RISCV_TOOL_PATH)/bin/riscv32-unknown-elf-objcopy
+GDB = $(RISCV_TOOL_PATH)/bin/riscv32-unknown-elf-gdb
+
+VX_SRCS =  $(VX_RT_PATH)/newlib/newlib.c
+VX_SRCS += $(VX_RT_PATH)/startup/vx_start.s
+VX_SRCS += $(VX_RT_PATH)/intrinsics/vx_intrinsics.s
+VX_SRCS += $(VX_RT_PATH)/io/vx_io.s $(VX_RT_PATH)/io/vx_io.c
+VX_SRCS += $(VX_RT_PATH)/fileio/fileio.s
+VX_SRCS += $(VX_RT_PATH)/tests/tests.c
+VX_SRCS += $(VX_RT_PATH)/vx_api/vx_api.c
+VX_SRCS += $(VX_STR) $(VX_FIO) $(VX_NEWLIB) $(VX_INT) $(VX_IO) $(VX_API) $(VX_TEST)
+
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/mains/vortex_link.ld
+
+CXXFLAGS =  -g -O0 -march=rv32im -mabi=ilp32 
+CXXFLAGS += -ffreestanding # program may not begin at main()
+CXXFLAGS += -Wl,--gc-sections # enable garbage collection of unused input sections
+CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
+CXXFLAGS += -I$(POCL_INC_PATH)
+
+VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
+
+PROJECT = convolution
+
+SRCS = main.cpp utils.cpp
+
+all: $(PROJECT).dump $(PROJECT).hex
+
+lib$(PROJECT).a: kernel.cl
+	POCL_DEBUG=all POCL_DEBUG_LLVM_PASSES=1 LD_LIBRARY_PATH=$(RISCV_TOOL_PATH)/lib:$(POCL_CC_PATH)/lib $(POCL_CC_PATH)/bin/poclcc -o lib$(PROJECT).a kernel.cl
+
+$(PROJECT).elf: $(SRCS) lib$(PROJECT).a
+	$(CXX) $(CXXFLAGS) $(VX_CFLAGS) $(VX_SRCS) $(SRCS) $(VX_LIBS) -o $(PROJECT).elf
+
+$(PROJECT).qemu: $(SRCS) lib$(PROJECT).a
+	$(CXX) $(CXXFLAGS) $(SRCS) $(QEMU_LIBS) -o $(PROJECT).qemu
+
+$(PROJECT).hex: $(PROJECT).elf
+	$(HEX) -O ihex $(PROJECT).elf $(PROJECT).hex
+
+$(PROJECT).dump: $(PROJECT).elf
+	$(DMP) -D $(PROJECT).elf > $(PROJECT).dump
+
+run: $(PROJECT).hex
+	POCL_DEBUG=all $(VX_SIMX_PATH)/Vcache_simX -E -a rv32i --core $(PROJECT).hex -s -b 1> emulator.debug
+
+qemu: $(PROJECT).qemu
+	POCL_DEBUG=all $(RISCV_TOOL_PATH)/bin/qemu-riscv32 -d in_asm -D debug.log $(PROJECT).qemu
+
+gdb-s: $(PROJECT).qemu
+	POCL_DEBUG=all $(RISCV_TOOL_PATH)/bin/qemu-riscv32 -g 1234 -d in_asm -D debug.log $(PROJECT).qemu
+
+gdb-c: $(PROJECT).qemu
+	$(GDB) $(PROJECT).qemu
+
+clean:
+	rm -rf *.o *.elf *.dump *.hex *.qemu *.log *.debug

benchmarks/opencl/convolution/kernel.cl

@@ -0,0 +1,54 @@
+__kernel
+void convolution(
+   __read_only  image2d_t  sourceImage,
+   __write_only image2d_t  outputImage, 
+   int rows,
+   int cols,
+   __constant float* filter, 
+   int filterWidth,
+   sampler_t sampler) 
+{
+   // Store each work-item’s unique row and column
+   int column = get_global_id(0);
+   int row    = get_global_id(1);
+
+   // Half the width of the filter is needed for indexing 
+   // memory later
+   int halfWidth = (int)(filterWidth/2);
+
+   // All accesses to images return data as four-element vector 
+   // (i.e., float4), although only the 'x' component will contain 
+   // meaningful data in this code
+   float4 sum = {0.0f, 0.0f, 0.0f, 0.0f};
+    
+   // Iterator for the filter
+   int filterIdx = 0;
+    
+   // Each work-item iterates around its local area based on the 
+   // size of the filter
+   int2 coords;  // Coordinates for accessing the image
+   // Iterate the filter rows
+   for(int i = -halfWidth; i <= halfWidth; i++) {
+      coords.y = row + i;
+
+      // Iterate over the filter columns
+      for(int j = -halfWidth; j <= halfWidth; j++) {
+         coords.x = column + j;
+
+         float4 pixel;
+         // Read a pixel from the image.  A single channel image 
+         // stores the pixel in the 'x' coordinate of the returned
+         // vector.
+         pixel = read_imagef(sourceImage, sampler, coords);
+         sum.x += pixel.x * filter[filterIdx++];
+      }
+   }
+    
+   // Copy the data to the output image if the
+   // work-item is in bounds
+   if(row < rows && column < cols) {
+      coords.x = column;
+      coords.y = row;
+      write_imagef(outputImage, coords, sum);
+   } 
+}

benchmarks/opencl/convolution/main.cpp

@@ -0,0 +1,261 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <CL/cl.h>
+
+#include "utils.h"
+
+// This function takes a positive integer and rounds it up to
+// the nearest multiple of another provided integer
+unsigned int roundUp(unsigned int value, unsigned int multiple) {
+
+  // Determine how far past the nearest multiple the value is
+  unsigned int remainder = value % multiple;
+
+  // Add the difference to make the value a multiple
+  if(remainder != 0) {
+          value += (multiple-remainder);
+  }
+
+  return value;
+}
+
+// This function reads in a text file and stores it as a char pointer
+char* readSource(char* kernelPath) {
+
+   cl_int status;
+   FILE *fp;
+   char *source;
+   long int size;
+
+   printf("Program file is: %s\n", kernelPath);
+
+   fp = fopen(kernelPath, "rb");
+   if(!fp) {
+      printf("Could not open kernel file\n");
+      exit(-1);
+   }
+   status = fseek(fp, 0, SEEK_END);
+   if(status != 0) {
+      printf("Error seeking to end of file\n");
+      exit(-1);
+   }
+   size = ftell(fp);
+   if(size < 0) {
+      printf("Error getting file position\n");
+      exit(-1);
+   }
+
+   rewind(fp);
+
+   source = (char *)malloc(size + 1);
+
+   int i;
+   for (i = 0; i < size+1; i++) {
+      source[i]='\0';
+   }
+
+   if(source == NULL) {
+      printf("Error allocating space for the kernel source\n");
+      exit(-1);
+   }
+
+   fread(source, 1, size, fp);
+   source[size] = '\0';
+
+   return source;
+}
+
+void chk(cl_int status, const char* cmd) {
+
+   if(status != CL_SUCCESS) {
+      printf("%s failed (%d)\n", cmd, status);
+      exit(-1);
+   }
+}
+
+int main() {
+
+   int i, j, k, l;
+
+   // Rows and columns in the input image
+   int imageHeight;
+   int imageWidth;
+
+   const char* inputFile = "input.bmp";
+   const char* outputFile = "output.bmp";
+
+   // Homegrown function to read a BMP from file
+   float* inputImage = readImage(inputFile, &imageWidth,
+      &imageHeight);
+
+   // Size of the input and output images on the host
+   int dataSize = imageHeight*imageWidth*sizeof(float);
+
+   // Output image on the host
+   float* outputImage = NULL;
+   outputImage = (float*)malloc(dataSize);
+   float* refImage = NULL;
+   refImage = (float*)malloc(dataSize);
+
+   // 45 degree motion blur
+   float filter[49] =
+      {0,      0,      0,      0,      0,      0,      0,
+       0,      0,      0,      0,      0,      0,      0,
+       0,      0,     -1,      0,      1,      0,      0,
+       0,      0,     -2,      0,      2,      0,      0,
+       0,      0,     -1,      0,      1,      0,      0,
+       0,      0,      0,      0,      0,      0,      0,
+       0,      0,      0,      0,      0,      0,      0};
+
+   // The convolution filter is 7x7
+   int filterWidth = 7;  
+   int filterSize  = filterWidth*filterWidth;  // Assume a square kernel
+
+   // Set up the OpenCL environment
+   cl_int status;
+
+   // Discovery platform
+   cl_platform_id platform;
+   status = clGetPlatformIDs(1, &platform, NULL);
+   chk(status, "clGetPlatformIDs");
+
+   // Discover device
+   cl_device_id device;
+   clGetDeviceIDs(platform, CL_DEVICE_TYPE_ALL, 1, &device, NULL);
+   chk(status, "clGetDeviceIDs");
+
+   // Create context
+   cl_context_properties props[3] = {CL_CONTEXT_PLATFORM,
+       (cl_context_properties)(platform), 0};
+   cl_context context;
+   context = clCreateContext(props, 1, &device, NULL, NULL, &status);
+   chk(status, "clCreateContext");
+
+   // Create command queue
+   cl_command_queue queue;
+   queue = clCreateCommandQueue(context, device, 0, &status);
+   chk(status, "clCreateCommandQueue");
+
+   // The image format describes how the data will be stored in memory
+   cl_image_format format;
+   format.image_channel_order     = CL_R;     // single channel
+   format.image_channel_data_type = CL_FLOAT; // float data type
+
+   // Create space for the source image on the device
+   cl_mem d_inputImage = clCreateImage2D(context, 0, &format, imageWidth, 
+      imageHeight, 0, NULL, &status);
+   chk(status, "clCreateImage2D");
+
+   // Create space for the output image on the device
+   cl_mem d_outputImage = clCreateImage2D(context, 0, &format, imageWidth, 
+      imageHeight, 0, NULL, &status);
+   chk(status, "clCreateImage2D");
+
+   // Create space for the 7x7 filter on the device
+   cl_mem d_filter = clCreateBuffer(context, 0, filterSize*sizeof(float), 
+      NULL, &status);
+   chk(status, "clCreateBuffer");
+
+   // Copy the source image to the device
+   size_t origin[3] = {0, 0, 0};  // Offset within the image to copy from
+   size_t region[3] = {imageWidth, imageHeight, 1}; // Elements to per dimension
+   status = clEnqueueWriteImage(queue, d_inputImage, CL_FALSE, origin, region, 
+      0, 0, inputImage, 0, NULL, NULL);
+   chk(status, "clEnqueueWriteImage");
+    
+   // Copy the 7x7 filter to the device
+   status = clEnqueueWriteBuffer(queue, d_filter, CL_FALSE, 0, 
+      filterSize*sizeof(float), filter, 0, NULL, NULL);
+   chk(status, "clEnqueueWriteBuffer");
+
+   // Create the image sampler
+   cl_sampler sampler = clCreateSampler(context, CL_FALSE, 
+      CL_ADDRESS_CLAMP_TO_EDGE, CL_FILTER_NEAREST, &status);
+   chk(status, "clCreateSampler");
+
+   const char* source = readSource("kernel.cl");
+
+   // Create a program object with source and build it
+   cl_program program;
+   program = clCreateProgramWithSource(context, 1, &source, NULL, NULL);
+   chk(status, "clCreateProgramWithSource");
+   status = clBuildProgram(program, 1, &device, NULL, NULL, NULL);
+   chk(status, "clBuildProgram");
+ 
+   // Create the kernel object
+   cl_kernel kernel;
+   kernel = clCreateKernel(program, "convolution", &status);
+   chk(status, "clCreateKernel");
+
+   // Set the kernel arguments
+   status  = clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_inputImage);
+   status |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_outputImage);
+   status |= clSetKernelArg(kernel, 2, sizeof(int), &imageHeight);
+   status |= clSetKernelArg(kernel, 3, sizeof(int), &imageWidth);
+   status |= clSetKernelArg(kernel, 4, sizeof(cl_mem), &d_filter);
+   status |= clSetKernelArg(kernel, 5, sizeof(int), &filterWidth);
+   status |= clSetKernelArg(kernel, 6, sizeof(cl_sampler), &sampler);
+   chk(status, "clSetKernelArg");
+
+   // Set the work item dimensions
+   size_t globalSize[2] = {imageWidth, imageHeight};
+   status = clEnqueueNDRangeKernel(queue, kernel, 2, NULL, globalSize, NULL, 0,
+      NULL, NULL);
+   chk(status, "clEnqueueNDRange");
+
+   // Read the image back to the host
+   status = clEnqueueReadImage(queue, d_outputImage, CL_TRUE, origin, 
+      region, 0, 0, outputImage, 0, NULL, NULL); 
+   chk(status, "clEnqueueReadImage");
+
+   // Write the output image to file
+   storeImage(outputImage, outputFile, imageHeight, imageWidth, inputFile);
+
+   // Compute the reference image
+   for(i = 0; i < imageHeight; i++) {
+      for(j = 0; j < imageWidth; j++) {
+         refImage[i*imageWidth+j] = 0;
+      }
+   }
+
+   // Iterate over the rows of the source image
+   int halfFilterWidth = filterWidth/2;
+   float sum;
+   for(i = 0; i < imageHeight; i++) {
+      // Iterate over the columns of the source image
+      for(j = 0; j < imageWidth; j++) {
+         sum = 0; // Reset sum for new source pixel
+         // Apply the filter to the neighborhood
+         for(k = - halfFilterWidth; k <= halfFilterWidth; k++) {
+            for(l = - halfFilterWidth; l <= halfFilterWidth; l++) {
+               if(i+k >= 0 && i+k < imageHeight && 
+                  j+l >= 0 && j+l < imageWidth) {
+                  sum += inputImage[(i+k)*imageWidth + j+l] * 
+                         filter[(k+halfFilterWidth)*filterWidth + 
+                            l+halfFilterWidth];
+               }
+            }
+         }
+         refImage[i*imageWidth+j] = sum;
+      }
+   }
+
+   int failed = 0;
+   for(i = 0; i < imageHeight; i++) {
+      for(j = 0; j < imageWidth; j++) {
+         if(abs(outputImage[i*imageWidth+j]-refImage[i*imageWidth+j]) > 0.01) {
+            printf("Results are INCORRECT\n");
+            printf("Pixel mismatch at <%d,%d> (%f vs. %f)\n", i, j,
+               outputImage[i*imageWidth+j], refImage[i*imageWidth+j]);
+            failed = 1;
+         }
+         if(failed) break;
+      }
+      if(failed) break;
+   }
+   if(!failed) {
+      printf("Results are correct\n");
+   }
+             
+   return 0;
+}

benchmarks/opencl/convolution/utils.cpp

@@ -0,0 +1,180 @@
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "utils.h"
+
+void storeImage(float *imageOut, 
+                const char *filename, 
+                int rows, 
+                int cols, 
+                const char* refFilename) {
+
+   FILE *ifp, *ofp;
+   unsigned char tmp;
+   int offset;
+   unsigned char *buffer;
+   int i, j;
+
+   int bytes;
+
+   int height, width;
+
+   ifp = fopen(refFilename, "rb");
+   if(ifp == NULL) {
+      perror(filename);
+      exit(-1);
+   }
+
+   fseek(ifp, 10, SEEK_SET);
+   fread(&offset, 4, 1, ifp);
+
+   fseek(ifp, 18, SEEK_SET);
+   fread(&width, 4, 1, ifp);
+   fread(&height, 4, 1, ifp);
+
+   fseek(ifp, 0, SEEK_SET);
+
+   buffer = (unsigned char *)malloc(offset);
+   if(buffer == NULL) {
+      perror("malloc");
+      exit(-1);
+   }
+
+   fread(buffer, 1, offset, ifp);
+
+   printf("Writing output image to %s\n", filename);
+   ofp = fopen(filename, "wb");
+   if(ofp == NULL) {
+      perror("opening output file");
+      exit(-1);
+   }
+   bytes = fwrite(buffer, 1, offset, ofp);
+   if(bytes != offset) {
+      printf("error writing header!\n");
+      exit(-1);
+   }
+
+   // NOTE bmp formats store data in reverse raster order (see comment in
+   // readImage function), so we need to flip it upside down here.  
+   int mod = width % 4;
+   if(mod != 0) {
+      mod = 4 - mod;
+   }
+   //   printf("mod = %d\n", mod);
+   for(i = height-1; i >= 0; i--) {
+      for(j = 0; j < width; j++) {
+         tmp = (unsigned char)imageOut[i*cols+j];
+         fwrite(&tmp, sizeof(char), 1, ofp);
+      }
+      // In bmp format, rows must be a multiple of 4-bytes.  
+      // So if we're not at a multiple of 4, add junk padding.
+      for(j = 0; j < mod; j++) {
+         fwrite(&tmp, sizeof(char), 1, ofp);
+      }
+   } 
+
+   fclose(ofp);
+   fclose(ifp);
+
+   free(buffer);
+}
+
+/*
+ * Read bmp image and convert to byte array. Also output the width and height
+ */
+float* readImage(const char *filename, int* widthOut, int* heightOut) {
+
+   uchar* imageData;
+
+   int height, width;
+   uchar tmp;
+   int offset;
+   int i, j;
+
+   printf("Reading input image from %s\n", filename);
+   FILE *fp = fopen(filename, "rb");
+   if(fp == NULL) {
+       perror(filename);
+       exit(-1);
+   }
+
+   fseek(fp, 10, SEEK_SET);
+   fread(&offset, 4, 1, fp);
+
+   fseek(fp, 18, SEEK_SET);
+   fread(&width, 4, 1, fp);
+   fread(&height, 4, 1, fp);
+
+   printf("width = %d\n", width);
+   printf("height = %d\n", height);
+
+   *widthOut = width;
+   *heightOut = height;    
+
+   imageData = (uchar*)malloc(width*height);
+   if(imageData == NULL) {
+       perror("malloc");
+       exit(-1);
+   }
+
+   fseek(fp, offset, SEEK_SET);
+   fflush(NULL);
+
+   int mod = width % 4;
+   if(mod != 0) {
+       mod = 4 - mod;
+   }
+
+   // NOTE bitmaps are stored in upside-down raster order.  So we begin
+   // reading from the bottom left pixel, then going from left-to-right, 
+   // read from the bottom to the top of the image.  For image analysis, 
+   // we want the image to be right-side up, so we'll modify it here.
+
+   // First we read the image in upside-down
+
+   // Read in the actual image
+   for(i = 0; i < height; i++) {
+
+      // add actual data to the image
+      for(j = 0; j < width; j++) {
+         fread(&tmp, sizeof(char), 1, fp);
+         imageData[i*width + j] = tmp;
+      }
+      // For the bmp format, each row has to be a multiple of 4, 
+      // so I need to read in the junk data and throw it away
+      for(j = 0; j < mod; j++) {
+         fread(&tmp, sizeof(char), 1, fp);
+      }
+   }
+
+   // Then we flip it over
+   int flipRow;
+   for(i = 0; i < height/2; i++) {
+      flipRow = height - (i+1);
+      for(j = 0; j < width; j++) {
+         tmp = imageData[i*width+j];
+         imageData[i*width+j] = imageData[flipRow*width+j];
+         imageData[flipRow*width+j] = tmp;
+      }
+   }
+
+   fclose(fp);
+
+   // Input image on the host
+   float* floatImage = NULL;
+   floatImage = (float*)malloc(sizeof(float)*width*height);
+   if(floatImage == NULL) {
+      perror("malloc");
+      exit(-1);
+   }
+
+   // Convert the BMP image to float (not required)
+   for(i = 0; i < height; i++) {
+      for(j = 0; j < width; j++) {
+         floatImage[i*width+j] = (float)imageData[i*width+j];
+      }
+   }
+
+   free(imageData);
+   return floatImage;
+}

benchmarks/opencl/convolution/utils.h

@@ -0,0 +1,11 @@
+#ifndef __UTILS__
+#define __UTILS__
+
+typedef unsigned char uchar;
+
+float* readImage(const char *filename, int* widthOut, int* heightOut);
+
+void storeImage(float *imageOut, const char *filename, int rows, int cols, 
+                const char* refFilename);
+
+#endif

benchmarks/opencl/cutcp/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = cutcp
 

benchmarks/opencl/guassian/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH)
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = gaussian
 

benchmarks/opencl/kmeans/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH)
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = kmeans
 SRCS = main.cc read_input.c rmse.c cluster.c kmeans_clustering.c

benchmarks/opencl/lbm/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = lbm
  

benchmarks/opencl/mri-q/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH) -I.
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = mri-q
 

benchmarks/opencl/nearn/Makefile

@@ -29,7 +29,7 @@ CXXFLAGS += -fno-rtti -fno-non-call-exceptions # disable RTTI and exceptions
 CXXFLAGS += -I$(POCL_INC_PATH)
 
 VX_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
-QEMU_LIBS = -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/qemu/libOpenCL.a
+QEMU_LIBS = $(VX_RT_PATH)/qemu/vx_api.c -Wl,--whole-archive lib$(PROJECT).a -Wl,--no-whole-archive $(POCL_LIB_PATH)/libOpenCL.a
 
 PROJECT = nearn