fpga_synthesis merge

.gitignore

@@ -1,6 +0,0 @@
-./rtl/obj_dir/
-./rtl/obj_dir/*.vcd
-./rtl/.*
-./rtl/modelsim/*.vcd
-*.vcd
-.*

Makefile

@@ -0,0 +1,6 @@
+
+all:
+	$(MAKE) -C hw
+	$(MAKE) -C driver
+	$(MAKE) -C runtime
+	$(MAKE) -C simX

README.md

@@ -6,13 +6,15 @@ Vortex currently supported RISC-V RV32I ISA
 
 /docs contains documentation.
 
-/runtime contains the runtime software support for Vortex.
+/hw constains hardware sources.
 
-/emulator contains a software emulator for Vortex.
+/driver contains the driver software.
+
+/runtime contains the kernel runtime software.
 
 /SimX contains a cycle-approximate simulator for Vortex.
 
-/rtl constains Vortex processor hardware description.
+/evaluation contains the synthesis/runtime reports.
 
 Basic Instructions to run OpenCL Benchmarks on Vortex
 -----------------------------------------------------
@@ -24,6 +26,8 @@ Install development tools
 
 Install gnu-riscv-tools
 
+    $ export RISC_GNU_TOOLS_PATH=/opt/riscv-gnu-toolchain
+
     $ sudo apt-get -y install \
         binutils build-essential libtool texinfo \
         gzip zip unzip patchutils curl git \
@@ -35,26 +39,27 @@ Install gnu-riscv-tools
     $ cd riscv-gnu-toolchain
     $ git submodule update --init --recursive
     $ mkdir build
-    $ cd build
-    $ export RISC_GNU_TOOLS_PATH=$PWD/../drops
+    $ cd build    
     $ ../configure --prefix=$RISC_GNU_TOOLS_PATH --with-arch=rv32im --with-abi=ilp32
     $ make -j`nproc`  
     $ make -j`nproc` build-qemu
 
+Install Verilator
+
+    You need into build the latest version using the instructions on their website
+    $ https://www.veripool.org/projects/verilator/wiki/Installing 
+
 Install Vortex 
 
-    $ sudo apt-get install verilator
     $ git clone https://github.gatech.edu/casl/Vortex.git
-
-Build SimX
-
-    $ cd Vortex/simx
+    $ cd Vortex
     $ make
 
 Run SGEMM OpenCL Benchmark
 
     $ cd Vortex/benchmarks/opencl/sgemm
     $ make
+<<<<<<< HEAD
     $ make run
 
 Basic Instructions to build the OpenCL Compiler for Vortex
@@ -82,3 +87,6 @@ Build pocl for RISCV
     $ rm -rf *
     $ cmake -G Ninja -DCMAKE_INSTALL_PREFIX=$POCL_RT_PATH -DCMAKE_BUILD_TYPE=Debug -DOCS_AVAILABLE=OFF -DBUILD_SHARED_LIBS=OFF -DNEWLIB_BSP=ON -DNEWLIB_DEVICE_ADDRESS_BIT=32 -DNEWLIB_DEVICE_MARCH=rv32im -DBUILD_TESTS=OFF -DHOST_DEVICE_BUILD_HASH=basic-riscv32-unknown-elf -DCMAKE_TOOLCHAIN_FILE=../RISCV_newlib.cmake -DENABLE_TRACING=OFF -DENABLE_ICD=OFF -DPOCL_DEBUG_MESSAGES=ON ..
     $ cmake --build . --target install
+=======
+    $ make run
+>>>>>>> fpga_synthesis

benchmarks/old_opencl/BlackScholes/Makefile

@@ -12,15 +12,14 @@ 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)/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,$(VX_RT_PATH)/mains/vortex_link.ld
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_link.ld
 
 CXXFLAGS =  -g -O0 -march=rv32im -mabi=ilp32
 CXXFLAGS += -ffreestanding # program may not begin at main()

benchmarks/old_opencl/DotProduct/Makefile

@@ -12,15 +12,14 @@ 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)/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,$(VX_RT_PATH)/mains/vortex_link.ld
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_link.ld
 
 CXXFLAGS =  -g -O0 -march=rv32im -mabi=ilp32 
 CXXFLAGS += -ffreestanding # program may not begin at main()

benchmarks/old_opencl/VectorHypot/Makefile

@@ -12,15 +12,14 @@ 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)/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,$(VX_RT_PATH)/mains/vortex_link.ld
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_link.ld
 
 CXXFLAGS =  -g -O0 -march=rv32im -mabi=ilp32 
 CXXFLAGS += -ffreestanding # program may not begin at main()

benchmarks/old_opencl/bfs/CLHelper.h

@@ -0,0 +1,816 @@
+//------------------------------------------
+//--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)
+
+/*
+ * 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);*/
+  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/old_opencl/bfs/Makefile

@@ -0,0 +1,67 @@
+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 += $(VX_RT_PATH)/vx_api/vx_api.c
+
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_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 = bfs
+
+SRCS = main.cc
+
+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 -strace -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/old_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/old_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/old_opencl/bfs/run

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

benchmarks/old_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/old_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/old_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/old_opencl/convolution/Makefile

@@ -0,0 +1,67 @@
+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 += $(VX_RT_PATH)/vx_api/vx_api.c
+
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_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/old_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/old_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/old_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/old_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/old_opencl/cutcp/Makefile

@@ -12,15 +12,14 @@ 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)/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,$(VX_RT_PATH)/mains/vortex_link.ld
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_link.ld
 
 CXXFLAGS =  -g -O0 -march=rv32im -mabi=ilp32 
 CXXFLAGS += -ffreestanding # program may not begin at main()

benchmarks/old_opencl/guassian/Makefile

@@ -0,0 +1,67 @@
+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 += $(VX_RT_PATH)/vx_api/vx_api.c
+
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_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 = gaussian
+
+SRCS = main.cc clutils.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/old_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/old_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/old_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/old_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/old_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/old_opencl/guassian/main.cc

@@ -0,0 +1,412 @@
+#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/old_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/old_opencl/guassian/run

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

benchmarks/old_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/old_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/old_opencl/kmeans/Makefile

@@ -0,0 +1,78 @@
+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 += $(VX_RT_PATH)/vx_api/vx_api.c
+
+VX_CFLAGS = -nostartfiles -Wl,-Bstatic,-T,$(VX_RT_PATH)/startup/vx_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 = kmeans
+SRCS = main.cc read_input.c rmse.c cluster.c kmeans_clustering.c
+
+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
+
+kmeans_clustering.o: kmeans_clustering.c
+	$(CC) $(CXXFLAGS) -c kmeans_clustering.c
+
+cluster.o: cluster.c 
+	$(CC) $(CXXFLAGS) -c cluster.c
+
+read_input.o: read_input.c 
+	$(CC) $(CXXFLAGS) -c read_input.c
+
+rmse.o: rmse.c
+	$(CC) $(CXXFLAGS) -c rmse.c
+
+$(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 -strace -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/old_opencl/kmeans/cluster.c

@@ -0,0 +1,155 @@
+/*****************************************************************************/
+/*IMPORTANT:  READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.         */
+/*By downloading, copying, installing or using the software you agree        */
+/*to this license.  If you do not agree to this license, do not download,    */
+/*install, copy or use the software.                                         */
+/*                                                                           */
+/*                                                                           */
+/*Copyright (c) 2005 Northwestern University                                 */
+/*All rights reserved.                                                       */
+
+/*Redistribution of the software in source and binary forms,                 */
+/*with or without modification, is permitted provided that the               */
+/*following conditions are met:                                              */
+/*                                                                           */
+/*1       Redistributions of source code must retain the above copyright     */
+/*        notice, this list of conditions and the following disclaimer.      */
+/*                                                                           */
+/*2       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.*/ 
+/*                                                                            */
+/*3       Neither the name of Northwestern University 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, NON-INFRINGEMENT AND         */
+/*FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL          */
+/*NORTHWESTERN UNIVERSITY 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 THIS SOFTWARE, EVEN IF ADVISED OF THE             */
+/*POSSIBILITY OF SUCH DAMAGE.                                                 */
+/******************************************************************************/
+
+/*************************************************************************/
+/**   File:         cluster.c                                           **/
+/**   Description:  Takes as input a file, containing 1 data point per  **/
+/**                 per line, and performs a fuzzy c-means clustering   **/
+/**                 on the data. Fuzzy clustering is performed using    **/
+/**                 min to max clusters and the clustering that gets    **/
+/**                 the best score according to a compactness and       **/
+/**                 separation criterion are returned.                  **/
+/**   Author:  Brendan McCane                                           **/
+/**            James Cook University of North Queensland.               **/
+/**            Australia. email: mccane@cs.jcu.edu.au                   **/
+/**                                                                     **/
+/**   Edited by: Jay Pisharath, Wei-keng Liao                           **/
+/**              Northwestern University.                               **/
+/**																		**/
+/**   ================================================================  **/
+/**																		**/
+/**   Edited by: Shuai Che, David Tarjan, Sang-Ha Lee					**/
+/**				 University of Virginia									**/
+/**																		**/
+/**   Description:	No longer supports fuzzy c-means clustering;	 	**/
+/**					only regular k-means clustering.					**/
+/**					No longer performs "validity" function to analyze	**/
+/**					compactness and separation crietria; instead		**/
+/**					calculate root mean squared error.					**/
+/**                                                                     **/
+/*************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+#include <math.h>
+#include <float.h>
+#include "kmeans.h"
+
+float	min_rmse_ref = FLT_MAX;		
+extern double wtime(void);
+	/* reference min_rmse value */
+
+/*---< cluster() >-----------------------------------------------------------*/
+int cluster(int      npoints,				/* number of data points */
+            int      nfeatures,				/* number of attributes for each point */
+            float  **features,			/* array: [npoints][nfeatures] */                  
+            int      min_nclusters,			/* range of min to max number of clusters */
+			int		 max_nclusters,
+            float    threshold,				/* loop terminating factor */
+            int     *best_nclusters,		/* out: number between min and max with lowest RMSE */
+            float ***cluster_centres,		/* out: [best_nclusters][nfeatures] */
+			float	*min_rmse,				/* out: minimum RMSE */
+			int		 isRMSE,				/* calculate RMSE */
+			int		 nloops					/* number of iteration for each number of clusters */
+			)
+{    
+	int		nclusters;						/* number of clusters k */	
+	int		index =0;						/* number of iteration to reach the best RMSE */
+	int		rmse;							/* RMSE for each clustering */
+    int    *membership;						/* which cluster a data point belongs to */
+    float **tmp_cluster_centres;			/* hold coordinates of cluster centers */
+	int		i;
+
+	/* allocate memory for membership */
+    membership = (int*) malloc(npoints * sizeof(int));
+
+	/* sweep k from min to max_nclusters to find the best number of clusters */
+	for(nclusters = min_nclusters; nclusters <= max_nclusters; nclusters++)
+	{
+		if (nclusters > npoints) break;	/* cannot have more clusters than points */
+
+		/* allocate device memory, invert data array (@ kmeans_cuda.cu) */
+		allocate(npoints, nfeatures, nclusters, features);
+
+		/* iterate nloops times for each number of clusters */
+		for(i = 0; i < nloops; i++)
+		{
+			/* initialize initial cluster centers, CUDA calls (@ kmeans_cuda.cu) */
+			tmp_cluster_centres = kmeans_clustering(features,
+													nfeatures,
+													npoints,
+													nclusters,
+													threshold,
+													membership);
+
+			if (*cluster_centres) {
+				free((*cluster_centres)[0]);
+				free(*cluster_centres);
+			}
+			*cluster_centres = tmp_cluster_centres;
+	        
+					
+			/* find the number of clusters with the best RMSE */
+			if(isRMSE)
+			{
+				rmse = rms_err(features,
+							   nfeatures,
+							   npoints,
+							   tmp_cluster_centres,
+							   nclusters);
+				
+				if(rmse < min_rmse_ref){
+					min_rmse_ref = rmse;			//update reference min RMSE
+					*min_rmse = min_rmse_ref;		//update return min RMSE
+					*best_nclusters = nclusters;	//update optimum number of clusters
+					index = i;						//update number of iteration to reach best RMSE
+				}
+			}			
+		}
+		
+		deallocateMemory();							/* free device memory (@ kmeans_cuda.cu) */
+	}
+
+    free(membership);
+
+    return index;
+}
+

benchmarks/old_opencl/kmeans/getopt.h

@@ -0,0 +1,191 @@
+
+
+/* getopt.h */
+/* Declarations for getopt.
+   Copyright (C) 1989-1994, 1996-1999, 2001 Free Software 
+   Foundation, Inc. This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute 
+   it and/or modify it under the terms of the GNU Lesser 
+   General Public License as published by the Free Software 
+   Foundation; either version 2.1 of the License, or 
+   (at your option) any later version.
+
+   The GNU C Library 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 Lesser General Public 
+   License for more details.
+
+   You should have received a copy of the GNU Lesser General 
+   Public License along with the GNU C Library; if not, write 
+   to the Free Software Foundation, Inc., 59 Temple Place,
+   Suite 330, Boston, MA 02111-1307 USA.  */
+
+  
+
+  
+
+#ifndef _GETOPT_H
+
+#ifndef __need_getopt
+# define _GETOPT_H 1
+#endif
+
+/* If __GNU_LIBRARY__ is not already defined, either we are being used
+   standalone, or this is the first header included in the source file.
+   If we are being used with glibc, we need to include <features.h>, but
+   that does not exist if we are standalone.  So: if __GNU_LIBRARY__ is
+   not defined, include <ctype.h>, which will pull in <features.h> for us
+   if it's from glibc.  (Why ctype.h?  It's guaranteed to exist and it
+   doesn't flood the namespace with stuff the way some other headers do.)  */
+#if !defined __GNU_LIBRARY__
+# include <ctype.h>
+#endif
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+/* For communication from `getopt' to the caller.
+   When `getopt' finds an option that takes an argument,
+   the argument value is returned here.
+   Also, when `ordering' is RETURN_IN_ORDER,
+   each non-option ARGV-element is returned here.  */
+
+extern char *optarg;
+
+/* Index in ARGV of the next element to be scanned.
+   This is used for communication to and from the caller
+   and for communication between successive calls to `getopt'.
+
+   On entry to `getopt', zero means this is the first call; initialize.
+
+   When `getopt' returns -1, this is the index of the first of the
+   non-option elements that the caller should itself scan.
+
+   Otherwise, `optind' communicates from one call to the next
+   how much of ARGV has been scanned so far.  */
+
+extern int optind;
+
+/* Callers store zero here to inhibit the error message `getopt' prints
+   for unrecognized options.  */
+
+extern int opterr;
+
+/* Set to an option character which was unrecognized.  */
+
+extern int optopt;
+
+#ifndef __need_getopt
+/* Describe the long-named options requested by the application.
+   The LONG_OPTIONS argument to getopt_long or getopt_long_only is a vector
+   of `struct option' terminated by an element containing a name which is
+   zero.
+
+   The field `has_arg' is:
+   no_argument          (or 0) if the option does not take an argument,
+   required_argument    (or 1) if the option requires an argument,
+   optional_argument    (or 2) if the option takes an optional argument.
+
+   If the field `flag' is not NULL, it points to a variable that is set
+   to the value given in the field `val' when the option is found, but
+   left unchanged if the option is not found.
+
+   To have a long-named option do something other than set an `int' to
+   a compiled-in constant, such as set a value from `optarg', set the
+   option's `flag' field to zero and its `val' field to a nonzero
+   value (the equivalent single-letter option character, if there is
+   one).  For long options that have a zero `flag' field, `getopt'
+   returns the contents of the `val' field.  */
+
+struct option
+{
+# if (defined __STDC__ && __STDC__) || defined __cplusplus
+  const char *name;
+# else
+  char *name;
+# endif
+  /* has_arg can't be an enum because some compilers complain about
+     type mismatches in all the code that assumes it is an int.  */
+  int has_arg;
+  int *flag;
+  int val;
+};
+
+/* Names for the values of the `has_arg' field of `struct option'.  */
+
+# define no_argument            0
+# define required_argument      1
+# define optional_argument      2
+#endif  /* need getopt */
+
+
+/* Get definitions and prototypes for functions to process the
+   arguments in ARGV (ARGC of them, minus the program name) for
+   options given in OPTS.
+
+   Return the option character from OPTS just read.  Return -1 when
+   there are no more options.  For unrecognized options, or options
+   missing arguments, `optopt' is set to the option letter, and '?' is
+   returned.
+
+   The OPTS string is a list of characters which are recognized option
+   letters, optionally followed by colons, specifying that that letter
+   takes an argument, to be placed in `optarg'.
+
+   If a letter in OPTS is followed by two colons, its argument is
+   optional.  This behavior is specific to the GNU `getopt'.
+
+   The argument `--' causes premature termination of argument
+   scanning, explicitly telling `getopt' that there are no more
+   options.
+
+   If OPTS begins with `--', then non-option arguments are treated as
+   arguments to the option '\0'.  This behavior is specific to the GNU
+   `getopt'.  */
+
+#if (defined __STDC__ && __STDC__) || defined __cplusplus
+# ifdef __GNU_LIBRARY__
+/* Many other libraries have conflicting prototypes for getopt, with
+   differences in the consts, in stdlib.h.  To avoid compilation
+   errors, only prototype getopt for the GNU C library.  */
+extern int getopt (int ___argc, char *const *___argv, const char *__shortopts);
+# else /* not __GNU_LIBRARY__ */
+extern int getopt ();
+# endif /* __GNU_LIBRARY__ */
+
+# ifndef __need_getopt
+extern int getopt_long (int ___argc, char *const *___argv,
+                        const char *__shortopts,
+                        const struct option *__longopts, int *__longind);
+extern int getopt_long_only (int ___argc, char *const *___argv,
+                             const char *__shortopts,
+                             const struct option *__longopts, int *__longind);
+
+/* Internal only.  Users should not call this directly.  */
+extern int _getopt_internal (int ___argc, char *const *___argv,
+                             const char *__shortopts,
+                             const struct option *__longopts, int *__longind,
+                             int __long_only);
+# endif
+#else /* not __STDC__ */
+extern int getopt ();
+# ifndef __need_getopt
+extern int getopt_long ();
+extern int getopt_long_only ();
+
+extern int _getopt_internal ();
+# endif
+#endif /* __STDC__ */
+
+#ifdef  __cplusplus
+}
+#endif
+
+/* Make sure we later can get all the definitions and declarations.  */
+#undef __need_getopt
+
+#endif /* getopt.h */
+

benchmarks/old_opencl/kmeans/kernel.cl

@@ -0,0 +1,61 @@
+#ifndef FLT_MAX
+#define FLT_MAX 3.40282347e+38
+#endif
+
+__kernel void
+kmeans_kernel_c(__global float  *feature,   
+			  __global float  *clusters,
+			  __global int    *membership,
+			    int     npoints,
+				int     nclusters,
+				int     nfeatures,
+				int		offset,
+				int		size
+			  ) 
+{
+	unsigned int point_id = get_global_id(0);
+    int index = 0;
+    //const unsigned int point_id = get_global_id(0);
+		if (point_id < npoints)
+		{
+			float min_dist=FLT_MAX;
+			for (int i=0; i < nclusters; i++) {
+				
+				float dist = 0;
+				float ans  = 0;
+				for (int l=0; l<nfeatures; l++){
+						ans += (feature[l * npoints + point_id]-clusters[i*nfeatures+l])* 
+							   (feature[l * npoints + point_id]-clusters[i*nfeatures+l]);
+				}
+
+				dist = ans;
+				if (dist < min_dist) {
+					min_dist = dist;
+					index    = i;
+					
+				}
+			}
+		  //printf("%d\n", index);
+		  membership[point_id] = index;
+		}	
+	
+	return;
+}
+
+__kernel void
+kmeans_swap(__global float  *feature,   
+			__global float  *feature_swap,
+			int     npoints,
+			int     nfeatures
+){
+
+	unsigned int tid = get_global_id(0);
+	//for(int i = 0; i <  nfeatures; i++)
+	//	feature_swap[i * npoints + tid] = feature[tid * nfeatures + i];
+    //Lingjie Zhang modificated at 11/05/2015
+    if (tid < npoints){
+	    for(int i = 0; i <  nfeatures; i++)
+		    feature_swap[i * npoints + tid] = feature[tid * nfeatures + i];
+    }
+    // end of Lingjie Zhang's modification
+}