testPerformance.cc

Go to the documentation of this file.

/*
 * Advanced Simulation Library <http://asl.org.il>
 * 
 * Copyright 2015 Avtech Scientific <http://avtechscientific.com>
 *
 *
 * This file is part of Advanced Simulation Library (ASL).
 *
 * ASL is free software: you can redistribute it and/or modify it
 * under the terms of the GNU Affero General Public License as
 * published by the Free Software Foundation, version 3 of the License.
 *
 * ASL 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 Affero General Public License for more details.
 *
 * You should have received a copy of the GNU Affero General Public License
 * along with ASL. If not, see <http://www.gnu.org/licenses/>.
 *
 */
 

 
#include "acl/acl.h"
#include "acl/aclHardware.h"
#include "acl/DataTypes/aclIndex.h"
#include "acl/DataTypes/aclConstant.h"
#include "acl/DataTypes/aclVariable.h"
#include "acl/DataTypes/aclPrivateVariable.h"
#include "acl/DataTypes/aclArray.h"
#include "acl/DataTypes/aclSubvector.h"
#include "acl/Kernels/aclKernel.h" 
#include "aslUtilities.h"
#include "utilities/aslTimer.h"
#include <math.h>
#include <fstream>
 
#include <acl/Kernels/aclKernelConfigurationTemplates.h>
 
#define ARRAY_SIZE 10000000
//#define ITERATIONS_NUM 20
#define TIME_INTERVAL 5 // in seconds
 
using namespace acl;
using namespace std;
 
 
// Operators: +
template <typename T> inline T testSum(T x1, T x2)
{ 
  using namespace elementOperators;
  return x1 + x2;
}
 
 
// Operators: + (non sequential)
template <typename T> inline T testSumNonSequential(T x1, T x2)
{
  Element index(new Index(ARRAY_SIZE));
  Element c(new Constant<cl_uint>(17));
  Element cArraySize(new Constant<cl_uint>(ARRAY_SIZE));
  using namespace elementOperators;
  return excerpt(x1 + x2, (index * c) % cArraySize);
}
 
 
// Operators: +, -, *, /
template <typename T> inline T testBasicOperators(T x1, T x2)
{
  using namespace elementOperators;
  return x1 * (x2 + x2) * x1 + (x2 + x1 * x1) * x2 * x1 * x2 * x1 - x1 / x2 + 
        x1 * (x1 + x2) * x1 + (x2 - x1 * x2) * x2 - x1 * x2 * x1 - x2 / x1 + 
        x1 * (x2 + x1) * x1 - (x2 + x1 * x1) * x2 * x1 * x2 * x1 + x1 / x2 -
        x1 * (x1 + x2) * x2 + (x2 - x1 * x2) * x2 - x1 * x2 * x1 - x2 / x1 + 
        x1 * (x2 + x1) * x1 - (x2 + x1 * x1) * x2 * x1 * x2 * x1 + x1 / x2;
}
 
 
// Operators: +, -, *, /, sin, cos, sqrt
template <typename T> inline T testSpecialOperators(T x1, T x2)
{
  using namespace elementOperators;
  return cos(x1*sin(x2+x2)*x1+sqrt(x2+x1*x1)*x2*sin(x1)*x1*sqrt(x2)*x2*x1*sqrt(x1*x2)-x1/x2);
}
 
 
/*
void print(Glib::KeyFile * keyFile,
           string deviceName,
           string key,
           string value,
           bool writeToKeyFile)
{
  cout << key + ": " + value << endl;
  
  // Save the output for this device to the performance report file if needed
  if (writeToKeyFile)
  {
    keyFile->set_string((deviceName),
                        key,
                        value);
  }
}
*/
 
/*
template <typename T> inline void testKernelPerformance(const KernelConfiguration kernelConfig,
                                                        const CommandQueue & queue,
                                                        Glib::KeyFile * keyFile,
                                                        bool writeToKeyFile)
{
  Element arr1(new Array<T>(ARRAY_SIZE, queue));
  Element arr2(new Array<T>(ARRAY_SIZE, queue));
  Element arrResult(new Array<T>(ARRAY_SIZE, queue));
  Element c(new Constant<T>(5));
 
  string typeStr = typeToStr<T>();
  
  string kernelConfigStr;
  if (kernelConfig == KERNEL_BASIC)
    kernelConfigStr = "KERNEL_BASIC";
  else
    if (kernelConfig == KERNEL_SIMD)
      kernelConfigStr = "KERNEL_SIMD";
    else
      if (kernelConfig == KERNEL_SIMDUA)
        kernelConfigStr = "KERNEL_SIMDUA";
 
  
  Kernel k(kernelConfig);
  
  // Initialization of the arrays
  {
    using namespace elementOperators;
    k.addExpression(operatorAssignmentSafe(arr1, c));
    k.addExpression(operatorAssignmentSafe(arr2, c));
  }
  k.compute();
 
  // Test of sum
  k.clear();
  { 
    using namespace elementOperators;
    k.addExpression(operatorAssignmentSafe(arrResult, testSum(arr1, arr2)));
  }
  k.setup();
  
  asl::Timer timer;
  unsigned int iterationsNum;
 
  timer.start();
  timer.stop();
  iterationsNum = 0;
  do
  {
    timer.resume();
    k.compute();
    ++iterationsNum;
    timer.stop();
  }
  while (timer.getTime() < TIME_INTERVAL);
 
  print(keyFile,
        getDeviceName(queue),
        kernelConfigStr + "---" + typeStr + "---" + "Sum",
        numToStr((float) iterationsNum / timer.getTime()),
        writeToKeyFile);
 
  // Test of non sequential sum (only for non SIMD kernels)
  if (kernelConfig.vectorWidth == 1)
  {
    k.clear();
    { 
      using namespace elementOperators;
      k.addExpression(operatorAssignmentSafe(arrResult,
                                             testSumNonSequential(arr1, arr2)));
    }
    k.setup();
    
    timer.start();
    timer.stop();
    iterationsNum = 0;
    do
    {
      timer.resume();
      k.compute();
      ++iterationsNum;
      timer.stop();
    }
    while (timer.getTime() < TIME_INTERVAL);
 
    print(keyFile,
          getDeviceName(queue),
          kernelConfigStr + "---" + typeStr + "---" + "NonSequentialSum",
          numToStr((float) iterationsNum / timer.getTime()),
          writeToKeyFile);
 
  }
 
  // Test of basic operators
  k.clear();
  { 
    using namespace elementOperators;
    k.addExpression(operatorAssignmentSafe(arrResult,
                                           testBasicOperators(arr1, arr2)));
  }
  k.setup();
  
  timer.start();
  timer.stop();
  iterationsNum = 0;
  do
  {
    timer.resume();
    k.compute();
    ++iterationsNum;
    timer.stop();
  }
  while (timer.getTime() < TIME_INTERVAL);
 
  print(keyFile,
        getDeviceName(queue),
        kernelConfigStr + "---" + typeStr + "---" + "BasicOperators",
        numToStr((float) iterationsNum / timer.getTime()),
        writeToKeyFile);
 
 
  // Test of special operators
  k.clear();
  if (arr1->getTypeID() != TYPE_INT)
  {
    { 
      using namespace elementOperators;
      k.addExpression(operatorAssignmentSafe(arrResult,
                                             testSpecialOperators(arr1, arr2)));
    }
    k.setup();
  
    timer.start();
    timer.stop();
    iterationsNum = 0;
    do
    {
      timer.resume();
      k.compute();
      ++iterationsNum;
      timer.stop();
    }
    while (timer.getTime() < TIME_INTERVAL);
 
    print(keyFile,
          getDeviceName(queue),
          kernelConfigStr + "---" + typeStr + "---" + "SpecialOperators",
          numToStr((float) iterationsNum / timer.getTime()),
          writeToKeyFile);
  }
}
*/
 
int main()
{
  /*
  const string FILE_NAME("./PerformanceReport.txt");
  bool writeToKeyFile = false;
  bool dumpKeyFile = false;
  
  Glib::KeyFile *keyFile = new Glib::KeyFile;
 
  ifstream fileCheck(FILE_NAME);
  if (fileCheck.good())
  {
    keyFile->load_from_file(FILE_NAME);
  }
  else
  {
    warningMessage("Failed to open " + FILE_NAME + " . Creating new file");
    keyFile->set_comment("\n Performance test report\n\n");
  }
 
  
  for (unsigned int i = 0; i < hardware.queues.size(); ++i)
  {
    if (!keyFile->has_group(getDeviceName(hardware.queues[i])))
    {
      writeToKeyFile = true;
      dumpKeyFile = true;
    }
    
    cout << "\nDevice: " + getDeviceName(hardware.queues[i]) + "\n" << endl;
    
    testKernelPerformance<cl_int>(KERNEL_BASIC,
                                  hardware.queues[i],
                                  keyFile,
                                  writeToKeyFile);
    testKernelPerformance<cl_float>(KERNEL_BASIC,
                                    hardware.queues[i],
                                    keyFile,
                                    writeToKeyFile);
    if (doublePrecisionSupport(hardware.queues[i]))
      testKernelPerformance<cl_double>(KERNEL_BASIC,
                                       hardware.queues[i],
                                       keyFile,
                                       writeToKeyFile);
 
 
    testKernelPerformance<cl_int>(KERNEL_SIMD,
                                  hardware.queues[i],
                                  keyFile,
                                  writeToKeyFile);
    testKernelPerformance<cl_float>(KERNEL_SIMD,
                                    hardware.queues[i],
                                    keyFile,
                                    writeToKeyFile);
    if (doublePrecisionSupport(hardware.queues[i]))
      testKernelPerformance<cl_double>(KERNEL_SIMD,
                                       hardware.queues[i],
                                       keyFile,
                                       writeToKeyFile);
 
 
    testKernelPerformance<cl_int>(KERNEL_SIMDUA,
                                  hardware.queues[i],
                                  keyFile,
                                  writeToKeyFile);
    testKernelPerformance<cl_float>(KERNEL_SIMDUA,
                                    hardware.queues[i],
                                    keyFile,
                                    writeToKeyFile);
    if (doublePrecisionSupport(hardware.queues[i]))
      testKernelPerformance<cl_double>(KERNEL_SIMDUA,
                                       hardware.queues[i],
                                       keyFile,
                                       writeToKeyFile);
 
    writeToKeyFile = false;
  }
 
  if (dumpKeyFile)
  {
    ofstream file;
    file.open(FILE_NAME, ios::app);
    if (!file.good())
    {
      errorMessage("Opening file " + FILE_NAME + " failed");
    }
    else
    {
      file << keyFile->to_data();
      if (!file.good())
        errorMessage("Writing to file " + FILE_NAME + " failed");
    }
    file.close();
  }
 
  delete keyFile;
  */
  return 0;
}