cppOverload.cu

/* Copyright (c) 2021, NVIDIA CORPORATION. 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 NVIDIA CORPORATION 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 ``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 OWNER 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.
 */

#define THREAD_N 256
#define N 1024
#define DIV_UP(a, b) (((a) + (b)-1) / (b))

// Includes, system
#include <stdio.h>
#include <helper_cuda.h>
#include <helper_string.h>
#include <helper_math.h>
#include "cppOverload_kernel.cuh"

const char *sampleName = "C++ Function Overloading";

#define OUTPUT_ATTR(attr)                                         \
  printf("Shared Size:   %d\n", (int)attr.sharedSizeBytes);       \
  printf("Constant Size: %d\n", (int)attr.constSizeBytes);        \
  printf("Local Size:    %d\n", (int)attr.localSizeBytes);        \
  printf("Max Threads Per Block: %d\n", attr.maxThreadsPerBlock); \
  printf("Number of Registers: %d\n", attr.numRegs);              \
  printf("PTX Version: %d\n", attr.ptxVersion);                   \
  printf("Binary Version: %d\n", attr.binaryVersion);

bool check_func1(int *hInput, int *hOutput, int a) {
  for (int i = 0; i < N; ++i) {
    int cpuRes = hInput[i] * a + i;

    if (hOutput[i] != cpuRes) {
      return false;
    }
  }

  return true;
}

bool check_func2(int2 *hInput, int *hOutput, int a) {
  for (int i = 0; i < N; i++) {
    int cpuRes = (hInput[i].x + hInput[i].y) * a + i;

    if (hOutput[i] != cpuRes) {
      return false;
    }
  }

  return true;
}

bool check_func3(int *hInput1, int *hInput2, int *hOutput, int a) {
  for (int i = 0; i < N; i++) {
    if (hOutput[i] != (hInput1[i] + hInput2[i]) * a + i) {
      return false;
    }
  }

  return true;
}

int main(int argc, const char *argv[]) {
  int *hInput = NULL;
  int *hOutput = NULL;
  int *dInput = NULL;
  int *dOutput = NULL;

  printf("%s starting...\n", sampleName);

  int deviceCount;
  checkCudaErrors(cudaGetDeviceCount(&deviceCount));
  printf("Device Count: %d\n", deviceCount);

  int deviceID = findCudaDevice(argc, argv);
  cudaDeviceProp prop;
  checkCudaErrors(cudaGetDeviceProperties(&prop, deviceID));
  if (prop.major < 2) {
    printf(
        "ERROR: cppOverload requires GPU devices with compute SM 2.0 or "
        "higher.\n");
    printf("Current GPU device has compute SM%d.%d, Exiting...", prop.major,
           prop.minor);
    exit(EXIT_WAIVED);
  }

  checkCudaErrors(cudaSetDevice(deviceID));

  // Allocate device memory
  checkCudaErrors(cudaMalloc(&dInput, sizeof(int) * N * 2));
  checkCudaErrors(cudaMalloc(&dOutput, sizeof(int) * N));

  // Allocate host memory
  checkCudaErrors(cudaMallocHost(&hInput, sizeof(int) * N * 2));
  checkCudaErrors(cudaMallocHost(&hOutput, sizeof(int) * N));

  for (int i = 0; i < N * 2; i++) {
    hInput[i] = i;
  }

  // Copy data from host to device
  checkCudaErrors(
      cudaMemcpy(dInput, hInput, sizeof(int) * N * 2, cudaMemcpyHostToDevice));

  // Test C++ overloading
  bool testResult = true;
  bool funcResult = true;
  int a = 1;

  void (*func1)(const int *, int *, int);
  void (*func2)(const int2 *, int *, int);
  void (*func3)(const int *, const int *, int *, int);
  struct cudaFuncAttributes attr;

  // overload function 1
  func1 = simple_kernel;
  memset(&attr, 0, sizeof(attr));
  checkCudaErrors(cudaFuncSetCacheConfig(*func1, cudaFuncCachePreferShared));
  checkCudaErrors(cudaFuncGetAttributes(&attr, *func1));
  OUTPUT_ATTR(attr);
  (*func1)<<<DIV_UP(N, THREAD_N), THREAD_N>>>(dInput, dOutput, a);
  checkCudaErrors(
      cudaMemcpy(hOutput, dOutput, sizeof(int) * N, cudaMemcpyDeviceToHost));
  funcResult = check_func1(hInput, hOutput, a);
  printf("simple_kernel(const int *pIn, int *pOut, int a) %s\n\n",
         funcResult ? "PASSED" : "FAILED");
  testResult &= funcResult;

  // overload function 2
  func2 = simple_kernel;
  memset(&attr, 0, sizeof(attr));
  checkCudaErrors(cudaFuncSetCacheConfig(*func2, cudaFuncCachePreferShared));
  checkCudaErrors(cudaFuncGetAttributes(&attr, *func2));
  OUTPUT_ATTR(attr);
  (*func2)<<<DIV_UP(N, THREAD_N), THREAD_N>>>((int2 *)dInput, dOutput, a);
  checkCudaErrors(
      cudaMemcpy(hOutput, dOutput, sizeof(int) * N, cudaMemcpyDeviceToHost));
  funcResult = check_func2(reinterpret_cast<int2 *>(hInput), hOutput, a);
  printf("simple_kernel(const int2 *pIn, int *pOut, int a) %s\n\n",
         funcResult ? "PASSED" : "FAILED");
  testResult &= funcResult;

  // overload function 3
  func3 = simple_kernel;
  memset(&attr, 0, sizeof(attr));
  checkCudaErrors(cudaFuncSetCacheConfig(*func3, cudaFuncCachePreferShared));
  checkCudaErrors(cudaFuncGetAttributes(&attr, *func3));
  OUTPUT_ATTR(attr);
  (*func3)<<<DIV_UP(N, THREAD_N), THREAD_N>>>(dInput, dInput + N, dOutput, a);
  checkCudaErrors(
      cudaMemcpy(hOutput, dOutput, sizeof(int) * N, cudaMemcpyDeviceToHost));
  funcResult = check_func3(&hInput[0], &hInput[N], hOutput, a);
  printf(
      "simple_kernel(const int *pIn1, const int *pIn2, int *pOut, int a) "
      "%s\n\n",
      funcResult ? "PASSED" : "FAILED");
  testResult &= funcResult;

  checkCudaErrors(cudaFree(dInput));
  checkCudaErrors(cudaFree(dOutput));
  checkCudaErrors(cudaFreeHost(hOutput));
  checkCudaErrors(cudaFreeHost(hInput));

  checkCudaErrors(cudaDeviceSynchronize());

  exit(testResult ? EXIT_SUCCESS : EXIT_FAILURE);
}