/* Copyright (c) 2018-2022 Advanced Micro Devices, Inc.

 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal
 in the Software without restriction, including without limitation the rights
 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 copies of the Software, and to permit persons to whom the Software is
 furnished to do so, subject to the following conditions:

 The above copyright notice and this permission notice shall be included in
 all copies or substantial portions of the Software.

 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE. */

#include <iostream>

// hip header file
#include <hip/hip_runtime.h>
#include "roctracer_ext.h"
// roctx header file
#include <roctx.h>


#define WIDTH 1024


#define NUM (WIDTH * WIDTH)

#define THREADS_PER_BLOCK_X 4
#define THREADS_PER_BLOCK_Y 4
#define THREADS_PER_BLOCK_Z 1

// Mark API
extern "C" void roctracer_mark(const char* str);

// Device (Kernel) function, it must be void
__global__ void matrixTranspose(float* out, float* in, const int width) {
  int x = hipBlockDim_x * hipBlockIdx_x + hipThreadIdx_x;
  int y = hipBlockDim_y * hipBlockIdx_y + hipThreadIdx_y;

  out[y * width + x] = in[x * width + y];
}

// CPU implementation of matrix transpose
void matrixTransposeCPUReference(float* output, float* input, const unsigned int width) {
  for (unsigned int j = 0; j < width; j++) {
    for (unsigned int i = 0; i < width; i++) {
      output[i * width + j] = input[j * width + i];
    }
  }
}

int main() {
  float* Matrix;
  float* TransposeMatrix;
  float* cpuTransposeMatrix;

  float* gpuMatrix;
  float* gpuTransposeMatrix;

  hipDeviceProp_t devProp;
  hipGetDeviceProperties(&devProp, 0);

  std::cerr << "Device name " << devProp.name << std::endl;

  int i;
  int errors;

  Matrix = (float*)malloc(NUM * sizeof(float));
  TransposeMatrix = (float*)malloc(NUM * sizeof(float));
  cpuTransposeMatrix = (float*)malloc(NUM * sizeof(float));

  // initialize the input data
  for (i = 0; i < NUM; i++) {
    Matrix[i] = (float)i * 10.0f;
  }

  // allocate the memory on the device side
  hipMalloc((void**)&gpuMatrix, NUM * sizeof(float));
  hipMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(float));

  uint32_t iterations = 100;
  while (iterations-- > 0) {
    std::cerr << "## Iteration (" << iterations << ") #################" << std::endl;

    // Memory transfer from host to device
    hipMemcpy(gpuMatrix, Matrix, NUM * sizeof(float), hipMemcpyHostToDevice);

    roctracer_mark("before HIP LaunchKernel");
    roctxMark("before hipLaunchKernel");
    int rangeId = roctxRangeStart("hipLaunchKernel range");
    roctxRangePush("hipLaunchKernel");
    // Lauching kernel from host
    hipLaunchKernelGGL(
        matrixTranspose, dim3(WIDTH / THREADS_PER_BLOCK_X, WIDTH / THREADS_PER_BLOCK_Y),
        dim3(THREADS_PER_BLOCK_X, THREADS_PER_BLOCK_Y), 0, 0, gpuTransposeMatrix, gpuMatrix, WIDTH);
    roctracer_mark("after HIP LaunchKernel");
    roctxMark("after hipLaunchKernel");

    // Memory transfer from device to host
    roctxRangePush("hipMemcpy");

    hipMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(float), hipMemcpyDeviceToHost);

    roctxRangePop();  // for "hipMemcpy"
    roctxRangePop();  // for "hipLaunchKernel"
    roctxRangeStop(rangeId);

    // CPU MatrixTranspose computation
    matrixTransposeCPUReference(cpuTransposeMatrix, Matrix, WIDTH);

    // verify the results
    errors = 0;
    double eps = 1.0E-6;
    for (i = 0; i < NUM; i++) {
      if (std::abs(TransposeMatrix[i] - cpuTransposeMatrix[i]) > eps) {
        errors++;
      }
    }
    if (errors != 0) {
      fprintf(stderr, "FAILED: %d errors\n", errors);
    } else {
      fprintf(stderr, "PASSED!\n");
    }
  }

  // free the resources on device side
  hipFree(gpuMatrix);
  hipFree(gpuTransposeMatrix);

  // free the resources on host side
  free(Matrix);
  free(TransposeMatrix);
  free(cpuTransposeMatrix);

  return errors;
}