TheRock/rocm-systems/projects/hip-tests/catch/unit/deviceLib/AtomicsWithRandomActiveLanesInWavefront.cc

/*
 * Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
 *
 * SPDX-License-Identifier: MIT
 */

// Testcase Description:
// Verifies working of following Global Atomic Operations:
// 1. atomicAdd,
// 2. atomicSub,
// 3. atomicMax,
// 4. atomicMin,
// 5. atomicAnd,
// 6. atomicOr, and
// 7. atomicXor
// for INT, UNSIGNED and FLOAT input types.

// Tests consist of following scenarios:
// 1) Uniform input value to atomic operation (i.e. every thread will be
//    performing atomic operation with the same value)
// 2) Divergent input value (i.e. every thread with be performing
//    atomic operation on different values)

#include <hip_test_common.hh>

using namespace std;

////////////////////////////////////////////////////////////////////////////////
// Auto-Verification Code
////////////////////////////////////////////////////////////////////////////////

// Atomic Operations
enum AtomicOp : unsigned {
  Add,
  Sub,
  Max,
  Min,
  And,
  Or,
  Xor,
};

// Initial values for atomic ops and various types
template <typename T, AtomicOp Op> struct Initial {
  constexpr static T value;
};

template <> const int Initial<int, AtomicOp::Add>::value = 0;
template <> const unsigned Initial<unsigned, AtomicOp::Add>::value = 0;
template <> const float Initial<float, AtomicOp::Add>::value = 0.0f;

template <> const int Initial<int, AtomicOp::Sub>::value = 0;
template <> const unsigned Initial<unsigned, AtomicOp::Sub>::value = 0;
template <> const float Initial<float, AtomicOp::Sub>::value = 0.0f;

template <> const int Initial<int, AtomicOp::Min>::value = INT_MAX;
template <> const unsigned Initial<unsigned, AtomicOp::Min>::value = UINT_MAX;
template <> const float Initial<float, AtomicOp::Min>::value = std::numeric_limits<float>::min();

template <> const int Initial<int, AtomicOp::Max>::value = INT_MIN;
template <> const unsigned Initial<unsigned, AtomicOp::Max>::value = 0;
template <> const float Initial<float, AtomicOp::Max>::value = std::numeric_limits<float>::max();

template <> const int Initial<int, AtomicOp::And>::value = 0xffffffff;
template <> const unsigned Initial<unsigned, AtomicOp::And>::value = 0xffffffff;

template <> const int Initial<int, AtomicOp::Or>::value = 0;
template <> const unsigned Initial<unsigned, AtomicOp::Or>::value = 0;

template <> const int Initial<int, AtomicOp::Xor>::value = 0x5a5a5a5a;
template <> const unsigned Initial<unsigned, AtomicOp::Xor>::value = 0x5a5a5a5a;


// Uniform values for atomic ops and various types
template <typename T, AtomicOp Op> struct Uniform {
  constexpr static T value;
};

template <> const int Uniform<int, AtomicOp::Add>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Add>::value = 10;
template <> const float Uniform<float, AtomicOp::Add>::value = 10.0f;

template <> const int Uniform<int, AtomicOp::Sub>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Sub>::value = 10;
template <> const float Uniform<float, AtomicOp::Sub>::value = 10.0f;

template <> const int Uniform<int, AtomicOp::Min>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Min>::value = 10;
template <> const float Uniform<float, AtomicOp::Min>::value = 10.0f;

template <> const int Uniform<int, AtomicOp::Max>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Max>::value = 10;
template <> const float Uniform<float, AtomicOp::Max>::value = 10.0f;

template <> const int Uniform<int, AtomicOp::And>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::And>::value = 10;

template <> const int Uniform<int, AtomicOp::Or>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Or>::value = 10;

template <> const int Uniform<int, AtomicOp::Xor>::value = 10;
template <> const unsigned Uniform<unsigned, AtomicOp::Xor>::value = 10;

// Auto-verification APIs for uniform values for various atomic ops
template <typename T> bool verifyAdd(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Add>::value;
  T uniformValue = Uniform<T, AtomicOp::Add>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val += uniformValue;
  }
  return val == gpuData[0];
}

template <typename T> bool verifySub(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Sub>::value;
  T uniformValue = Uniform<T, AtomicOp::Sub>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val -= uniformValue;
  }
  return val == gpuData[1];
}

template <typename T> bool verifyMax(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Max>::value;
  T uniformValue = Uniform<T, AtomicOp::Max>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val = std::max<T>(val, uniformValue);
  }
  return val == gpuData[2];
}

template <typename T> bool verifyMin(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Min>::value;
  T uniformValue = Uniform<T, AtomicOp::Min>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val = std::min<T>(val, uniformValue);
  }
  return val == gpuData[3];
}

template <typename T> bool verifyAnd(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::And>::value;
  T uniformValue = Uniform<T, AtomicOp::And>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val &= uniformValue;
  }
  return val == gpuData[4];
}

template <typename T> bool verifyOr(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Or>::value;
  T uniformValue = Uniform<T, AtomicOp::Or>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val |= uniformValue;
  }
  return val == gpuData[5];
}

template <typename T> bool verifyXor(T* gpuData, int len, bool* activeLanes) {
  T val = Initial<T, AtomicOp::Xor>::value;
  T uniformValue = Uniform<T, AtomicOp::Xor>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val ^= uniformValue;
  }
  return val == gpuData[6];
}

// Auto-verification APIs for divergent values for various atomic ops
template <typename T>
bool verifyAdd_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Add>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val += divergentValue[i];
  }
  if (std::is_same<T, float>::value) {
    REQUIRE(val == Catch::Approx(gpuData[0]));
    return true;
  }
  return val == gpuData[0];
}

template <typename T>
bool verifySub_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Sub>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val -= divergentValue[i];
  }
  if (std::is_same<T, float>::value) {
    REQUIRE(val == Catch::Approx(gpuData[1]));
    return true;
  }
  return val == gpuData[1];
}

template <typename T>
bool verifyMax_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Max>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val = std::max<T>(val, divergentValue[i]);
  }

  if (std::is_same<T, float>::value) {
    REQUIRE(val == Catch::Approx(gpuData[2]));
    return true;
  }
  return val == gpuData[2];
}

template <typename T>
bool verifyMin_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Min>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val = std::min<T>(val, divergentValue[i]);
  }

  if (std::is_same<T, float>::value) {
    REQUIRE(val == Catch::Approx(gpuData[3]));
    return true;
  }
  return val == gpuData[3];
}

template <typename T>
bool verifyAnd_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::And>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val &= ~(1 << divergentValue[i]);
  }
  return val == gpuData[4];
}

template <typename T>
bool verifyOr_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Or>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val |= (1 << divergentValue[i]);
  }
  return val == gpuData[5];
}

template <typename T>
bool verifyXor_divValue(T* gpuData, int len, bool* activeLanes, T* divergentValue) {
  T val = Initial<T, AtomicOp::Xor>::value;
  for (int i = 0; i < len; ++i) {
    if (activeLanes[i]) val ^= divergentValue[i];
  }
  return val == gpuData[6];
}

// Kernels exercising atomic ops on uniform value for random set of active lanes in wavefront
template <typename T> __global__ void testAtomicAdd_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Add>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicAdd(&g_odata[0], uniformValue);
}

template <typename T> __global__ void testAtomicSub_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Sub>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicSub(&g_odata[1], uniformValue);
}


template <typename T> __global__ void testAtomicMax_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Max>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicMax(&g_odata[2], uniformValue);
}

template <typename T> __global__ void testAtomicMin_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Min>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicMin(&g_odata[3], uniformValue);
}


template <typename T> __global__ void testAtomicAnd_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::And>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicAnd(&g_odata[4], uniformValue);
}

template <typename T> __global__ void testAtomicOr_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Or>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicOr(&g_odata[5], uniformValue);
}

template <typename T> __global__ void testAtomicXor_uniValue(T* g_odata, bool* g_activelanes) {
  T uniformValue = Uniform<T, AtomicOp::Xor>::value;
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicXor(&g_odata[6], uniformValue);
}

// Kernels exercising atomic ops on divergent values for random set of active lanes in wavefront
template <typename T>
__global__ void testAtomicAdd_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicAdd(&g_odata[0], g_divergentvalue[tid]);
  ;
}

template <typename T>
__global__ void testAtomicSub_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicSub(&g_odata[1], g_divergentvalue[tid]);
}


template <typename T>
__global__ void testAtomicMax_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicMax(&g_odata[2], g_divergentvalue[tid]);
}

template <typename T>
__global__ void testAtomicMin_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicMin(&g_odata[3], g_divergentvalue[tid]);
}


template <typename T>
__global__ void testAtomicAnd_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicAnd(&g_odata[4], ~(1 << g_divergentvalue[tid]));
}

template <typename T>
__global__ void testAtomicOr_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicOr(&g_odata[5], (1 << g_divergentvalue[tid]));
}

template <typename T>
__global__ void testAtomicXor_divValue(T* g_odata, bool* g_activelanes, T* g_divergentvalue) {
  const unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
  if (g_activelanes[tid]) atomicXor(&g_odata[6], g_divergentvalue[tid]);
}


template <typename T> static void runIntTest() {
  bool testResult = true;
  unsigned int numThreads = 256;
  unsigned int numBlocks = 64;
  // This test exercises 7 atomic operations
  // 1. atomicAdd
  // 2. atomicSub
  // 3. atomicMax
  // 4. atomicMin
  // 5. atomicAnd
  // 6. atomicOr
  // 7. atomicXor
  unsigned int numData = 7;
  unsigned int memSize = sizeof(T) * numData;
  unsigned int N = numThreads * numBlocks;

  // allocate mem for the result on host side
  T* hOData = reinterpret_cast<T*>(malloc(memSize));
  bool* hIActiveLanes = (bool*)(malloc(N * sizeof(bool)));

  // initialize the memory
  hOData[0] = Initial<T, AtomicOp::Add>::value;
  hOData[1] = Initial<T, AtomicOp::Sub>::value;
  hOData[2] = Initial<T, AtomicOp::Max>::value;
  hOData[3] = Initial<T, AtomicOp::Min>::value;
  hOData[4] = Initial<T, AtomicOp::And>::value;
  hOData[5] = Initial<T, AtomicOp::Or>::value;
  hOData[6] = Initial<T, AtomicOp::Xor>::value;


  for (unsigned int i = 0; i < N; i++) {
    // Genearte random values betwenn 0 to 9 to get random active and inactive lanes for atomic
    // operations
    hIActiveLanes[i] = std::rand() / ((RAND_MAX + 1u) / 9) % 2 == 0;
  }

  // allocate device memory for result
  T* dOData;
  bool* dIActiveLanes;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dOData), memSize));
  HIP_CHECK(hipMalloc(reinterpret_cast<bool**>(&dIActiveLanes), N * sizeof(bool)));

  // copy host memory to device to initialize to zero
  HIP_CHECK(hipMemcpy(dOData, hOData, memSize, hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIActiveLanes, hIActiveLanes, N, hipMemcpyHostToDevice));

  // execute the atomicAdd kernel
  hipLaunchKernelGGL(testAtomicAdd_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyAdd(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicSub kernel
  hipLaunchKernelGGL(testAtomicSub_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifySub(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicMax kernel
  hipLaunchKernelGGL(testAtomicMax_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyMax(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicMin kernel
  hipLaunchKernelGGL(testAtomicMin_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyMin(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicAnd kernel
  hipLaunchKernelGGL(testAtomicAnd_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyAnd(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicOr kernel
  hipLaunchKernelGGL(testAtomicOr_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyOr(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicXor kernel
  hipLaunchKernelGGL(testAtomicXor_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyXor(hOData, numThreads * numBlocks, hIActiveLanes));

  // Cleanup memory
  free(hOData);
  free(hIActiveLanes);
  HIP_CHECK(hipFree(dOData));
  HIP_CHECK(hipFree(dIActiveLanes));
}

static void runFloatTest() {
  bool testResult = true;
  unsigned int numThreads = 256;
  unsigned int numBlocks = 64;
  unsigned int numData = 4;
  unsigned int memSize = sizeof(float) * numData;
  unsigned int N = numThreads * numBlocks;

  // allocate mem for the result on host side
  float* hOData = reinterpret_cast<float*>(malloc(memSize));
  bool* hIActiveLanes = (bool*)(malloc(N * sizeof(bool)));

  // initialize the memory
  hOData[0] = Initial<float, AtomicOp::Add>::value;
  hOData[1] = Initial<float, AtomicOp::Sub>::value;
  hOData[2] = Initial<float, AtomicOp::Max>::value;
  hOData[3] = Initial<float, AtomicOp::Min>::value;

  for (unsigned int i = 0; i < N; i++) {
    // Genearte random values betwenn 0 to 9 to get random active and inactive lanes for atomic
    // operations
    hIActiveLanes[i] = std::rand() / ((RAND_MAX + 1u) / 9) % 2 == 0;
  }

  // allocate device memory for result
  float* dOData;
  bool* dIActiveLanes;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dOData), memSize));
  HIP_CHECK(hipMalloc(reinterpret_cast<bool**>(&dIActiveLanes), N * sizeof(bool)));

  // copy host memory to device to initialize to zero
  HIP_CHECK(hipMemcpy(dOData, hOData, memSize, hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIActiveLanes, hIActiveLanes, N * sizeof(bool), hipMemcpyHostToDevice));

  // execute the atomicAdd kernel
  hipLaunchKernelGGL(testAtomicAdd_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyAdd(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicSub kernel
  hipLaunchKernelGGL(testAtomicSub_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifySub(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicMax kernel
  hipLaunchKernelGGL(testAtomicMax_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyMax(hOData, numThreads * numBlocks, hIActiveLanes));

  // execute the atomicMin kernel
  hipLaunchKernelGGL(testAtomicMin_uniValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult == verifyMin(hOData, numThreads * numBlocks, hIActiveLanes));

  // Cleanup memory
  free(hOData);
  free(hIActiveLanes);
  HIP_CHECK(hipFree(dOData));
  HIP_CHECK(hipFree(dIActiveLanes));
}


template <typename T> static void runDivIntTest() {
  bool testResult = true;
  unsigned int numThreads = 256;
  unsigned int numBlocks = 64;
  unsigned int numData = 7;
  unsigned int memSize = sizeof(T) * numData;
  unsigned int N = numThreads * numBlocks;

  // allocate mem for the result on host side
  T* hOData = reinterpret_cast<T*>(malloc(memSize));
  T* hIDivValues = reinterpret_cast<T*>(malloc(N * sizeof(T)));
  bool* hIActiveLanes = (bool*)(malloc(N * sizeof(bool)));

  // initialize the memory
  hOData[0] = Initial<T, AtomicOp::Add>::value;
  hOData[1] = Initial<T, AtomicOp::Sub>::value;
  hOData[2] = Initial<T, AtomicOp::Max>::value;
  hOData[3] = Initial<T, AtomicOp::Min>::value;
  hOData[4] = Initial<T, AtomicOp::And>::value;
  hOData[5] = Initial<T, AtomicOp::Or>::value;
  hOData[6] = Initial<T, AtomicOp::Xor>::value;

  for (unsigned int i = 0; i < N; i++) {
    // Genearte random values betwenn 0 to 9 to get radom active and inactive lanes for atomic
    // operations
    unsigned randomValue = std::rand() / ((RAND_MAX + 1u) / 9);
    hIDivValues[i] = randomValue;
    hIActiveLanes[i] = randomValue % 2 == 0;
  }

  // allocate device memory for result
  T* dOData;
  bool* dIActiveLanes;
  T* dIDivValues;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dOData), memSize));
  HIP_CHECK(hipMalloc(reinterpret_cast<bool**>(&dIActiveLanes), N * sizeof(bool)));
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dIDivValues), N * sizeof(T)));

  // copy host memory to device to initialize to zero
  HIP_CHECK(hipMemcpy(dOData, hOData, memSize, hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIActiveLanes, hIActiveLanes, N * sizeof(bool), hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIDivValues, hIDivValues, N * sizeof(T), hipMemcpyHostToDevice));

  // execute the atomicAdd kernel
  hipLaunchKernelGGL(testAtomicAdd_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyAdd_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicSub kernel
  hipLaunchKernelGGL(testAtomicSub_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifySub_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicMax kernel
  hipLaunchKernelGGL(testAtomicMax_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyMax_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicMin kernel
  hipLaunchKernelGGL(testAtomicMin_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyMin_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicAnd kernel
  hipLaunchKernelGGL(testAtomicAnd_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyAnd_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicOr kernel
  hipLaunchKernelGGL(testAtomicOr_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyOr_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicXor kernel
  hipLaunchKernelGGL(testAtomicXor_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyXor_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // Cleanup memory
  free(hOData);
  free(hIDivValues);
  free(hIActiveLanes);
  HIP_CHECK(hipFree(dOData));
  HIP_CHECK(hipFree(dIActiveLanes));
  HIP_CHECK(hipFree(dIDivValues));
}

static void runDivFloatTest() {
  bool testResult = true;
  unsigned int numThreads = 256;
  unsigned int numBlocks = 64;
  unsigned int numData = 4;
  unsigned int memSize = sizeof(float) * numData;
  unsigned int N = numThreads * numBlocks;

  // allocate mem for the result on host side
  float* hOData = reinterpret_cast<float*>(malloc(memSize));
  float* hIDivValues = reinterpret_cast<float*>(malloc(N * sizeof(float)));
  bool* hIActiveLanes = (bool*)(malloc(N * sizeof(bool)));

  // initialize the memory
  hOData[0] = Initial<float, AtomicOp::Add>::value;
  hOData[1] = Initial<float, AtomicOp::Sub>::value;
  hOData[2] = Initial<float, AtomicOp::Max>::value;
  hOData[3] = Initial<float, AtomicOp::Min>::value;

  for (unsigned int i = 0; i < N; i++) {
    // Genearte random values betwenn 0 to 9 to get random active and inactive lanes for atomic
    // operations
    unsigned randomValue = std::rand() / ((RAND_MAX + 1u) / 9);
    hIDivValues[i] = (float)std::rand() / (float)randomValue;
    hIActiveLanes[i] = randomValue % 2 == 0;
  }

  // allocate device memory for result
  float* dOData;
  bool* dIActiveLanes;
  float* dIDivValues;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dOData), memSize));
  HIP_CHECK(hipMalloc(reinterpret_cast<bool**>(&dIActiveLanes), N * sizeof(bool)));
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&dIDivValues), N * sizeof(float)));

  // copy host memory to device to initialize to zero
  HIP_CHECK(hipMemcpy(dOData, hOData, memSize, hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIActiveLanes, hIActiveLanes, N * sizeof(bool), hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dIDivValues, hIDivValues, N * sizeof(float), hipMemcpyHostToDevice));

  // execute the atomicAdd kernel
  hipLaunchKernelGGL(testAtomicAdd_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyAdd_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicSub kernel
  hipLaunchKernelGGL(testAtomicSub_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifySub_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicMax kernel
  hipLaunchKernelGGL(testAtomicMax_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyMax_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // execute the atomicMin kernel
  hipLaunchKernelGGL(testAtomicMin_divValue, dim3(numBlocks), dim3(numThreads), 0, 0, dOData,
                     dIActiveLanes, dIDivValues);

  // Copy result from device to host
  HIP_CHECK(hipMemcpy(hOData, dOData, memSize, hipMemcpyDeviceToHost));

  // Compute reference solution
  REQUIRE(testResult ==
          verifyMin_divValue(hOData, numThreads * numBlocks, hIActiveLanes, hIDivValues));

  // Cleanup memory
  free(hOData);
  free(hIDivValues);
  free(hIActiveLanes);
  HIP_CHECK(hipFree(dOData));
  HIP_CHECK(hipFree(dIActiveLanes));
  HIP_CHECK(hipFree(dIDivValues));
}

/*
This testcases perform the following scenario of atomic opearations on Uniform and Divergent values
for INT, UNSIGNED INT, and FLOAT types:
  Uniform and Divergent Integer (int, unsigned int):
    // 1. atomicAdd
    // 2. atomicSub
    // 3. atomicMax
    // 4. atomicMin
    // 5. atomicAnd
    // 6. atomicOr
    // 7. atomicXor
  Uniform and Divergent Float:
    // 1. atomicAdd
    // 2. atomicSub
    // 3. atomicMax
    // 4. atomicMin
*/
HIP_TEST_CASE(Unit_AtomicsWithRandomActiveLanesInWavefront) {
  SECTION("Uniform Integer - int") { runIntTest<int>(); }
  SECTION("Uniform Integer - unsigned int") { runIntTest<unsigned int>(); }
  SECTION("Uniform Float") { runFloatTest(); }
  SECTION("Divergent Integer - int") { runDivIntTest<int>(); }
  SECTION("Divergent Integer - unsigned int") { runDivIntTest<unsigned int>(); }
  SECTION("Divergent Float") { runDivFloatTest(); }
}