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

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

#include <hip/hip_fp16.h>
#include <hip_test_common.hh>

#pragma clang diagnostic ignored "-Wunused-variable"
#pragma clang diagnostic ignored "-Wunused-parameter"

__device__ void test_convert() {
  __half x;
  float y = static_cast<float>(x);
}

__global__ void __halfMath(bool* result, __half a) {
  result[0] = __heq(__hadd(a, __half{1}), __half{2});
  result[0] = __heq(__hadd_sat(a, __half{1}), __half{1}) && result[0];
  result[0] = __heq(__hfma(a, __half{2}, __half{3}), __half{5}) && result[0];
  result[0] = __heq(__hfma_sat(a, __half{2}, __half{3}), __half{1}) && result[0];
  result[0] = __heq(__hsub(a, __half{1}), __half{0}) && result[0];
  result[0] = __heq(__hsub_sat(a, __half{2}), __half{0}) && result[0];
  result[0] = __heq(__hmul(a, __half{2}), __half{2}) && result[0];
  result[0] = __heq(__hmul_sat(a, __half{2}), __half{1}) && result[0];
  result[0] = __heq(__hdiv(a, __half{2}), __half{0.5}) && result[0];
}

__device__ bool to_bool(const __half2& x) {
  auto r = static_cast<const __half2_raw&>(x);

  return r.data.x != 0 && r.data.y != 0;
}

__global__ void __half2Math(bool* result, __half2 a) {
  result[0] = to_bool(__heq2(__hadd2(a, __half2{1, 1}), __half2{2, 2}));
  result[0] = to_bool(__heq2(__hadd2_sat(a, __half2{1, 1}), __half2{1, 1})) && result[0];
  result[0] = to_bool(__heq2(__hfma2(a, __half2{2, 2}, __half2{3, 3}), __half2{5, 5})) && result[0];
  result[0] =
      to_bool(__heq2(__hfma2_sat(a, __half2{2, 2}, __half2{3, 3}), __half2{1, 1})) && result[0];
  result[0] = to_bool(__heq2(__hsub2(a, __half2{1, 1}), __half2{0, 0})) && result[0];
  result[0] = to_bool(__heq2(__hsub2_sat(a, __half2{2, 2}), __half2{0, 0})) && result[0];
  result[0] = to_bool(__heq2(__hmul2(a, __half2{2, 2}), __half2{2, 2})) && result[0];
  result[0] = to_bool(__heq2(__hmul2_sat(a, __half2{2, 2}), __half2{1, 1})) && result[0];
  result[0] = to_bool(__heq2(__h2div(a, __half2{2, 2}), __half2{0.5, 0.5})) && result[0];
}

__global__ void kernel_hisnan(__half* input, int* output) {
  int tx = threadIdx.x;
  output[tx] = __hisnan(input[tx]);
}

__global__ void kernel_hisinf(__half* input, int* output) {
  int tx = threadIdx.x;
  output[tx] = __hisinf(input[tx]);
}

__global__ void testHalfAbs(float* p) {
  auto a = __float2half(*p);
  a = __habs(a);
  *p = __half2float(a);
}

__global__ void testHalf2Abs(float2* p) {
  auto a = __float22half2_rn(*p);
  a = __habs2(a);
  *p = __half22float2(a);
}

__half host_ushort_as_half(uint32_t s) {
  union {
    __half h;
    uint32_t s;
  } converter;
  converter.s = s;
  return converter.h;
}

void check_hisnan(int NUM_INPUTS, __half* inputCPU, __half* inputGPU) {
  // allocate memory
  auto memsize = NUM_INPUTS * sizeof(int);
  int* outputGPU = nullptr;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&outputGPU), memsize));

  // launch the kernel
  hipLaunchKernelGGL(kernel_hisnan, dim3(1), dim3(NUM_INPUTS), 0, 0, inputGPU, outputGPU);

  // copy output from device
  int* outputCPU = reinterpret_cast<int*>(malloc(memsize));
  HIP_CHECK(hipMemcpy(outputCPU, outputGPU, memsize, hipMemcpyDeviceToHost));

  // check output
  for (int i = 0; i < NUM_INPUTS; i++) {
    if ((2 <= i) && (i <= 5)) {  // inputs are nan, output should be true
      REQUIRE(outputCPU[i] == true);
    } else {  // inputs are NOT nan, output should be false
      REQUIRE(outputCPU[i] == false);
    }
  }

  // free memory
  free(outputCPU);
  HIP_CHECK(hipFree(outputGPU));
}


void check_hisinf(int NUM_INPUTS, __half* inputCPU, __half* inputGPU) {
  // allocate memory
  auto memsize = NUM_INPUTS * sizeof(int);
  int* outputGPU = nullptr;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&outputGPU), memsize));

  // launch the kernel
  hipLaunchKernelGGL(kernel_hisinf, dim3(1), dim3(NUM_INPUTS), 0, 0, inputGPU, outputGPU);

  // copy output from device
  int* outputCPU = reinterpret_cast<int*>(malloc(memsize));
  HIP_CHECK(hipMemcpy(outputCPU, outputGPU, memsize, hipMemcpyDeviceToHost));

  // check output
  for (int i = 0; i < NUM_INPUTS; i++) {
    if ((0 <= i) && (i <= 1)) {  // inputs are inf, output should be true
      REQUIRE(outputCPU[i] == true);
    } else {  // inputs are NOT inf, output should be false
      REQUIRE(outputCPU[i] == false);
    }
  }
  // free memory
  free(outputCPU);
  HIP_CHECK(hipFree(outputGPU));
}


void checkFunctional() {
  // allocate memory
  const int NUM_INPUTS = 16;
  auto memsize = NUM_INPUTS * sizeof(__half);
  __half* inputCPU = reinterpret_cast<__half*>(malloc(memsize));

  // populate inputs
  inputCPU[0] = host_ushort_as_half(0x7c00);   //  inf
  inputCPU[1] = host_ushort_as_half(0xfc00);   // -inf
  inputCPU[2] = host_ushort_as_half(0x7c01);   //  nan
  inputCPU[3] = host_ushort_as_half(0x7e00);   //  nan
  inputCPU[4] = host_ushort_as_half(0xfc01);   //  nan
  inputCPU[5] = host_ushort_as_half(0xfe00);   //  nan
  inputCPU[6] = host_ushort_as_half(0x0000);   //  0
  inputCPU[7] = host_ushort_as_half(0x8000);   // -0
  inputCPU[8] = host_ushort_as_half(0x7bff);   // max +ve normal
  inputCPU[9] = host_ushort_as_half(0xfbff);   // max -ve normal
  inputCPU[10] = host_ushort_as_half(0x0400);  // min +ve normal
  inputCPU[11] = host_ushort_as_half(0x8400);  // min -ve normal
  inputCPU[12] = host_ushort_as_half(0x03ff);  // max +ve sub-normal
  inputCPU[13] = host_ushort_as_half(0x83ff);  // max -ve sub-normal
  inputCPU[14] = host_ushort_as_half(0x0001);  // min +ve sub-normal
  inputCPU[15] = host_ushort_as_half(0x8001);  // min -ve sub-normal

  // copy inputs to the GPU
  __half* inputGPU = nullptr;
  HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&inputGPU), memsize));
  HIP_CHECK(hipMemcpy(inputGPU, inputCPU, memsize, hipMemcpyHostToDevice));

  // run checks
  check_hisnan(NUM_INPUTS, inputCPU, inputGPU);
  check_hisinf(NUM_INPUTS, inputCPU, inputGPU);

  // free memory
  HIP_CHECK(hipFree(inputGPU));
  free(inputCPU);
}

void checkHalfAbs() {
  SECTION("Half Abs") {
    float* p;
    HIP_CHECK(hipMalloc(&p, sizeof(float)));
    float pp = -2.1f;
    HIP_CHECK(hipMemcpy(p, &pp, sizeof(float), hipMemcpyDefault));
    hipLaunchKernelGGL(testHalfAbs, 1, 1, 0, 0, p);
    HIP_CHECK(hipMemcpy(&pp, p, sizeof(float), hipMemcpyDefault));
    HIP_CHECK(hipFree(p));
    REQUIRE(pp >= 0.0f);
  }
  SECTION("Half2 Abs") {
    float2* p;
    HIP_CHECK(hipMalloc(&p, sizeof(float2)));
    float2 pp;
    pp.x = -2.1f;
    pp.y = -1.1f;
    HIP_CHECK(hipMemcpy(p, &pp, sizeof(float2), hipMemcpyDefault));
    hipLaunchKernelGGL(testHalf2Abs, 1, 1, 0, 0, p);
    HIP_CHECK(hipMemcpy(&pp, p, sizeof(float2), hipMemcpyDefault));
    HIP_CHECK(hipFree(p));
    bool result = true;
    if (pp.x < 0.0f || pp.y < 0.0f) {
      result = false;
    }
    REQUIRE(result == true);
  }
}

HIP_TEST_CASE(Unit_hipTestHalf) {
  bool* result{nullptr};
  HIP_CHECK(hipHostMalloc(&result, sizeof(result)));

  SECTION("Test half math") {
    result[0] = false;
    hipLaunchKernelGGL(__halfMath, dim3(1, 1, 1), dim3(1, 1, 1), 0, 0, result, __half{1});
    HIP_CHECK(hipDeviceSynchronize());
    REQUIRE(result[0] == true);
  }
  SECTION("Test half math") {
    result[0] = false;
    hipLaunchKernelGGL(__half2Math, dim3(1, 1, 1), dim3(1, 1, 1), 0, 0, result, __half2{1, 1});
    HIP_CHECK(hipDeviceSynchronize());
    REQUIRE(result[0] == true);
  }
  SECTION("Functional checks") {
    checkFunctional();
    checkHalfAbs();
  }
  HIP_CHECK(hipHostFree(result));
}