rocm-systems/projects/hip-tests/catch/unit/rtc/hipRTCDeviceMalloc.cc

#include <hip_test_common.hh>

#include <hip/hiprtc.h>
#include <hip/hip_runtime.h>


#include <cassert>
#include <cstddef>
#include <memory>
#include <iostream>
#include <iterator>
#include <vector>

static constexpr auto NUM_THREADS{32};
static constexpr auto NUM_BLOCKS{10};

// Tests device side malloc and free functions within a hiprtc kernel using simple add operation
static constexpr auto devicemalloc{
    R"(
extern "C"
__global__
void devicemalloc(float* x, float* y, float* out, float** px, float** py, size_t n)
{
  size_t tid = blockIdx.x * blockDim.x + threadIdx.x;
  if (tid == 0) {
    *px = (float*)malloc(sizeof(float) * n);
    *py = (float*)malloc(sizeof(float) * n);
     for (int i = 0; i < n; i++) {
       (*px)[i] = x[i];
       (*py)[i] = y[i];
       out[i] = (*px)[i] + (*py)[i];
     }
     free(*px);
     free(*py);
  }
}
)"};

TEST_CASE("Unit_hiprtc_devicemalloc") {
  int pcieAtomic = 0;
  HIP_CHECK(hipDeviceGetAttribute(&pcieAtomic, hipDeviceAttributeHostNativeAtomicSupported, 0));
  if (!pcieAtomic) {
    HipTest::HIP_SKIP_TEST("Device doesn't support pcie atomic, Skipped");
    return;
  }

  using namespace std;
  hiprtcProgram prog;
  hiprtcCreateProgram(&prog,       // prog
                      devicemalloc,       // buffer
                      "devicemalloc.cu",  // name
                      0, nullptr, nullptr);
  hipDeviceProp_t props;
  int device = 0;
  HIP_CHECK(hipGetDeviceProperties(&props, device));
#ifdef __HIP_PLATFORM_AMD__
  std::string sarg = std::string("--gpu-architecture=") + props.gcnArchName;
#else
  std::string sarg = std::string("--fmad=false");
#endif
  const char* options[] = {sarg.c_str()};
  hiprtcResult compileResult{hiprtcCompileProgram(prog, 1, options)};
  size_t logSize;
  HIPRTC_CHECK(hiprtcGetProgramLogSize(prog, &logSize));
  if (logSize) {
    string log(logSize, '\0');
    HIPRTC_CHECK(hiprtcGetProgramLog(prog, &log[0]));
    std::cout << log << '\n';
  }
  REQUIRE(compileResult == HIPRTC_SUCCESS);
  size_t codeSize;
  HIPRTC_CHECK(hiprtcGetCodeSize(prog, &codeSize));

  vector<char> code(codeSize);
  HIPRTC_CHECK(hiprtcGetCode(prog, code.data()));

  HIPRTC_CHECK(hiprtcDestroyProgram(&prog));

  // Do hip malloc first so that we donot need to do a cuInit manually before calling hipModule APIs
  size_t n = NUM_THREADS * NUM_BLOCKS;
  size_t bufferSize = n * sizeof(float);

  float *dX, *dY, *dOut;
  HIP_CHECK(hipMalloc(&dX, bufferSize));
  HIP_CHECK(hipMalloc(&dY, bufferSize));
  HIP_CHECK(hipMalloc(&dOut, bufferSize));

  hipModule_t module;
  hipFunction_t kernel;
  HIP_CHECK(hipModuleLoadData(&module, code.data()));
  HIP_CHECK(hipModuleGetFunction(&kernel, module, "devicemalloc"));

  unique_ptr<float[]> hX{new float[n]};
  unique_ptr<float[]> hY{new float[n]};
  unique_ptr<float[]> hOut{new float[n]};
  for (size_t i = 0; i < n; ++i) {
    hX[i] = static_cast<float>(i);
    hY[i] = static_cast<float>(i * 2);
  }

  HIP_CHECK(hipMemcpy(dX, hX.get(), bufferSize, hipMemcpyHostToDevice));
  HIP_CHECK(hipMemcpy(dY, hY.get(), bufferSize, hipMemcpyHostToDevice));

  float **pA, **pB;
  HIP_CHECK(hipMalloc((float***)&pA, sizeof(float*)));
  HIP_CHECK(hipMalloc((float***)&pB, sizeof(float*)));

  struct {
    float* b_;
    float* c_;
    float* d_;
    float** e_;
    float** f_;
    size_t g_;
  } args{dX, dY, dOut, pA, pB, n};

  auto size = sizeof(args);
  void* config[] = {HIP_LAUNCH_PARAM_BUFFER_POINTER, &args, HIP_LAUNCH_PARAM_BUFFER_SIZE, &size,
                    HIP_LAUNCH_PARAM_END};

  HIP_CHECK(hipModuleLaunchKernel(kernel, NUM_BLOCKS, 1, 1, NUM_THREADS, 1, 1, 0, nullptr, nullptr, config));

  HIP_CHECK(hipMemcpy(hOut.get(), dOut, bufferSize, hipMemcpyDeviceToHost));

  HIP_CHECK(hipFree(dX));
  HIP_CHECK(hipFree(dY));
  HIP_CHECK(hipFree(dOut));

  HIP_CHECK(hipModuleUnload(module));

  for (size_t i = 0; i < n; ++i) {
    INFO("For " << i << " Value: " << fabs(hX[i] + hY[i] - hOut[i])
                << " with: " << (fabs(hOut[i] * 1.0f) * 1e-6));
    REQUIRE(fabs(hX[i] + hY[i] - hOut[i]) <= fabs(hOut[i]) * 1e-6);
  }
}