rocm-systems/catch/unit/stream/hipStreamWaitEvent.cc

/*
Copyright (c) 2022 Advanced Micro Devices, Inc. All rights reserved.
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.
*/
/*
Testcase Scenarios :
Unit_hipStreamWaitEvent_Negative - Test unsuccessful hipStreamWaitEvent when either event or flags are invalid
Unit_hipStreamWaitEvent_UninitializedStream_Negative - Test unsuccessful hipStreamWaitEvent when stream is uninitialized
Unit_hipStreamWaitEvent_Default - Test simple waiting for an event with hipStreamWaitEvent api
Unit_hipStreamWaitEvent_DifferentStreams - Test waiting for an event on a different stream with hipStreamWaitEvent api
*/

#include <hip_test_common.hh>

TEST_CASE("Unit_hipStreamWaitEvent_Negative") {
  enum class StreamTestType { NullStream = 0, StreamPerThread, CreatedStream };

  auto streamType = GENERATE(StreamTestType::NullStream, StreamTestType::StreamPerThread,
                             StreamTestType::CreatedStream);

  hipStream_t stream{nullptr};
  hipEvent_t event{nullptr};

  if (streamType == StreamTestType::StreamPerThread) {
    stream = hipStreamPerThread;
  } else if (streamType == StreamTestType::CreatedStream) {
    HIP_CHECK(hipStreamCreate(&stream));
  }

  HIP_CHECK(hipEventCreate(&event));

  REQUIRE((stream != nullptr) != (streamType == StreamTestType::NullStream));
  REQUIRE(event != nullptr);

  SECTION("Invalid Event") {
    INFO("Running against Invalid Event");
    HIP_CHECK_ERROR(hipStreamWaitEvent(stream, nullptr, 0), hipErrorInvalidResourceHandle);
  }

  SECTION("Invalid Flags") {
    INFO("Running against Invalid Flags");
    constexpr unsigned flag = ~0u;
    REQUIRE(flag != 0);
    HIP_CHECK_ERROR(hipStreamWaitEvent(stream, event, flag), hipErrorInvalidValue);
  }

  HIP_CHECK(hipEventDestroy(event));

  if (streamType == StreamTestType::CreatedStream) {
    HIP_CHECK(hipStreamDestroy(stream));
  }
}

 /* Test removed for Nvidia devices because it returns unexpected error */
#if !HT_NVIDIA
TEST_CASE("Unit_hipStreamWaitEvent_UninitializedStream_Negative") {
  hipStream_t stream{reinterpret_cast<hipStream_t>(0xFFFF)};
  hipEvent_t event{nullptr};

  HIP_CHECK(hipEventCreate(&event));

  HIP_CHECK_ERROR(hipStreamWaitEvent(stream, event, 0), hipErrorContextIsDestroyed);

  HIP_CHECK(hipEventDestroy(event));
}
#endif

// Since we can not use atomic*_system on every gpu, we will use wait based on clock rate.
// This wont be accurate since clock rate of a GPU varies depending on many variables including
// thermals, load, utilization
__global__ void waitKernel(int clockRate, int seconds) {
  auto start = clock();
  auto ms = seconds * 1000;
  long long waitTill = clockRate * (long long)ms;
  while (1) {
    auto end = clock();
    if ((end - start) > waitTill) {
      return;
    }
  }
}

__global__ void waitKernel_gfx11(int clockRate, int seconds) {
#if HT_AMD
  auto start = wall_clock64();
  auto ms = seconds * 1000;
  long long waitTill = clockRate * (long long)ms;
  while (1) {
    auto end = wall_clock64();
    if ((end - start) > waitTill) {
      return;
    }
  }
#endif
}

TEST_CASE("Unit_hipStreamWaitEvent_Default") {
  hipStream_t stream{nullptr};
  hipEvent_t waitEvent{nullptr};

  HIP_CHECK(hipStreamCreate(&stream));
  HIP_CHECK(hipEventCreate(&waitEvent));

  REQUIRE(stream != nullptr);
  REQUIRE(waitEvent != nullptr);

  int deviceId {};
  HIP_CHECK(hipGetDevice(&deviceId));

  hipDeviceProp_t prop{};
  HIP_CHECK(hipGetDeviceProperties(&prop, deviceId));
  auto clockRate = prop.clockRate;

  auto waitKernel_used = IsGfx11() ? waitKernel_gfx11 : waitKernel;
  waitKernel_used<<<1, 1, 0, stream>>>(clockRate, 2);  // Wait for 2 seconds

  HIP_CHECK(hipEventRecord(waitEvent, stream));

  // Make sure stream is waiting for data to be set
  HIP_CHECK_ERROR(hipEventQuery(waitEvent), hipErrorNotReady);

  HIP_CHECK(hipStreamWaitEvent(stream, waitEvent, 0));

  HIP_CHECK(hipStreamSynchronize(stream));

  HIP_CHECK(hipStreamDestroy(stream));
  HIP_CHECK(hipEventDestroy(waitEvent));
}

TEST_CASE("Unit_hipStreamWaitEvent_DifferentStreams") {
  hipStream_t blockedStreamA{nullptr}, streamBlockedOnStreamA{nullptr}, unblockingStream{nullptr};
  hipEvent_t waitEvent{nullptr};

  HIP_CHECK(hipStreamCreate(&blockedStreamA));
  HIP_CHECK(hipStreamCreate(&streamBlockedOnStreamA));
  HIP_CHECK(hipStreamCreate(&unblockingStream));
  HIP_CHECK(hipEventCreate(&waitEvent));

  REQUIRE(blockedStreamA != nullptr);
  REQUIRE(streamBlockedOnStreamA != nullptr);
  REQUIRE(waitEvent != nullptr);

  int deviceId {};
  HIP_CHECK(hipGetDevice(&deviceId));

  hipDeviceProp_t prop{};
  HIP_CHECK(hipGetDeviceProperties(&prop, deviceId));
  auto clockRate = prop.clockRate;
  auto waitKernel_used = IsGfx11() ? waitKernel_gfx11 : waitKernel;
  waitKernel_used<<<1, 1, 0, blockedStreamA>>>(clockRate,
                                          3);  // wait for 3 seconds
  HIP_CHECK(hipEventRecord(waitEvent, blockedStreamA));

  // Make sure stream is waiting for data to be set
  HIP_CHECK_ERROR(hipEventQuery(waitEvent), hipErrorNotReady);

  HIP_CHECK(hipStreamWaitEvent(streamBlockedOnStreamA, waitEvent, 0));

  waitKernel_used<<<1, 1, 0, streamBlockedOnStreamA>>>(clockRate, 2);  // Wait for 2 seconds

  HIP_CHECK(hipStreamSynchronize(unblockingStream));

  HIP_CHECK(hipStreamSynchronize(blockedStreamA));

  // Make sure streamBlockedOnStreamA waited for event on blockedStreamA
  HIP_CHECK_ERROR(hipStreamQuery(streamBlockedOnStreamA), hipErrorNotReady);
  HIP_CHECK(hipStreamSynchronize(streamBlockedOnStreamA));

  // Check that both streams have finished
  HIP_CHECK(hipStreamQuery(blockedStreamA));
  HIP_CHECK(hipStreamQuery(streamBlockedOnStreamA));

  HIP_CHECK(hipStreamDestroy(blockedStreamA));
  HIP_CHECK(hipStreamDestroy(streamBlockedOnStreamA));
  HIP_CHECK(hipStreamDestroy(unblockingStream));
  HIP_CHECK(hipEventDestroy(waitEvent));
}