Danylo Lytovchenko
203 行
7.2 KiB
C++
203 行
7.2 KiB
C++
/*
|
|
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.
|
|
*/
|
|
|
|
#include "memcpy_performance_common.hh"
|
|
|
|
/**
|
|
* @addtogroup memcpy memcpy
|
|
* @{
|
|
* @ingroup PerformanceTest
|
|
*/
|
|
|
|
class MemcpyAsyncBenchmark : public Benchmark<MemcpyAsyncBenchmark> {
|
|
public:
|
|
void operator()(void* dst, const void* src, size_t size, hipMemcpyKind kind,
|
|
const hipStream_t& stream) {
|
|
TIMED_SECTION_STREAM(kTimerTypeEvent, stream) {
|
|
HIP_CHECK(hipMemcpyAsync(dst, src, size, kind, stream));
|
|
}
|
|
HIP_CHECK(hipStreamSynchronize(stream));
|
|
}
|
|
};
|
|
|
|
static void RunBenchmark(LinearAllocs dst_allocation_type, LinearAllocs src_allocation_type,
|
|
size_t size, hipMemcpyKind kind, bool enable_peer_access = false) {
|
|
MemcpyAsyncBenchmark benchmark;
|
|
benchmark.AddSectionName(std::to_string(size));
|
|
benchmark.AddSectionName(GetAllocationSectionName(src_allocation_type));
|
|
benchmark.AddSectionName(GetAllocationSectionName(dst_allocation_type));
|
|
|
|
const StreamGuard stream_guard{Streams::created};
|
|
const hipStream_t stream = stream_guard.stream();
|
|
if (kind != hipMemcpyDeviceToDevice) {
|
|
LinearAllocGuard<int> src_allocation(src_allocation_type, size);
|
|
LinearAllocGuard<int> dst_allocation(dst_allocation_type, size);
|
|
benchmark.Run(dst_allocation.ptr(), src_allocation.ptr(), size, kind, stream);
|
|
} else {
|
|
int src_device = std::get<0>(GetDeviceIds(enable_peer_access));
|
|
int dst_device = std::get<1>(GetDeviceIds(enable_peer_access));
|
|
if (src_device == -1 && dst_device == -1) {
|
|
return;
|
|
}
|
|
|
|
LinearAllocGuard<int> src_allocation(src_allocation_type, size);
|
|
HIP_CHECK(hipSetDevice(dst_device));
|
|
LinearAllocGuard<int> dst_allocation(dst_allocation_type, size);
|
|
HIP_CHECK(hipSetDevice(src_device));
|
|
benchmark.Run(dst_allocation.ptr(), src_allocation.ptr(), size, kind, stream);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Test Description
|
|
* ------------------------
|
|
* - Executes `hipMemcpyAsync` from Device to Host:
|
|
* -# Allocation size
|
|
* - Small: 4 KB
|
|
* - Medium: 4 MB
|
|
* - Large: 16 MB
|
|
* -# Allocation type
|
|
* - Source: device malloc
|
|
* - Destination: host pinned and pageable
|
|
* Test source
|
|
* ------------------------
|
|
* - performance/memcpy/hipMemcpyAsync.cc
|
|
* Test requirements
|
|
* ------------------------
|
|
* - HIP_VERSION >= 5.2
|
|
*/
|
|
TEST_CASE("Performance_hipMemcpyAsync_DeviceToHost") {
|
|
const auto allocation_size = GENERATE(4_KB, 4_MB, 16_MB);
|
|
const auto src_allocation_type = LinearAllocs::hipMalloc;
|
|
const auto dst_allocation_type = GENERATE(LinearAllocs::malloc, LinearAllocs::hipHostMalloc);
|
|
RunBenchmark(dst_allocation_type, src_allocation_type, allocation_size, hipMemcpyDeviceToHost);
|
|
}
|
|
|
|
/**
|
|
* Test Description
|
|
* ------------------------
|
|
* - Executes `hipMemcpyAsync` from Host to Device:
|
|
* -# Allocation size
|
|
* - Small: 4 KB
|
|
* - Medium: 4 MB
|
|
* - Large: 16 MB
|
|
* -# Allocation type
|
|
* - Source: host pinned and pageable
|
|
* - Destination: device malloc
|
|
* Test source
|
|
* ------------------------
|
|
* - performance/memcpy/hipMemcpyAsync.cc
|
|
* Test requirements
|
|
* ------------------------
|
|
* - HIP_VERSION >= 5.2
|
|
*/
|
|
TEST_CASE("Performance_hipMemcpyAsync_HostToDevice") {
|
|
const auto allocation_size = GENERATE(4_KB, 4_MB, 16_MB);
|
|
const auto src_allocation_type = GENERATE(LinearAllocs::malloc, LinearAllocs::hipHostMalloc);
|
|
const auto dst_allocation_type = LinearAllocs::hipMalloc;
|
|
RunBenchmark(dst_allocation_type, src_allocation_type, allocation_size, hipMemcpyHostToDevice);
|
|
}
|
|
|
|
/**
|
|
* Test Description
|
|
* ------------------------
|
|
* - Executes `hipMemcpyAsync` from Host to Host:
|
|
* -# Allocation size
|
|
* - Small: 4 KB
|
|
* - Medium: 4 MB
|
|
* - Large: 16 MB
|
|
* -# Allocation type
|
|
* - Source: host pinned and pageable
|
|
* - Destination: host pinned and pageable
|
|
* Test source
|
|
* ------------------------
|
|
* - performance/memcpy/hipMemcpyAsync.cc
|
|
* Test requirements
|
|
* ------------------------
|
|
* - HIP_VERSION >= 5.2
|
|
*/
|
|
TEST_CASE("Performance_hipMemcpyAsync_HostToHost") {
|
|
const auto allocation_size = GENERATE(4_KB, 4_MB, 16_MB);
|
|
const auto src_allocation_type = GENERATE(LinearAllocs::malloc, LinearAllocs::hipHostMalloc);
|
|
const auto dst_allocation_type = GENERATE(LinearAllocs::malloc, LinearAllocs::hipHostMalloc);
|
|
RunBenchmark(dst_allocation_type, src_allocation_type, allocation_size, hipMemcpyHostToHost);
|
|
}
|
|
|
|
/**
|
|
* Test Description
|
|
* ------------------------
|
|
* - Executes `hipMemcpyAsync` from Device to Device with peer access disabled:
|
|
* -# Allocation size
|
|
* - Small: 4 KB
|
|
* - Medium: 4 MB
|
|
* - Large: 16 MB
|
|
* -# Allocation type
|
|
* - Source: device malloc
|
|
* - Destination: device malloc
|
|
* Test source
|
|
* ------------------------
|
|
* - performance/memcpy/hipMemcpyAsync.cc
|
|
* Test requirements
|
|
* ------------------------
|
|
* - HIP_VERSION >= 5.2
|
|
*/
|
|
TEST_CASE("Performance_hipMemcpyAsync_DeviceToDevice_DisablePeerAccess") {
|
|
const auto allocation_size = GENERATE(4_KB, 4_MB, 16_MB);
|
|
const auto src_allocation_type = LinearAllocs::hipMalloc;
|
|
const auto dst_allocation_type = LinearAllocs::hipMalloc;
|
|
RunBenchmark(dst_allocation_type, src_allocation_type, allocation_size, hipMemcpyDeviceToDevice);
|
|
}
|
|
|
|
/**
|
|
* Test Description
|
|
* ------------------------
|
|
* - Executes `hipMemcpyAsync` from Device to Device with peer access enabled:
|
|
* -# Allocation size
|
|
* - Small: 4 KB
|
|
* - Medium: 4 MB
|
|
* - Large: 16 MB
|
|
* -# Allocation type
|
|
* - Source: device malloc
|
|
* - Destination: device malloc
|
|
* Test source
|
|
* ------------------------
|
|
* - performance/memcpy/hipMemcpyAsync.cc
|
|
* Test requirements
|
|
* ------------------------
|
|
* - Multi-device
|
|
* - Device supports Peer-to-Peer access
|
|
* - HIP_VERSION >= 5.2
|
|
*/
|
|
TEST_CASE("Performance_hipMemcpyAsync_DeviceToDevice_EnablePeerAccess") {
|
|
if (HipTest::getDeviceCount() < 2) {
|
|
HipTest::HIP_SKIP_TEST("This test requires 2 GPUs. Skipping.");
|
|
return;
|
|
}
|
|
const auto allocation_size = GENERATE(4_KB, 4_MB, 16_MB);
|
|
const auto src_allocation_type = LinearAllocs::hipMalloc;
|
|
const auto dst_allocation_type = LinearAllocs::hipMalloc;
|
|
RunBenchmark(dst_allocation_type, src_allocation_type, allocation_size, hipMemcpyDeviceToDevice,
|
|
true);
|
|
}
|
|
|
|
/**
|
|
* End doxygen group memcpy.
|
|
* @}
|
|
*/
|