mirror of
https://gitlab.com/libeigen/eigen.git
synced 2025-10-12 16:11:29 +08:00
8fbd47052b
This commit enables the use of Eigen on HIP kernels / AMD GPUs. Support has been added along the same lines as what already exists for using Eigen in CUDA kernels / NVidia GPUs. Application code needs to explicitly define EIGEN_USE_HIP when using Eigen in HIP kernels. This is because some of the CUDA headers get picked up by default during Eigen compile (irrespective of whether or not the underlying compiler is CUDACC/NVCC, for e.g. Eigen/src/Core/arch/CUDA/Half.h). In order to maintain this behavior, the EIGEN_USE_HIP macro is used to switch to using the HIP version of those header files (see Eigen/Core and unsupported/Eigen/CXX11/Tensor) Use the "-DEIGEN_TEST_HIP" cmake option to enable the HIP specific unit tests.
86 lines
2.3 KiB
Plaintext
86 lines
2.3 KiB
Plaintext
// This file is part of Eigen, a lightweight C++ template library
|
|
// for linear algebra.
|
|
//
|
|
// Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com>
|
|
//
|
|
// This Source Code Form is subject to the terms of the Mozilla
|
|
// Public License v. 2.0. If a copy of the MPL was not distributed
|
|
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
|
|
|
|
#define EIGEN_TEST_NO_LONGDOUBLE
|
|
#define EIGEN_TEST_NO_COMPLEX
|
|
#define EIGEN_TEST_FUNC cxx11_tensor_random_hip
|
|
#define EIGEN_DEFAULT_DENSE_INDEX_TYPE int
|
|
#define EIGEN_USE_GPU
|
|
|
|
#include "main.h"
|
|
#include <Eigen/CXX11/Tensor>
|
|
|
|
|
|
void test_hip_random_uniform()
|
|
{
|
|
Tensor<float, 2> out(72,97);
|
|
out.setZero();
|
|
|
|
std::size_t out_bytes = out.size() * sizeof(float);
|
|
|
|
float* d_out;
|
|
hipMalloc((void**)(&d_out), out_bytes);
|
|
|
|
Eigen::HipStreamDevice stream;
|
|
Eigen::GpuDevice gpu_device(&stream);
|
|
|
|
Eigen::TensorMap<Eigen::Tensor<float, 2> > gpu_out(d_out, 72,97);
|
|
|
|
gpu_out.device(gpu_device) = gpu_out.random();
|
|
|
|
assert(hipMemcpyAsync(out.data(), d_out, out_bytes, hipMemcpyDeviceToHost, gpu_device.stream()) == hipSuccess);
|
|
assert(hipStreamSynchronize(gpu_device.stream()) == hipSuccess);
|
|
|
|
// For now we just check thes code doesn't crash.
|
|
// TODO: come up with a valid test of randomness
|
|
}
|
|
|
|
|
|
void test_hip_random_normal()
|
|
{
|
|
Tensor<float, 2> out(72,97);
|
|
out.setZero();
|
|
|
|
std::size_t out_bytes = out.size() * sizeof(float);
|
|
|
|
float* d_out;
|
|
hipMalloc((void**)(&d_out), out_bytes);
|
|
|
|
Eigen::HipStreamDevice stream;
|
|
Eigen::GpuDevice gpu_device(&stream);
|
|
|
|
Eigen::TensorMap<Eigen::Tensor<float, 2> > gpu_out(d_out, 72,97);
|
|
|
|
Eigen::internal::NormalRandomGenerator<float> gen(true);
|
|
gpu_out.device(gpu_device) = gpu_out.random(gen);
|
|
|
|
assert(hipMemcpyAsync(out.data(), d_out, out_bytes, hipMemcpyDeviceToHost, gpu_device.stream()) == hipSuccess);
|
|
assert(hipStreamSynchronize(gpu_device.stream()) == hipSuccess);
|
|
}
|
|
|
|
static void test_complex()
|
|
{
|
|
Tensor<std::complex<float>, 1> vec(6);
|
|
vec.setRandom();
|
|
|
|
// Fixme: we should check that the generated numbers follow a uniform
|
|
// distribution instead.
|
|
for (int i = 1; i < 6; ++i) {
|
|
VERIFY_IS_NOT_EQUAL(vec(i), vec(i-1));
|
|
}
|
|
}
|
|
|
|
|
|
void test_cxx11_tensor_random_hip()
|
|
{
|
|
CALL_SUBTEST(test_hip_random_uniform());
|
|
CALL_SUBTEST(test_hip_random_normal());
|
|
CALL_SUBTEST(test_complex());
|
|
}
|