GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-17 22:25:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

Move HIP/CUDA defines to Core.

Browse Source
This commit is contained in:
Antonio Sánchez 2025-06-27 16:48:07 +00:00
parent 26616fe5b8
commit 7c636dd5db
10 changed files with 9 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h → Eigen/src/Core/util/GpuHipCudaDefines.inc
View File

@ -8,7 +8,7 @@
// Public License v. 2.0. If a copy of the MPL was not distributed // 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/. // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#if defined(EIGEN_USE_GPU) && !defined(EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H) #if defined(EIGEN_USE_GPU) && !defined(EIGEN_CORE_GPU_HIP_CUDA_DEFINES_H)
#define EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H #define EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H
// Note that we are using EIGEN_USE_HIP here instead of EIGEN_HIPCC...this is by design // Note that we are using EIGEN_USE_HIP here instead of EIGEN_HIPCC...this is by design
@ -98,4 +98,4 @@
#endif // gpu_assert #endif // gpu_assert
#endif // EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H #endif // EIGEN_CORE_GPU_HIP_CUDA_DEFINES_H

6
unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaUndefines.h → Eigen/src/Core/util/GpuHipCudaUndefines.inc
View File

@ -8,7 +8,7 @@
// Public License v. 2.0. If a copy of the MPL was not distributed // 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/. // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#if defined(EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H) #if defined(EIGEN_CORE_GPU_HIP_CUDA_DEFINES_H)
#ifndef EIGEN_PERMANENTLY_ENABLE_GPU_HIP_CUDA_DEFINES #ifndef EIGEN_PERMANENTLY_ENABLE_GPU_HIP_CUDA_DEFINES
@ -40,6 +40,6 @@
#endif // EIGEN_PERMANENTLY_ENABLE_GPU_HIP_CUDA_DEFINES #endif // EIGEN_PERMANENTLY_ENABLE_GPU_HIP_CUDA_DEFINES
#undef EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H #undef EIGEN_CORE_GPU_HIP_CUDA_DEFINES_H
#endif // EIGEN_CXX11_TENSOR_GPU_HIP_CUDA_DEFINES_H #endif // EIGEN_CORE_GPU_HIP_CUDA_DEFINES_H

2
test/gpu_test_helper.h
View File

@ -4,7 +4,7 @@
#include <Eigen/Core> #include <Eigen/Core>
// Allow gpu** macros for generic tests. // Allow gpu** macros for generic tests.
#include <unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h> #include <Eigen/src/Core/util/GpuHipCudaDefines.inc>
// std::tuple cannot be used on device, and there is a bug in cuda < 9.2 that // std::tuple cannot be used on device, and there is a bug in cuda < 9.2 that
// doesn't allow std::tuple to compile for host code either. In these cases, // doesn't allow std::tuple to compile for host code either. In these cases,

9
unsupported/Eigen/CXX11/src/Tensor/TensorDeviceGpu.h
View File

@ -10,14 +10,11 @@
#if defined(EIGEN_USE_GPU) && !defined(EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H) #if defined(EIGEN_USE_GPU) && !defined(EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H)
#define EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H #define EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H
// This header file container defines fo gpu* macros which will resolve to
// their equivalent hip* or cuda* versions depending on the compiler in use
// A separate header (included at the end of this file) will undefine all
#include "TensorGpuHipCudaDefines.h"
// IWYU pragma: private // IWYU pragma: private
#include "./InternalHeaderCheck.h" #include "./InternalHeaderCheck.h"
#include "../../../../../Eigen/src/Core/util/GpuHipCudaDefines.inc"
namespace Eigen { namespace Eigen {
static const int kGpuScratchSize = 1024; static const int kGpuScratchSize = 1024;
@ -390,6 +387,6 @@ static EIGEN_DEVICE_FUNC inline void setGpuSharedMemConfig(gpuSharedMemConfig co
} // end namespace Eigen } // end namespace Eigen
// undefine all the gpu* macros we defined at the beginning of the file // undefine all the gpu* macros we defined at the beginning of the file
#include "TensorGpuHipCudaUndefines.h" #include "../../../../../Eigen/src/Core/util/GpuHipCudaUndefines.inc"
#endif // EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H #endif // EIGEN_CXX11_TENSOR_TENSOR_DEVICE_GPU_H

2
unsupported/test/cxx11_tensor_argmax_gpu.cu
View File

@ -14,8 +14,6 @@
#include "main.h" #include "main.h"
#include <unsupported/Eigen/CXX11/Tensor> #include <unsupported/Eigen/CXX11/Tensor>
#include <unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
using Eigen::Tensor; using Eigen::Tensor;
template <int Layout> template <int Layout>

2
unsupported/test/cxx11_tensor_contract_gpu.cu
View File

@ -17,8 +17,6 @@
#include "main.h" #include "main.h"
#include <unsupported/Eigen/CXX11/Tensor> #include <unsupported/Eigen/CXX11/Tensor>
#include <unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
using Eigen::Tensor; using Eigen::Tensor;
typedef Tensor<float, 1>::DimensionPair DimPair; typedef Tensor<float, 1>::DimensionPair DimPair;

2
unsupported/test/cxx11_tensor_device.cu
View File

@ -17,8 +17,6 @@
#include "OffByOneScalar.h" #include "OffByOneScalar.h"
#include <unsupported/Eigen/CXX11/Tensor> #include <unsupported/Eigen/CXX11/Tensor>
#include <unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
using Eigen::RowMajor; using Eigen::RowMajor;
using Eigen::Tensor; using Eigen::Tensor;

2
unsupported/test/cxx11_tensor_gpu.cu
View File

@ -15,8 +15,6 @@
#include "main.h" #include "main.h"
#include <unsupported/Eigen/CXX11/Tensor> #include <unsupported/Eigen/CXX11/Tensor>
#include <unsupported/Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
using Eigen::Tensor; using Eigen::Tensor;
void test_gpu_nullary() { void test_gpu_nullary() {

2
unsupported/test/cxx11_tensor_random_gpu.cu
View File

@ -16,8 +16,6 @@
#include "main.h" #include "main.h"
#include <Eigen/CXX11/Tensor> #include <Eigen/CXX11/Tensor>
#include <Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
void test_gpu_random_uniform() { void test_gpu_random_uniform() {
Tensor<float, 2> out(72, 97); Tensor<float, 2> out(72, 97);
out.setZero(); out.setZero();

2
unsupported/test/cxx11_tensor_scan_gpu.cu
View File

@ -16,8 +16,6 @@
#include "main.h" #include "main.h"
#include <unsupported/Eigen/CXX11/Tensor> #include <unsupported/Eigen/CXX11/Tensor>
#include <Eigen/CXX11/src/Tensor/TensorGpuHipCudaDefines.h>
using Eigen::Tensor; using Eigen::Tensor;
typedef Tensor<float, 1>::DimensionPair DimPair; typedef Tensor<float, 1>::DimensionPair DimPair;