devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updates corresponding to the latest round of PR feedback

Browse Source 
The major changes are

1. Moving CUDA/PacketMath.h to GPU/PacketMath.h
2. Moving CUDA/MathFunctions.h to GPU/MathFunction.h
3. Moving CUDA/CudaSpecialFunctions.h to GPU/GpuSpecialFunctions.h
    The above three changes effectively enable the Eigen "Packet" layer for the HIP platform

4. Merging the "hip_basic" and "cuda_basic" unit tests into one ("gpu_basic")
5. Updating the "EIGEN_DEVICE_FUNC" marking in some places

The change has been tested on the HIP and CUDA platforms.
This commit is contained in:
Deven Desai
2018-07-11 10:39:54 -04:00
parent 1fe0b74904
commit 876f392c39
18 changed files with 158 additions and 285 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h
View File

@@ -28,6 +28,20 @@ class TensorContractionBlocking {
typedef typename LhsMapper::Scalar LhsScalar;
typedef typename RhsMapper::Scalar RhsScalar;
/*
adding EIGEN_DEVICE_FUNC unconditionally to 'TensorContractionBlocking' constructor in `TensorContractionBlocking.h`
requires adding EIGEN_DEVICE_FUNC to `computeProductBlockingSizes` in `GeneralBlockPanelKernel.h`
which in turn, requires adding EIGEN_DEVICE_FUNC to `evaluateProductBlockingSizesHeuristic` in `GeneralBlockPanelKernel.h`
which in turn, requires adding EIGEN_DEVICE_FUNC to `manage_caching_sizes` in `GeneralBlockPanelKernel.h`
(else HIPCC will error out)
However adding EIGEN_DEVICE_FUNC to `manage_caching_sizes` in `GeneralBlockPanelKernel.h`
results in NVCC erroring out with the following error
../Eigen/src/Core/products/GeneralBlockPanelKernel.h(57): error #2901:
dynamic initialization is not supported for function-scope static variables within a __device__/__global__ function
*/
#if !defined(EIGEN_HIPCC)
EIGEN_DEVICE_FUNC
#endif