Blaze CUDA compatibility: adding "IsCUDAAssignable" typetrait and "BLAZE_HOST_DEVICE" macro

#35 Merged at 87fefb4
Repository
JPenuchot
Branch
blaze_cuda
Repository
blaze-lib
Branch
master
Author
  1. Jules Pénuchot
Reviewers
Description

Klaus,

Here’s the list of the commits, however some of them such as the cuBLAS-related commits have been revoked like the ones related to cuBLAS. Now it simply consists in adding the IsCUDAAssignable typetrait and the BLAZE_HOST_DEVICE macro.

  • Added BLAZE_HOST_DEVICE macro
  • Made DVecDVec basic expressions CUDA-compatible
  • Added BLAZE_HOST_DEVICE macros to DVecMapExpr iterators
  • Annotated DVecScalMultExpr for CUDA support, fixed operators (constructor was called without scalar_ being passed)
  • Added BLAZE_HOST_DEVICE macros to functors
  • Added global, host and device CUDA attribute macros, changed filename accordingly
  • Added documentation for CUDA attributes
  • Changed BLAZE_HOST_DEVICE to BLAZE_DEVICE_CALLABLE
  • Changed BLAZE_HOST_DEVICE macros to newer BLAZE_DEVICE_CALLABLE
  • Changed BLAZE_HOST_DEVICE macros to newer BLAZE_DEVICE_CALLABLE
  • Addeed IsCUDAAssignable header to TypeTraits.h's includes
  • Added CUDA entry point to HPX backend
  • Made dense subvector iterator CUDA-compatible
  • Extended expression iterators CUDA functionality
  • Annotated ConstIterator functions with BLAZE_DEVICE_CALLABLE + added include
  • Made DMatMatExpr CUDA-compatible
  • Fixed smpAssign dispatch
  • Added missing include statements for CUDAAttributes
  • Added cuBLAS interface (enabled only if compiled with BLAZE_CUDA_MODE)
  • Fixed DMatMapExpr CUDA compatibility
  • Added CUDA-specific includes, might get removed later
  • Reverted changes for DMatDMatMultExpr
  • Enlarged BLAS for cuBLAS
  • Reverted changes to original Blaze, made it independent from Blaze CUDA again, I might delete that branch of Blaze eventually...
  • Removed CUDA typetraits
  • Copyright notice...
  • Made useAssign_v public
  • Revert "Made useAssign_v public"

    This reverts commit 2fc39f4f8d96abce5f831a27b22fcdf1e8414427.

  • Reintroduced IsCUDAAssignable.h to Blaze

  • IsCUDAAssignable include added to TypeTraits.h
  • Changed CUDAAttributes back to HostDevice
  • Added BLAZE_HOST_DEVICE macro
  • Made DVecDVec basic expressions CUDA-compatible
  • Added BLAZE_HOST_DEVICE macros to DVecMapExpr iterators
  • Annotated DVecScalMultExpr for CUDA support, fixed operators (constructor was called without scalar_ being passed)
  • Added BLAZE_HOST_DEVICE macros to functors
  • Added global, host and device CUDA attribute macros, changed filename accordingly
  • Added documentation for CUDA attributes
  • Changed BLAZE_HOST_DEVICE to BLAZE_DEVICE_CALLABLE
  • Changed BLAZE_HOST_DEVICE macros to newer BLAZE_DEVICE_CALLABLE
  • Changed BLAZE_HOST_DEVICE macros to newer BLAZE_DEVICE_CALLABLE
  • Addeed IsCUDAAssignable header to TypeTraits.h's includes
  • Added CUDA entry point to HPX backend
  • Made dense subvector iterator CUDA-compatible
  • Extended expression iterators CUDA functionality
  • Annotated ConstIterator functions with BLAZE_DEVICE_CALLABLE + added include
  • Made DMatMatExpr CUDA-compatible
  • Fixed smpAssign dispatch
  • Added missing include statements for CUDAAttributes
  • Added cuBLAS interface (enabled only if compiled with BLAZE_CUDA_MODE)
  • Fixed DMatMapExpr CUDA compatibility
  • Added CUDA-specific includes, might get removed later
  • Reverted changes for DMatDMatMultExpr
  • Enlarged BLAS for cuBLAS
  • Reverted changes to original Blaze, made it independent from Blaze CUDA again, I might delete that branch of Blaze eventually...
  • Removed CUDA typetraits
  • Copyright notice...
  • Made useAssign_v public
  • Revert "Made useAssign_v public"

    This reverts commit 2fc39f4f8d96abce5f831a27b22fcdf1e8414427.

  • Reintroduced IsCUDAAssignable.h to Blaze

  • IsCUDAAssignable include added to TypeTraits.h
  • Changed CUDAAttributes back to HostDevice
  • Added Blaze CUDA to the contributions

Comments (0)