Commit Graph

48 Commits (e9aa9a52371d6c77ffc567e74581e58b0a79485f)

Author SHA1 Message Date
sfilippone ade79bcc7e Fixes for compilation with CUDA
sfilippone abdf7fc05a Fix constructor name for multivector
sfilippone 3d9fee2dd7 Fix DOT on CUDA vectors.
sfilippone c74be820ea Rework configry for CUDA
sfilippone 4461b44eda Change name abgdxyz into upd_xyz
sfilippone 9f2b8a2623 Cleanup
sfilippone e3a55967a5 Modify CUDA code to compile with 12.4/12.5
sfilippone d71d355b68 Refactor cusparse includes..
sfilippone 2e3f862e42 Start refactoring cusparse.h
sfilippone ee66db5efd Refactor interface to cusparse in preparation for CSR Adaptive
sfilippone c8cc2275d0 Fix cuda/makefile for make -j
sfilippone d01b8145c6 Fix cuda makefile dependencies
sfilippone e18de650f2 Take out debug print
sfilippone 0760e4d553 Fix C function declarations for compilation with LLVM/clang in CUDA
sfilippone 3a25d7b04a Fixes for LLVM compilation
sfilippone e0a4d362fa Define flag TRACK_CUDA_MALLOC
Salvatore Filippone b5f1442ac8 Merge branch 'nond-rep' into repackage
sfilippone 48455190ec Add GPU version of XYZW
sfilippone a11f328e62 Added CUDA version of XYZW
sfilippone b5d5f97661 Improve cuda%zero()
sfilippone 0e269ed641 typo in Cabgdxyz
Salvatore Filippone d95077ffd6 Fix typo in vectordev_mod
Salvatore Filippone 2d3773df98 CUDA kernels for ABGDXYZ
Salvatore Filippone 2a75d677d0 ABGDXYZ in vectordev_mod
sfilippone 2391f64df6 X_cuda_vect%abgdxyz
sfilippone 93c71c4316 Fix %ZERO() on cuda
sfilippone 0568a83734 Fix ifdef and old code
Salvatore Filippone 35d68aa4e3 Reuse calls to getDeviceProperties done at init time
Salvatore Filippone 1ba8dfc7b7 Switch FOR and IF in AXPBY
Salvatore Filippone f9677bc892 Enabled new CUDA version of ABGDXYZ
Salvatore Filippone 4681767ef8 New implementation for ABGDXYZ in CUDA
Salvatore Filippone 105aa3c570 Intermediate impl of ABGDXYZ
Salvatore Filippone 864872ecac Intermediate implementation of abgdxyz on cuda
Salvatore Filippone a41b209144 Better AXPBY implementation in CUDA.
Salvatore Filippone ebc7c6b3b4 Fix call to base%abgdxyz
Salvatore Filippone 14c4ff0f32 Added new methd for two combined axpbys
sfilippone 6433dc797e Fix CUDA implementation of %set_scal and %zero
sfilippone 097d63147a Fix cuda dir makefile
Salvatore Filippone 20a01d4d71 Attempt at fixing CSRG in CUDA 10.2. Not complete yet.
sfilippone 1bc2a884e2 Adjust conditional compilation on CUDA version
Salvatore Filippone 62db7c0449 Fix spsv with CSRG handling of descriptors.
Salvatore Filippone d28ea462d9 Modified CSRG to work with latest versions; cusparse docs are unclear
sfilippone 0230fbb7af Identufied problems with CSRG. Will fix in a branch
sfilippone b2b7b074df Fix usage of HAVE_CUDA/HAVE_GPU (mostly disappeared)
sfilippone 655c86caed Updated docs.
sfilippone 9b713c177b Fix cuda interfaces for renaming
sfilippone 6fa0bf7fe7 Complete cuda renaming
sfilippone 6aa7987d52 Rename GPU into cuda, and merge SPGPU code.