DPC++ multiGPU stencil and/or transpose#479

Draft

jeffhammond wants to merge 72 commits intoParRes:mainfrom

jeffhammond:dpcpp-multi-gpu-transpose

Commits on Jul 20, 2020

add stencil DPC++ skeletons
Jeff Hammond
committed
cleanup comments
Jeff Hammond
committed
some fixes - GPU broken
Jeff Hammond
committed
cleanup
Jeff Hammond
committed
workaround Level Zero SPIR detection
Jeff Hammond
committed
some fixes - GPU broken
Jeff Hammond
committed
undo minor mistake
Jeff Hammond
committed
add transpose multi-GPU DPC++ skeleton
Jeff Hammond
committed
WIP
Jeff Hammond
committed
add ignores
Jeff Hammond
committed
fix message
Jeff Hammond
committed

Commits on Jul 21, 2020

add fill, cleanup dead code
Jeff Hammond
committed
start decomposition
Jeff Hammond
committed
device queues class and related methods
Jeff Hammond
committed
use device queues stuff - currently broken
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
working with inlined methods
Jeff Hammond
committed
working with inlined methods
Jeff Hammond
committed
cleaned up
Jeff Hammond
committed
cleaned up
Jeff Hammond
committed
fixed bugs
Jeff Hammond
committed
working but to be replaced
Jeff Hammond
committed
working
Jeff Hammond
committed

Commits on Jul 22, 2020

fix input helper comment
Jeff Hammond
committed
add broadcast and reduce (unused and untested)
Jeff Hammond
committed

Commits on Jul 24, 2020

add stencil DPC++ skeletons
Jeff Hammond
committed
cleanup comments
Jeff Hammond
committed
some fixes - GPU broken
Jeff Hammond
committed
cleanup
Jeff Hammond
committed
workaround Level Zero SPIR detection
Jeff Hammond
committed
some fixes - GPU broken
Jeff Hammond
committed
undo minor mistake
Jeff Hammond
committed
add transpose multi-GPU DPC++ skeleton
Jeff Hammond
committed
WIP
Jeff Hammond
committed
add ignores
Jeff Hammond
committed
fix message
Jeff Hammond
committed
add fill, cleanup dead code
Jeff Hammond
committed
start decomposition
Jeff Hammond
committed
device queues class and related methods
Jeff Hammond
committed
use device queues stuff - currently broken
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
trying to fix
Jeff Hammond
committed
working with inlined methods
Jeff Hammond
committed
working with inlined methods
Jeff Hammond
committed
cleaned up
Jeff Hammond
committed
cleaned up
Jeff Hammond
committed
fixed bugs
Jeff Hammond
committed
working but to be replaced
Jeff Hammond
committed
working
Jeff Hammond
committed
fix input helper comment
Jeff Hammond
committed
add broadcast and reduce (unused and untested)
Jeff Hammond
committed
Merge branch 'dpcpp-multi-gpu-transpose' of https://github.com/jeffhammond/PRK into dpcpp-multi-gpu-transpose
Jeff Hammond
committed
Merge branch 'default' into dpcpp-multi-gpu-transpose
Jeff Hammond
committed
remove unnecessary preprocessor
Jeff Hammond
committed
remove 2D indexing code that won't work with USM
Jeff Hammond
committed
fix a bunch of problems
Jeff Hammond
committed

Commits on Jul 26, 2020

need a unit test for the collectives...
Jeff Hammond
committed
whitespace
Jeff Hammond
committed
hoist invariant; start alltoall
Jeff Hammond
committed
progress on unit test
Jeff Hammond
committed
work around issue with CPU-only in "multigpu" tester
Jeff Hammond
committed
broadcast and reduce tested
Jeff Hammond
committed
there are bugs somewhere
Jeff Hammond
committed

Commits on Sep 6, 2020

merge fix
Jeff Hammond
committed

Commits on Oct 14, 2020

Merge branch 'default' into dpcpp-multi-gpu-transpose
Jeff Hammond
committed

Commits on Oct 26, 2020

Merge branch 'default' into dpcpp-multi-gpu-transpose
Jeff Hammond
committed

Commits on Oct 28, 2020

never commit binaries
Jeff Hammond
committed
never commit binaries
Jeff Hammond
committed
Merge branch 'default' into dpcpp-multi-gpu-transpose
Jeff Hammond
committed