projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 3a90112) | patch
[SE] Pack global dev handle addresses
authorJason Henline <jhen@google.com>
Tue, 13 Sep 2016 23:59:10 +0000 (23:59 +0000)
committerJason Henline <jhen@google.com>
Tue, 13 Sep 2016 23:59:10 +0000 (23:59 +0000)
commitb38d8a3a3baabf759e819fdefd764462691f4048
tree27a2654ade62e9165035879d37817407790a8342tree | snapshot
parent3a90112591ba70d0631572c4cfa4d1fba7858b6acommit | diff
[SE] Pack global dev handle addresses

Summary:
We were packing global device memory handles in
`PackedKernelArgumentArray`, but as I was implementing the CUDA
platform, I realized that CUDA wants the address of the handle, not the
handle itself. So this patch switches to packing the address of the
handle.

Reviewers: jlebar

Subscribers: jprice, jlebar, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D24528

llvm-svn: 281424
parallel-libs/streamexecutor/examples/HostSaxpy.cpp diff | blob | history
parallel-libs/streamexecutor/include/streamexecutor/DeviceMemory.h diff | blob | history
parallel-libs/streamexecutor/include/streamexecutor/PackedKernelArgumentArray.h diff | blob | history
parallel-libs/streamexecutor/unittests/CoreTests/PackedKernelArgumentArrayTest.cpp diff | blob | history
Domain: System / Toolchain;