[SE] Host platform implementation
authorJason Henline <jhen@google.com>
Tue, 13 Sep 2016 19:28:02 +0000 (19:28 +0000)
committerJason Henline <jhen@google.com>
Tue, 13 Sep 2016 19:28:02 +0000 (19:28 +0000)
commit3088696499c2d883d8573eb384cd5d2455a3c6c1
treef1313406d5c01531ddb8c733bbd467b3b99af10e
parent1af1414d4d268b5b8580a56101be3e434f8c9291
[SE] Host platform implementation

Summary:
This implementation does not currently support multiple concurrent streams, and
it won't allow kernels to be launched with grids larger than one block or
blocks larger than one thread. These limitations could be removed in the future
by launching new threads on the host, but that is not done in this
implementation.

Reviewers: jlebar

Subscribers: beanz, mgorny, jprice, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D24473

llvm-svn: 281377
parallel-libs/streamexecutor/examples/CMakeLists.txt
parallel-libs/streamexecutor/examples/CUDASaxpy.cpp
parallel-libs/streamexecutor/examples/HostSaxpy.cpp [new file with mode: 0644]
parallel-libs/streamexecutor/include/streamexecutor/KernelSpec.h
parallel-libs/streamexecutor/include/streamexecutor/PlatformDevice.h
parallel-libs/streamexecutor/include/streamexecutor/platforms/host/HostPlatform.h [new file with mode: 0644]
parallel-libs/streamexecutor/include/streamexecutor/platforms/host/HostPlatformDevice.h [new file with mode: 0644]
parallel-libs/streamexecutor/lib/PlatformManager.cpp