]> granicus.if.org Git - llvm/commit
Merging r328884:
authorTom Stellard <tstellar@redhat.com>
Tue, 29 May 2018 22:55:51 +0000 (22:55 +0000)
committerTom Stellard <tstellar@redhat.com>
Tue, 29 May 2018 22:55:51 +0000 (22:55 +0000)
commitcc4798b5bf5bf5e76f39d2b34fdbe2872d18639e
tree77c3f9e4a06551c6149a5ada7ef4cb5c9a91bf44
parenta474ab4a392a097c3fe7d254776d28febee28ab9
Merging r328884:

------------------------------------------------------------------------
r328884 | timshen | 2018-03-30 10:51:00 -0700 (Fri, 30 Mar 2018) | 19 lines

[BlockPlacement] Disable block placement tail duplciation in structured CFG.

Summary:
Tail duplication easily breaks the structure of CFG, e.g. duplicating on
a region entry. If the structure is intended to be preserved, then we
may want to configure tail duplication, or disable it for structured
CFG. From our benchmark results disabling it doesn't cause performance
regression.

Notice that this currently affects AMDGPU backend. In the next patch, I
also plan to turn on requiresStructuredCFG for NVPTX.

All unit tests still pass.

Reviewers: jlebar, arsenm

Subscribers: jholewinski, sanjoy, wdng, tpr, hiraditya, llvm-commits

Differential Revision: https://reviews.llvm.org/D45008
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_60@333474 91177308-0d34-0410-b5e6-96231b3b80d8
lib/CodeGen/MachineBlockPlacement.cpp