projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 33f41ff) | patch
AMDGPU: Mark s_barrier as a high latency instruction
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Tue, 8 Sep 2015 19:54:32 +0000 (19:54 +0000)
committerMatt Arsenault <Matthew.Arsenault@amd.com>
Tue, 8 Sep 2015 19:54:32 +0000 (19:54 +0000)
commit55ec9f281ce3348a1a08792d96a34034ab19442d
tree9fd32d19dfbffbdeccc5907753a1a4e30f8ad5b0tree | snapshot
parent33f41ffd1ff15f4e527c829edd72eacb0ae89309commit | diff
AMDGPU: Mark s_barrier as a high latency instruction

These were marked as WriteSALU, which is low latency.
I'm guessing at the value to use, but it should probably
be considered the highest latency instruction.

I'm not sure this has any actual effect since hasSideEffects
probably is preventing any moving of these.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@247060 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/AMDGPU/SIInstructions.td diff | blob | history
lib/Target/AMDGPU/SISchedule.td diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors