Complete the MachineScheduler fix made way back in r210390.
authorAndrew Trick <atrick@apple.com>
Fri, 27 Mar 2015 06:10:13 +0000 (06:10 +0000)
committerAndrew Trick <atrick@apple.com>
Fri, 27 Mar 2015 06:10:13 +0000 (06:10 +0000)
commit9217916725713c00f17cb64123e8dffdae843eb7
tree76ef21428425e8503373c96592d1225385da3a66
parent7df7833eaf53a01c96e7b5cd82d1bb413e3d035f
Complete the MachineScheduler fix made way back in r210390.

"Fix the MachineScheduler's logic for updating ready times for in-order.
 Now the scheduler updates a node's ready time as soon as it is
 scheduled, before releasing dependent nodes."

This fix was only made in one variant of the ScheduleDAGMI driver.
Francois de Ferriere reported the issue in the other bit of code where
it was also needed.
I never got around to coming up with a test case, but it's an
obvious fix that shouldn't be delayed any longer.
I'll try to refactor this code a little better.

I did verify performance on a wide variety of targets and saw no
negative impact with this fix.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233366 91177308-0d34-0410-b5e6-96231b3b80d8
lib/CodeGen/MachineScheduler.cpp
test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll
test/CodeGen/AArch64/arm64-convert-v4f64.ll
test/CodeGen/AArch64/sibling-call.ll
test/CodeGen/ARM/saxpy10-a9.ll
test/CodeGen/PowerPC/vsx-fma-m.ll
test/CodeGen/R600/ftrunc.f64.ll
test/CodeGen/R600/local-memory-two-objects.ll