maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

Use a latency value of 0 for the artificial edges inserted by 

AddPseudoTwoAddrDeps. This lets the scheduling infrastructure
avoid recalculating node heights. In very large testcases this
was a major bottleneck. Thanks to Roman Levenstein for finding
this!

As a side effect, fold-pcmpeqd-0.ll is now scheduled better
and it no longer requires spilling on x86-32.

llvm-svn: 61778
This commit is contained in:
Dan Gohman 2009-01-06 01:19:04 +00:00
parent 04a2079726
commit 79c3516912
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llvm/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
View File

@ -1252,7 +1252,7 @@ void RegReductionPriorityQueue<SF>::AddPseudoTwoAddrDeps() {
!scheduleDAG->IsReachable(SuccSU, SU)) {
DOUT << "Adding a pseudo-two-addr edge from SU # " << SU->NodeNum
<< " to SU #" << SuccSU->NodeNum << "\n";
scheduleDAG->AddPred(SU, SDep(SuccSU, SDep::Order, /*Latency=*/1,
scheduleDAG->AddPred(SU, SDep(SuccSU, SDep::Order, /*Latency=*/0,
/*Reg=*/0, /*isMustAlias=*/false,
/*isArtificial=*/true));
}

2
llvm/test/CodeGen/X86/fold-pcmpeqd-0.ll
View File

@ -1,4 +1,4 @@
; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin | not grep pcmpeqd
; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin | grep pcmpeqd | count 1
; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin | grep pcmpeqd | count 1
; This testcase shouldn't need to spill the -1 value,