RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-12-23 03:59:12 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Sorting is now handled by both linearscan and iterative scan so live

Browse Source 
intervals need not be sorted anymore. Removing this redundant step
improves LiveIntervals running time by 5% on 176.gcc.

llvm-svn: 15106
This commit is contained in:
Alkis Evlogimenos 2004-07-22 15:18:10 +00:00
parent 717d6cd988
commit ba7fdf3568
1 changed files with 0 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
lib/CodeGen/LiveIntervals.cpp
View File

@ -159,7 +159,6 @@ bool LiveIntervals::runOnMachineFunction(MachineFunction &fn) {
}
}
intervals_.sort();
DEBUG(std::cerr << "********** INTERVALS **********\n");
DEBUG(std::copy(intervals_.begin(), intervals_.end(),
std::ostream_iterator<LiveInterval>(std::cerr, "\n")));
@ -263,15 +262,6 @@ std::vector<LiveInterval*> LiveIntervals::addIntervalsForSpills(
}
}
// FIXME: This method MUST return intervals in sorted order. If a
// particular machine instruction both uses and defines the vreg being
// spilled (e.g., vr = vr + 1) and if the def is processed before the
// use, the list ends up not sorted.
//
// The proper way to fix this is to process all uses of the vreg before we
// process any defs. However, this would require refactoring the above
// blob of code, which I'm not feeling up to right now.
std::sort(added.begin(), added.end(), less_ptr<LiveInterval>());
return added;
}