Displaying 2 results from an estimated 2 matches for "rl316295".
2018 Dec 05
3
Strange regalloc behaviour: one more available register causes much worse allocation
...:niravd at google.com]
Sent: 05 December 2018 17:14
To: John Brawn
Cc: llvm-dev; nd
Subject: Re: [llvm-dev] Strange regalloc behaviour: one more available register causes much worse allocation
This has cropped up before in X86 (https://bugs.llvm.org/show_bug.cgi?id=26810 / https://reviews.llvm.org/rL316295), and there's at least a partial mitigation
(I recently ran into an eviction change on X86 when trying variants of a MachineScheduler change, but couldn't find a reproduction post the landed patch).
I suggest you try enabling enableAdvancedRASplitCost() for ARM and seeing if that helps.
-...
2018 Dec 05
2
Strange regalloc behaviour: one more available register causes much worse allocation
Preamble
--------
While working on an IR-level optimisation completely unrelated to register
allocation I happened to trigger some really strange register allocator
behaviour causing a large regression in bzip2 in spec2006. I've been trying
to fix that regression before getting the optimisation patch committed, because
I don't want to regress spec2006, but I'm basically fumbling in