On 08/12/2016 03:13 AM, Bin Cheng wrote:
Hi,
This is a simple refactoring patch which can help long compile time issue
reported as PR72855.
Basically, it moves cheaper checks before expensive ones. By doing this,
doloop -optimization
fails on cheap checks, thus following expensive data-flow analysis is skipped.
Bill helped bootstrap/test on powerpc64le-unknown-linux-gnu. Is it OK?
Thanks,
bin
2016-08-12 Bin Cheng <bin.ch...@arm.com>
PR rtl-optimization/72855
* loop-doloop.c (doloop_insn_valid_p): New function.
(doloop_valid_p): Rename to...
(doloop_niter_valid_p): ...this. Factor out insn check to function
doloop_insn_valid_p.
(doloop_optimize): Update uses.
Presumably it's the call into the IV code that you're trying to avoid,
right?
I'd issue distinct messages for the !doloop_insn_valid_p and
!doloop_niter_valid_p cases. Then you can include the testcase and
check for which of the two exit paths you took from doloop_optimize to
ensure you're getting the early exit you want.
OK with that change.
jeff
