https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #10 from Andrew Pinski ---
(In reply to Martin Liška from comment #1)
> I think it's fixed since r11-2588-gc072fd236dc08f99.
Oh this changed from the shift/xor/sub to using cmov ...
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
Andrew Pinski changed:
What|Removed |Added
Target||x86_64
Keywords|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #8 from Jakub Jelinek ---
This is specific to x86, where if the inputs are inpredictable and results
aren't consumed too early that the cmov latency kills performance cmov
sometimes improves performance a lot, on the other side, if th
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #7 from Roger Sayle ---
I agree in the general case, a conditional jump (that depends only on the
condition flags) potentially has a shorter dependence chain than a cmov (which
depends on the condition flags and two registers). But i
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org
--- Comment #6
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #5 from Dávid Bolvanský ---
User knows the data better, so he/she may prefer abs with branch.
Also PGO may say that branch for abs is better based on profile data.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #4 from Uroš Bizjak ---
Please see PR 56309 (and PR 85559 meta bug).
Quote from Honza:
The decision on whether to use cmov or jmp was always tricky on x86
architectures. Cmov increase dependency chains, register pressure (both value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
--- Comment #3 from Richard Biener ---
A well-predicted branch will be faster than the cmov because of the shorter
data dependence path.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
Roger Sayle changed:
What|Removed |Added
CC||roger at nextmovesoftware dot
com
--- Com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98713
Martin Liška changed:
What|Removed |Added
Ever confirmed|0 |1
Status|UNCONFIRMED
10 matches
Mail list logo