[HN Gopher] GCC 11.1 RC Released, GCC 12 in Development on Trunk
___________________________________________________________________
GCC 11.1 RC Released, GCC 12 in Development on Trunk
Author : rbanffy
Score : 11 points
Date : 2021-04-20 21:01 UTC (2 hours ago)
(HTM) web link (www.phoronix.com)
(TXT) w3m dump (www.phoronix.com)
| ncmncm wrote:
| Recently I learned that it is Gcc policy to reject fixes for
| optimization regressions anywhere but in trunk releases.
|
| So, a regression seen in Gcc 9 and 10, where a switch statement
| is no longer converted, where possible, to a single instruction,
| was fixed in Gcc 11 before its release, but is permanently
| enshrined in Gcc 9.x and 10.x, for all x for all time. The person
| who had introduced the regression argued against fixing it.
|
| I don't know what optimizer regressions are permanently enshrined
| in Gcc 11. I also don't know Clang's policy.
| simcop2387 wrote:
| This likely isn't a policy out of malice or anything like that,
| it's entirely going to be because it's,otherwise difficult or
| impossible to keep any kind if stability and testing with a
| given branch of the compiler. They're just way too cimplicated
| to be able to easily tell and test what a set of changes might
| cause, and because it's udually hard to prove that the change
| you're making isn;t going to have some kind of butterfly effect
| on other changes and optimizations that were made. So in the
| end it's better for stability to fix regressions like that in
| the next release rather than to keep tweaking things in minor
| releases where you're otherwise only commiting to changes for
| correctness and compiler crashes.
___________________________________________________________________
(page generated 2021-04-20 23:02 UTC)