llvm are you seriously declaring that not all float bit patterns are actually valid... &$?! Signaling NaNs
-
-
Replying to @Gankro
I thought llvm just straight up didn’t believe in fp exceptions or rounding modes or any of that hard global state stuff
1 reply 0 retweets 3 likes -
Replying to @jckarter
It seems to be fixed now, but basically I've been backtracking some suboptimal codegen problems to a big scare in the Rust issue tracker where the llvm docs suggested `fdiv %x sNaN => undef`
1 reply 0 retweets 4 likes -
That should definitely not be a thing.
1 reply 0 retweets 2 likes -
Replying to @stephentyrone @jckarter
Wait no I found it, but this paragraph makes literally no sense (they claim fdiv by 0 is also UB???)pic.twitter.com/jHm7p8CcH0
2 replies 0 retweets 3 likes -
That’s C semantics, but really shouldn’t be true at the IR level, since it costs us nothing to do the right thing.
2 replies 0 retweets 4 likes
C without Annex F. Annex F defines fdiv by zero.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.