Can LehmannMay 31
John Regehr

RE: https://mastodon.social/@cfbolz/116670458563452360

if it's physically possible to do the same thing in different, incompatible ways, then you can trust programmers to do exactly that (here, representing "known bits" dataflow facts)

May 31 at 6:22pmWeb
Show threadLorin Hochstein May 31
@regehr @cfbolz The great thing about standards is that there are so many to choose from!
Show threadDougMerritt (log😅 = 💧log😄)May 31

@norootcause @regehr @cfbolz
It's not that programmers are particularly bad. After all, if you have non-technical people do such things, they will totally miss some of the ways of doing equivalent but incompatible things.

Gotta have them all!

Show threadCan LehmannMay 31
@regehr Would be interesting to see how the performance differs between them. Our Simplify pass which mainly does knownbits & usedbits actually has a non-negligible compile-time performance impact.
Show threadCan LehmannMay 31
@regehr Though I would personally expect instruction traversal, dispatch etc to dominate, but 🤷‍♂️
Show threaddukJun 1
@regehr what on god's green earth is qemu doing
Show threaddukJun 1
@regehr oh that's terrible