Beware of Fast-Math • Hacker News Board

Re: Beware of Fast-Math

#231

Post by CamouflagedKiwi » Wed, Jun 04, 2025, 6:40 AM UTC

rlpb wrote: post #229

CamouflagedKiwi wrote: post #188

Earlier quoted context omitted.

If they're things that everyone wants, why aren't they in -O1?

-O1 is defined not to perform "any optimizations that take a great deal of compilation time". See the definitions in the manpage.

Yes, exactly my point. Things in -O2 but not -O1 are hence not "things that everyone wants".

Re: Beware of Fast-Math

#232

Post by teleforce » Sat, Jun 07, 2025, 2:28 AM UTC

neepi wrote: post #189

kvemkon wrote: post #101

Earlier quoted context omitted.

I'm wondering, why there are still no announcements for hardware support of such approaches in CPUs.

HP had proper deterministic decimal arithmetic since the 1970s.

Any link to that?

Re: Beware of Fast-Math

#233

Post by kvemkon » Sun, Jun 15, 2025, 1:11 AM UTC

vanderZwan wrote: post #213

kvemkon wrote: post #101

Earlier quoted context omitted.

I'm wondering, why there are still no announcements for hardware support of such approaches in CPUs.

Gustavson's last presentation starts with him holding an actual piece of hardware supporting posits, fwiw. https://m.youtube.com/watch?v=vzVlQhaAZtQ

Wow, great! Thanks for pointing this out.

Just found an announcement:

Calligo secures funds to scale POSIT-based chip development (07.04.2025)

https://www.verdict.co.uk/calligo-secures-funds/