On Thu, 15 Jan 2026, David Guillen Fandos wrote: > > > > Is there any document showing the behavior of max.s/min.s of R5900 on > > > > sNaN vs NUM > > > > +0 vs -0 > > > > qNaN vs NUM > > > > sNaN vs qNAN > > > > sNaN vs sNaN > > > > > > Yeah, the manual indicates that NaNs are not properly handled. The R5900 > > > doesn't have proper IEEE754 handling (for all its FP instructions, like > > > "c.eq.s"). > > > > Thanks. I plan to accept this patch, while can I add this statement to > > commit message? > > Awesome! Yeah feel free to add the comment, it's a good idea.
FWIW I think this subtarget ought to default to `-ffinite-math-only' and IIRC `-fno-trapping-math' and fail any attempt to change these defaults. It might be worth checking what non-IEEE-FP targets do here. Maciej
