On Wed, 2003-11-05 at 04:12, Falk Hueffner wrote: > Phil Carmody <thefatphil@yahoo.co.uk> writes: > > > --- Thomas Evans <tom@23palmer.net> wrote: > > > I really hope that no one is doing severely high > > > performance, mission critical, FP code on pre-EV6 > > > alphas these days, so I really feel that defaulting > > > all Debian built Alpha images to IEEE shouldn't be > > > a big deal. > > > > I do all of my presieving and some of my testing for > > my PIES project on a pre-EV6 alpha. No, it's not > > exactly mission critical, but it _is_ high performance, > > and it is FP code. (And it doesn't depend on IEEE > > nappy-changing as it doesn't shit itself.) > > There was one suggestion to disable -mieee by -ffast-math. -ffast-math > also enables a *lot* of other optimizations, so if you don't want IEEE > compliance, you might want it anyway. Would that be acceptable to you? I'm okay with this sugesstion, as long as Debian on Alpha enables fast math by exception and not as the rule. My concern is also dependencies - people will insist on building glibc, etc with fast math enabled (because by default they'll work fine) and then some other application with throw down IEEE values to them an they'll crash. The application (which functions perfectly on every other architecture in the Debian world) would then be blamed (even though is was built IEEE compliant) and we are back at having random applications crash. BTW, I debugged the kde library that was causing my problem. Sure enough there is a comment in the code that they'll need to deal with -0, NaN and Inf. It's not done yet though - I wonder why? perhaps because everything worked anyway on the platform that the author was testing. ...tom
Attachment:
signature.asc
Description: This is a digitally signed message part