Subject: Re: printf() using floating point ...
To: Bakul Shah <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@lagavulin.pdl.cs.cmu.edu>
Date: 06/14/1995 14:53:28
> I believe the idea was to change the *kernel* to do lazy FP
> saving. How will that change vfprintf?
you have it backward:
the idea was to change vfprintf, because, on some systems that do lazy
FP saves, the current vfprintf causes FP context to be saved/restored
even when no FP is actually _used_.
it's worth noting that on some systems (e.g. most m68k systems), you
CANNOT do lazy FP save and restore...
As i've said before: i'd like to see some numbers that indicate that
this is worthwhile. I think i'd also like it to be reasonably shown
that saving registers when they're never touched isn't a compiler