[PD] denormals
Matteo Sisti Sette
matteo.sistisette at email.it
Thu Jun 7 00:33:16 CEST 2007
Hi,
I may be missing something, but...
I don't think it should be any surprise that intels and other processors
"still" have the denormal problem. AFAIU, a processor may handle denormals
in a more or less efficient way, but it will always be quite much slower
when handling denormals than when handling normal numbers, and it can't be
expected to avoid or eliminate them, since in applications other than DSP
they may be useful (otherwise they wouldn't exsist at all).
We shouldn't wait for the cpu manufacturers to "solve" this problem: it is
the DSP application's responsibility to avoid denormals (as other dsp
applications do), by using some technique such as those described in the
references you have already quoted.
And when I say "the DSP application", I mean in this case PD, since I think
it is a low-level issue that shouldn't be addressed at the patch level.
By browsing the archives I saw that some general solutions were proposed
years ago which allowed to make PD denormal-safe. I don't have the
knowledge to understand them or their related problems in depth, so forgive
me if I ask a stupid question:
Why didn't those lines of code become part of PD??
Didn't they really solve the problem? Or was there any drawback?
Or is there any good reason for not getting rid of denormals?
Thanks
m.
--
Email.it, the professional e-mail, gratis per te: http://www.email.it/f
Sponsor:
Non perderti nella giungla di facili promesse, Logos ti da credito sempre!
*
Clicca qui: http://adv.email.it/cgi-bin/foclick.cgi?mid=6559&d=7-6
More information about the Pd-list
mailing list