[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Mail archives
Greetings, Clint! OK, if you just replace the
static __inline__ void
in each of the three include files
maa.h dpa.h and ga.h
all should work! Something is getting hosed when the dot product
function is inlined. I may figure out a way to restore the inlining,
but it should be a trivial gain in any case.
R Clint Whaley <email@example.com> writes:
> >Greetings! I'm new to this list, and was wondering if there was an
> >archive setup somewhere.
> I am looking into getting this done, but it's not in place yet. As
> an extremely elegant stop-gap, you can find
> which is my personal mail file regarding all the "developer" style
> ATLAS mails. I show my age by admitting I use old "mail", so you
> can read it with mail -f +atlasD, with your mail path set appropriately.
> You can also read it into MH, but I don't know how since I'm still
> living in the 70's with good old mail . . .
> >Just as an FYI, I think my next attempt at atlas contribution will be
> >double precision level 2 kernels using the PIII prefetch command.
> >Does this sound like a good idea, i.e. consonant with the efforts of
> So, in one of my abortive attempts to get the new developer release out,
> I was attempting to add your stuff to the package, and the GER kernels
> all failed the tests described in atlas_os.ps (you need to apply the
> new makefile given on the developer homepage, though). Now, unfortunately,
> my laptop was between upgrades, the atlas directory I was using was
> screwed up, and so I was far from sure the error was in your kernel.
> Can you run the tests and tell me if your GER kernels indeed pass for
> To answer your question, I am *very* interested to see if you double
> precision can be helped by prefetch, and no one has admitted to working
> on it yet . . .
Camm Maguire firstname.lastname@example.org
"The earth is but one country, and mankind its citizens." -- Baha'u'llah