Skip to Content.
Sympa Menu

lapackers - Re: [lapackers] Re: LAPACK debugging

Subject: lapackers

List archive

Re: [lapackers] Re: LAPACK debugging


Chronological Thread 
  • From: Mark Hoemmen <mark.hoemmen@gmail.com>
  • To: Julien Langou <julien.langou@ucdenver.edu>
  • Cc: James Demmel <demmel@cs.berkeley.edu>, Jack Dongarra <dongarra@cs.utk.edu>, Julie Langou <julie@cs.utk.edu>, lapackers <lapackers@cs.berkeley.edu>, "lapackers@cs.utk.edu" <lapackers@cs.utk.edu>, "grevy@cs.berkeley.edu" <grevy@cs.berkeley.edu>, Koushik Sen <ksen@cs.berkeley.edu>, "W. Kahan" <wkahan@EECS.Berkeley.EDU>, David H Bailey <dhbailey@lbl.gov>
  • Subject: Re: [lapackers] Re: LAPACK debugging
  • Date: Thu, 20 May 2010 12:43:56 -0600
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:reply-to:in-reply-to:references:from:date:message-id :subject:to:cc:content-type; b=SG0IUDyOx7umtGvqbPFZ4G7zSelomBoa0xEYIl4ZJuwpGL7/tZPsoMQ//Hk5Hd2aFT sssoVsUzu/Cyku2VyIwiDvfwFImAlUtEIlhGErxw4w8FnXqrxOu72Bjmq369s+3KYi62 h4RIorfhGaw3B9TQkO2CCv+LYfYMJ8U1A+jgo=
On Wed, May 19, 2010 at 20:47, Julien Langou <julien.langou@ucdenver.edu>
wrote:
> Jim, I agree with you. This is a very good solution. I do not think the
> maintenance will be that bad. Julien.
>
> On Wed, 19 May 2010, James Demmel wrote:
>
>> I expect Guillaume to return tomorrow to finish collecting data on the
>> slarfp bug.
>>
>> But I am currently leaning to the following decision, and solicit your
>> comments:
>>
>> (1) Introduce a new QR routine, call it geqrp, with new computational
>> subroutines underneath it as needed to eventually call larfp instead of
>> larfg.
>> ...

As a consumer of nonnegative-diagonal QR factorizations, I would be
happy as long as _LARFP stays in LAPACK. I can and do maintain my own
QR factorizations that call _LARFP, and I don't even particularly need
a BLAS 3 QR for this application (or can write it myself as
necessary).

mfh

Archive powered by MHonArc 2.6.16.

Top of Page