Dear nmusers,
Do you have any recommendations regarding gfortran and Intel Visual Fortran
under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step
successful) comparable?
Many thanks for your response in advance.
Ronald
NONMEM compiler
Dear Ronald
To my understanding the IVF is considered to be somewhat faster than the gfortran compiler in general. But gfortran is on the other hand open source. Regarding "quality" I think it is more dependent upon model than compiler. Here are some non NONMEM related compiler benchmarks http://www.polyhedron.com/compare0html .
Kind Regards
Oskar
--
Oskar Clewe, MSc Pharm
PhD Student
Pharmacometrics Research Group
Department of Pharmacutical Biosciences
Uppsala University
P.O. Box 591
751 24 Uppsala
SWEDEN
[email protected]
Quoted reply history
On 2012-03-14 13:34, Gieschke, Ronald wrote:
> Dear nmusers,
>
> Do you have any recommendations regarding gfortran and Intel Visual Fortran under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step successful) comparable?
>
> Many thanks for your response in advance.
>
> Ronald
Ronald, The web site that Oskar posted seems extremely useful, and confirmed the rumors that have been around for awhile. But, I'm wondering why anyone would use 32 bit, when 64 bit is faster (about 20%) and seems to be better behaved numerically? Mark Mark Sale MD President, Next Level Solutions, LLC www.NextLevelSolns.com 919-846-9185 A carbon-neutral company See our real time solar energy production at: http://enlighten.enphaseenergy.com/public/systems/aSDz2458
Quoted reply history
-------- Original Message --------
Subject: Re: [NMusers] NONMEM compiler
From: Oskar Clewe < [email protected] >
Date: Wed, March 14, 2012 10:16 am
To: "Gieschke, Ronald" < [email protected] >, "'nmusers'"
< [email protected] >
Dear Ronald To my understanding the IVF is considered to be somewhat faster than the gfortran compiler in general. But gfortran is on the other hand open source. Regarding "quality" I think it is more dependent upon model than compiler. Here are some non NONMEM related compiler benchmarks http://www.polyhedron.com/compare0html . Kind Regards Oskar --
Oskar Clewe, MSc Pharm
PhD Student
Pharmacometrics Research Group
Department of Pharmacutical Biosciences
Uppsala University
P.O. Box 591
751 24 Uppsala
SWEDEN
[email protected] On 2012-03-14 13:34, Gieschke, Ronald wrote: Dear nmusers, Do you have any recommendations regarding gfortran and Intel Visual Fortran under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step successful) comparable? Many thanks for your response in advance. v
My concerns with the Polyhedron website’s numbers are the use of gfortran 4.1.2
which was published in February of 2007 and gfortran 4.44 [sic] from April 2010
for the Linux computing environment comparisons. The Intel compiler is always
from August 2011, which would align it with gfortran 4.6.1.
As noted though, the transition from 32-bit to 64-bit is a much bigger impact
on runtime than Intel v gfortran.
Jeb Adams, Research Operations Manager
PKDM IS Liaison
~~~~~~~~~~~~~~~~~~
Amgen, Inc.
Quoted reply history
From: [email protected] [mailto:[email protected]] On
Behalf Of Mark Sale - Next Level Solutions
Sent: Thursday, March 15, 2012 4:31 AM
To: 'nmusers'
Subject: RE: [NMusers] NONMEM compiler
Ronald,
The web site that Oskar posted seems extremely useful, and confirmed the
rumors that have been around for awhile. But, I'm wondering why anyone would
use 32 bit, when 64 bit is faster (about 20%) and seems to be better behaved
numerically?
Mark
Mark Sale MD
President, Next Level Solutions, LLC
www.NextLevelSolns.com
919-846-9185
A carbon-neutral company
See our real time solar energy production at:
http://enlighten.enphaseenergy.com/public/systems/aSDz2458
-------- Original Message --------
Subject: Re: [NMusers] NONMEM compiler
From: Oskar Clewe <[email protected]>
Date: Wed, March 14, 2012 10:16 am
To: "Gieschke, Ronald" <[email protected]>, "'nmusers'"
<[email protected]>
Dear Ronald
To my understanding the IVF is considered to be somewhat faster than the
gfortran compiler in general. But gfortran is on the other hand open source.
Regarding "quality" I think it is more dependent upon model than compiler. Here
are some non NONMEM related compiler benchmarks
http://www.polyhedron.com/compare0html.
Kind Regards
Oskar
-- Oskar Clewe, MSc Pharm PhD Student Pharmacometrics Research Group Department
of Pharmacutical Biosciences Uppsala University P.O. Box 591 751 24 Uppsala
SWEDEN [email protected]
On 2012-03-14 13:34, Gieschke, Ronald wrote:
Dear nmusers,
Do you have any recommendations regarding gfortran and Intel Visual Fortran
under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step
successful) comparable?
Many thanks for your response in advance.
v
I have not done 32 vs 64 bit comparisons, but on the 32 bit Windows machine, the models with differential equations run much faster (30% or more faster) with Intel than with g77 . This may depend of OC, and I do not have these numbers for Linux
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Quoted reply history
On 3/15/2012 6:47 PM, Adams, Jeb wrote:
> My concerns with the Polyhedron website’s numbers are the use of gfortran 4.1.2
> which was published in February of 2007 and gfortran 4.44 [sic] from April 2010
> for the Linux computing environment comparisons. The Intel compiler is always
> from August 2011, which would align it with gfortran 4.6.1.
>
> As noted though, the transition from 32-bit to 64-bit is a much bigger impact
> on runtime than Intel v gfortran.
>
> Jeb Adams, Research Operations Manager
> PKDM IS Liaison
> ~~~~~~~~~~~~~~~~~~
> Amgen, Inc.
>
> From: [email protected] [mailto:[email protected]] On
> Behalf Of Mark Sale - Next Level Solutions
> Sent: Thursday, March 15, 2012 4:31 AM
> To: 'nmusers'
> Subject: RE: [NMusers] NONMEM compiler
>
> Ronald,
> The web site that Oskar posted seems extremely useful, and confirmed the
> rumors that have been around for awhile. But, I'm wondering why anyone would
> use 32 bit, when 64 bit is faster (about 20%) and seems to be better behaved
> numerically?
>
> Mark
>
> Mark Sale MD
> President, Next Level Solutions, LLC
> www.NextLevelSolns.com
> 919-846-9185
> A carbon-neutral company
> See our real time solar energy production at:
> http://enlighten.enphaseenergy.com/public/systems/aSDz2458
>
> -------- Original Message --------
> Subject: Re: [NMusers] NONMEM compiler
> From: Oskar Clewe<[email protected]>
> Date: Wed, March 14, 2012 10:16 am
> To: "Gieschke, Ronald"<[email protected]>, "'nmusers'"
> <[email protected]>
>
> Dear Ronald
> To my understanding the IVF is considered to be somewhat faster than the gfortran
> compiler in general. But gfortran is on the other hand open source. Regarding
> "quality" I think it is more dependent upon model than compiler. Here are some
> non NONMEM related compiler benchmarks http://www.polyhedron.com/compare0html.
>
> Kind Regards
> Oskar
> -- Oskar Clewe, MSc Pharm PhD Student Pharmacometrics Research Group Department
> of Pharmacutical Biosciences Uppsala University P.O. Box 591 751 24 Uppsala
> SWEDEN [email protected]
>
> On 2012-03-14 13:34, Gieschke, Ronald wrote:
> Dear nmusers,
>
> Do you have any recommendations regarding gfortran and Intel Visual Fortran
> under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step
> successful) comparable?
>
> Many thanks for your response in advance.
>
> v
Hi,
Just to add my 2 cents…
In my opinion should performance measures like these not the primary goal of
the exercise…
The priority for NONMEM as a scientific tool should be consistency.
http://wiki.scinethpc.ca/wiki/images/f/f2/FP_Consistency.pdf
http://wiki.scinethpc.ca/wiki/images/f/f2/FP_Consistency.pdf
in an age of multi processor computing there are –in my opinion- no good
reasons to tinker with the math…
Going from 1 core to 2, 3 or more will improve performance far beyond
optimization.
K. Regards,
Xavier
Quoted reply history
From: [email protected] [mailto:[email protected]] On
Behalf Of Mark Sale - Next Level Solutions
Sent: 15 March 2012 12:31
To: 'nmusers'
Subject: RE: [NMusers] NONMEM compiler
Ronald,
The web site that Oskar posted seems extremely useful, and confirmed the
rumors that have been around for awhile. But, I'm wondering why anyone would
use 32 bit, when 64 bit is faster (about 20%) and seems to be better behaved
numerically?
Mark
Mark Sale MD
President, Next Level Solutions, LLC
www.NextLevelSolns.com http://www.NextLevelSolns.com
919-846-9185
A carbon-neutral company
See our real time solar energy production at:
http://enlighten.enphaseenergy.com/public/systems/aSDz2458
http://enlighten.enphaseenergy.com/public/systems/aSDz2458
-------- Original Message --------
Subject: Re: [NMusers] NONMEM compiler
From: Oskar Clewe <[email protected]
<mailto:[email protected]> >
Date: Wed, March 14, 2012 10:16 am
To: "Gieschke, Ronald" <[email protected]
<mailto:[email protected]> >, "'nmusers'"
<[email protected] <mailto:[email protected]> >
Dear Ronald
To my understanding the IVF is considered to be somewhat faster than
the gfortran compiler in general. But gfortran is on the other hand open
source. Regarding "quality" I think it is more dependent upon model than
compiler. Here are some non NONMEM related compiler benchmarks
http://www.polyhedron.com/compare0html http://www.polyhedron.com/compare0html
.
Kind Regards
Oskar
-- Oskar Clewe, MSc Pharm PhD Student Pharmacometrics Research Group
Department of Pharmacutical Biosciences Uppsala University P.O. Box 591 751 24
Uppsala SWEDEN [email protected] <mailto:[email protected]>
On 2012-03-14 13:34, Gieschke, Ronald wrote:
Dear nmusers,
Do you have any recommendations regarding gfortran and Intel Visual
Fortran under 32 bit Windows 7? Are runtimes and quality (e.g. covariance step
successful) comparable?
Many thanks for your response in advance.
v
<<image001.png>>