www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - LAPACK/BLAS/SciD Windows

reply dsimcha <dsimcha yahoo.com> writes:
Has anyone been able to get BLAS/LAPACK/SciD working on Windows?  If so, how?
 The furthest I've gotten is building SciD, downloading BLAS and LAPACK
prebuilt binaries from http://icl.cs.utk.edu/lapack-for-windows/, and using
Agner Fog's objconv to convert them from COFF to OMF.  Then, when I try to
build a trivial test program, I get:

Symbol Undefined _dgetri_
Symbol Undefined _dgetrf_

Grepping my lapack binaries indicates that getri and getrf are there, but
_getri_ and _getrf_ are not.  It's an underscore issue.

Can anyone either:

1.  Tell me how to fix this?

2.  Tell me where to get BLAS and LAPACK in OMF format that will "just work"
with DMD?
Feb 09 2011
next sibling parent reply Andrej Mitrovic <andrej.mitrovich gmail.com> writes:
AFAIK objconv has the +nu or a similar switch that adds underscores to
symbol names. Hope that helps.
Feb 09 2011
parent dsimcha <dsimcha yahoo.com> writes:
Should have mentioned in my original post, I tried that and it didn't 
work.  I have no idea why.

On 2/9/2011 5:41 PM, Andrej Mitrovic wrote:
 AFAIK objconv has the +nu or a similar switch that adds underscores to
 symbol names. Hope that helps.

Feb 09 2011
prev sibling next sibling parent Andrej Mitrovic <andrej.mitrovich gmail.com> writes:
That's -nu+ actually, or alternatively -au+ which just adds aliases to
existing symbols.

On 2/9/11, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:
 AFAIK objconv has the +nu or a similar switch that adds underscores to
 symbol names. Hope that helps.

Feb 09 2011
prev sibling parent reply "Lars T. Kyllingstad" <public kyllingen.NOSPAMnet> writes:
On Wed, 09 Feb 2011 22:30:23 +0000, dsimcha wrote:

 Has anyone been able to get BLAS/LAPACK/SciD working on Windows?  If so,
 how?
  The furthest I've gotten is building SciD, downloading BLAS and LAPACK
 prebuilt binaries from http://icl.cs.utk.edu/lapack-for-windows/, and
 using Agner Fog's objconv to convert them from COFF to OMF.  Then, when
 I try to build a trivial test program, I get:
 
 Symbol Undefined _dgetri_
 Symbol Undefined _dgetrf_
 
 Grepping my lapack binaries indicates that getri and getrf are there,
 but _getri_ and _getrf_ are not.  It's an underscore issue.
 
 Can anyone either:
 
 1.  Tell me how to fix this?
 
 2.  Tell me where to get BLAS and LAPACK in OMF format that will "just
 work" with DMD?

Have you tried asking Bill Baxter? He wrote the original BLAS/LAPACK bindings, so he should know. -Lars
Feb 10 2011
next sibling parent reply %fil <fil learningd.com> writes:
Lars, David,

Did you managed to get in touch with Bill for this? If so, would you be able to
provide a package with SciD including the BLAS/LAPACK in a DMD compatible
format?
I would also be interested in this...

Are there any plans to bring more of these numerical libraries (like SciD or
even
basic linear alg) into Phobos? I would assume this would be very welcomed by a
lot
of D developers, no?

Btw, Lars, many thx for your work on SciD....

fil
Feb 12 2011
parent %fil <fil somewhere.com> writes:
Caligo Wrote:
 
 How about Eigen?
 http://eigen.tuxfamily.org/index.php?title=Main_Page
 
 I've used it extensively in my C++ projects, and I'm very pleased.
 
 Last time I counted, Eigen was about 10-15k lines of C++ code.  It's all
 templates, and it might be the perfect project to translate to D to show off
 some of D's meta and template features.  I only say this because I really
 miss not having a linear algebra library in D, and I dislike language
 bindings.
 
 I'm still learning D from TDPL, but I would really like to see some sample
 code in D for doing expression templates.  For example, could anyone
 translate this to D? http://en.wikipedia.org/wiki/Expression_templates  It
 took me about half an hour to understand the code because it's kind of hard
 to read.  I wonder if D's version would be more readable.
 

I've never used eigen, only boost ublas. but I've worked mostly in c# for the last couple of years...but I like that it is fully independent.... I'm still learning D (driven by the need to go lower then c#, but at the same time trying to avoid c++ as much as possible), but would not mind trying to work on a translation of eigen to D with the help of others as it would solve the issues of relying on LAPACK or BLAS.... Any other volunteers?
Feb 15 2011
prev sibling next sibling parent dsimcha <dsimcha yahoo.com> writes:
Good hint.  I found some pre-built binaries at 
http://dsource.org/projects/multiarray , Bill Baxter's old project.  I 
tested them (superficially) and they WORK on Windows w/ DMD 2.052 beta 
and SciD SVN.  Lars, please post this zip file somewhere on your SciD 
site so that other Windows users who want to use SciD don't have to got 
through the same hassle I did to track them down.

On 2/10/2011 7:18 AM, Lars T. Kyllingstad wrote:
 On Wed, 09 Feb 2011 22:30:23 +0000, dsimcha wrote:

 Has anyone been able to get BLAS/LAPACK/SciD working on Windows?  If so,
 how?
   The furthest I've gotten is building SciD, downloading BLAS and LAPACK
 prebuilt binaries from http://icl.cs.utk.edu/lapack-for-windows/, and
 using Agner Fog's objconv to convert them from COFF to OMF.  Then, when
 I try to build a trivial test program, I get:

 Symbol Undefined _dgetri_
 Symbol Undefined _dgetrf_

 Grepping my lapack binaries indicates that getri and getrf are there,
 but _getri_ and _getrf_ are not.  It's an underscore issue.

 Can anyone either:

 1.  Tell me how to fix this?

 2.  Tell me where to get BLAS and LAPACK in OMF format that will "just
 work" with DMD?

Have you tried asking Bill Baxter? He wrote the original BLAS/LAPACK bindings, so he should know. -Lars

Feb 13 2011
prev sibling next sibling parent "Lars T. Kyllingstad" <public kyllingen.NOSPAMnet> writes:
On Mon, 14 Feb 2011 01:13:51 -0500, dsimcha wrote:

 Good hint.  I found some pre-built binaries at
 http://dsource.org/projects/multiarray , Bill Baxter's old project.  I
 tested them (superficially) and they WORK on Windows w/ DMD 2.052 beta
 and SciD SVN.  Lars, please post this zip file somewhere on your SciD
 site so that other Windows users who want to use SciD don't have to got
 through the same hassle I did to track them down.

Cool! I will do that. :) -Lars
Feb 14 2011
prev sibling next sibling parent "Lars T. Kyllingstad" <public kyllingen.NOSPAMnet> writes:
On Sat, 12 Feb 2011 21:05:37 +0000, %fil wrote:

 Lars, David,
 
 Did you managed to get in touch with Bill for this? If so, would you be
 able to provide a package with SciD including the BLAS/LAPACK in a DMD
 compatible format? I would also be interested in this...

I am thinking of cleaning up some small things here and there and making a 0.1 "release" in the form of an easily downloadable zip file. It will of course be a rather arbitrary development snapshot, and in no way a "finished" product, but it may make things easier for potential users. I'm not sure I want to include someone else's prebuilt libraries, but I'll be sure to add a link to the BLAS/LAPACK download David mentioned.
 Are there any plans to bring more of these numerical libraries (like
 SciD or even basic linear alg) into Phobos? I would assume this would be
 very welcomed by a lot of D developers, no?

IMO, there should definitely be some basic linear algebra functionality in Phobos.
 Btw, Lars, many thx for your work on SciD....

You're welcome. :) -Lars
Feb 14 2011
prev sibling parent Caligo <iteronvexor gmail.com> writes:
--0023547c95cb283677049c41bcaf
Content-Type: text/plain; charset=ISO-8859-1

On Mon, Feb 14, 2011 at 7:30 AM, Lars T. Kyllingstad
<public kyllingen.nospamnet> wrote:

 On Sat, 12 Feb 2011 21:05:37 +0000, %fil wrote:

 Lars, David,

 Did you managed to get in touch with Bill for this? If so, would you be
 able to provide a package with SciD including the BLAS/LAPACK in a DMD
 compatible format? I would also be interested in this...

I am thinking of cleaning up some small things here and there and making a 0.1 "release" in the form of an easily downloadable zip file. It will of course be a rather arbitrary development snapshot, and in no way a "finished" product, but it may make things easier for potential users. I'm not sure I want to include someone else's prebuilt libraries, but I'll be sure to add a link to the BLAS/LAPACK download David mentioned.
 Are there any plans to bring more of these numerical libraries (like
 SciD or even basic linear alg) into Phobos? I would assume this would be
 very welcomed by a lot of D developers, no?

IMO, there should definitely be some basic linear algebra functionality in Phobos.
 Btw, Lars, many thx for your work on SciD....

You're welcome. :) -Lars

How about Eigen? http://eigen.tuxfamily.org/index.php?title=Main_Page I've used it extensively in my C++ projects, and I'm very pleased. Last time I counted, Eigen was about 10-15k lines of C++ code. It's all templates, and it might be the perfect project to translate to D to show off some of D's meta and template features. I only say this because I really miss not having a linear algebra library in D, and I dislike language bindings. I'm still learning D from TDPL, but I would really like to see some sample code in D for doing expression templates. For example, could anyone translate this to D? http://en.wikipedia.org/wiki/Expression_templates It took me about half an hour to understand the code because it's kind of hard to read. I wonder if D's version would be more readable. --0023547c95cb283677049c41bcaf Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable <br><br><div class=3D"gmail_quote">On Mon, Feb 14, 2011 at 7:30 AM, Lars T.= Kyllingstad <span dir=3D"ltr">&lt;public kyllingen.nospamnet&gt;</span> wr= ote:<br><blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border= -left:1px #ccc solid;padding-left:1ex;"> <div class=3D"im">On Sat, 12 Feb 2011 21:05:37 +0000, %fil wrote:<br> <br> &gt; Lars, David,<br> &gt;<br> &gt; Did you managed to get in touch with Bill for this? If so, would you b= e<br> &gt; able to provide a package with SciD including the BLAS/LAPACK in a DMD= <br> &gt; compatible format? I would also be interested in this...<br> <br> </div>I am thinking of cleaning up some small things here and there and mak= ing<br> a 0.1 &quot;release&quot; in the form of an easily downloadable zip file. = =A0It will<br> of course be a rather arbitrary development snapshot, and in no way a<br> &quot;finished&quot; product, but it may make things easier for potential u= sers.<br> I&#39;m not sure I want to include someone else&#39;s prebuilt libraries, b= ut<br> I&#39;ll be sure to add a link to the BLAS/LAPACK download David mentioned.= <br> <div class=3D"im"><br> <br> &gt; Are there any plans to bring more of these numerical libraries (like<b= r> &gt; SciD or even basic linear alg) into Phobos? I would assume this would = be<br> &gt; very welcomed by a lot of D developers, no?<br> <br> </div>IMO, there should definitely be some basic linear algebra functionali= ty<br> in Phobos.<br> <div class=3D"im"><br> <br> &gt; Btw, Lars, many thx for your work on SciD....<br> <br> </div>You&#39;re welcome. :)<br> <font color=3D"#888888"><br> -Lars<br> </font></blockquote></div><br><div><br></div><div>How about Eigen?</div><di= v><a href=3D"http://eigen.tuxfamily.org/index.php?title=3DMain_Page">http:/= /eigen.tuxfamily.org/index.php?title=3DMain_Page</a></div><div><br></div><d= iv> I&#39;ve used it extensively in my C++ projects, and I&#39;m very pleased. = =A0</div><div><br></div><div>Last time I counted, Eigen was about 10-15k li= nes of C++ code. =A0It&#39;s all templates, and it might be the perfect pro= ject to translate to D to show off some of D&#39;s meta and template featur= es. =A0I only say this because I really miss not having a linear=A0algebra= =A0library in D, and I dislike language bindings.</div> <div><br></div><div>I&#39;m still learning D from TDPL, but I would really = like to see some sample code in D for doing expression templates. =A0For ex= ample, could anyone translate this to D?=A0<a href=3D"http://en.wikipedia.o= rg/wiki/Expression_templates">http://en.wikipedia.org/wiki/Expression_templ= ates</a>=A0=A0It took me about half an hour to understand the code because = it&#39;s kind of hard to read. =A0I wonder if D&#39;s version would be more= readable.</div> <div><br></div><div><br></div><div><br></div><div><br></div><div><br></div> --0023547c95cb283677049c41bcaf--
Feb 14 2011