Search found 33 matches

by evanlezar
Tue Jun 05, 2012 2:25 pm
Forum: User discussion
Topic: Incorrect determination of free memory for OOC GETRF
Replies: 1
Views: 1595

Incorrect determination of free memory for OOC GETRF

Hi, I have just run across the following problem: When I use my primary GPU under Windows to do some quick tests -- it's not very fast, so this is just to check if the code runs -- the the calls to magma_zgertf (and other LU-decomposition routines) fail in some cases when the OOC routines are used. ...
by evanlezar
Mon Jun 04, 2012 12:14 pm
Forum: User discussion
Topic: Progress output in MAGMA
Replies: 2
Views: 1866

Re: Progress output in MAGMA

Thanks mark.

I will add the progress code as you suggest. For now I am not concerned with terminating the calculation.

I will add the code with the relevant defines and submit a patch to see if it is to you liking.

Regards
by evanlezar
Thu May 31, 2012 2:49 am
Forum: User discussion
Topic: Progress output in MAGMA
Replies: 2
Views: 1866

Progress output in MAGMA

Hi, We use MAGMA predominantly for the LU decomposition. One nice to have feature would be support for displaying some form of progress for the computational processes -- something like the mkl_progress or acml_progress available in mkl and acml. This does not have to be particularly fine-grained. I...
by evanlezar
Wed May 30, 2012 10:31 am
Forum: User discussion
Topic: Incorrect LAPACK/BLAS uppercase symbols
Replies: 4
Views: 2465

Re: Incorrect LAPACK/BLAS uppercase symbols

Thanks Mark,

I also sent a patch to Stan and Mathieu containing this and a couple of other fixes.

I have not received any notification that they got this. Should I send this on to you as well?

Evan
by evanlezar
Wed May 16, 2012 3:28 pm
Forum: User discussion
Topic: Incorrect LAPACK/BLAS uppercase symbols
Replies: 4
Views: 2465

Re: Incorrect LAPACK/BLAS uppercase symbols

Thanks.

Also, keep up the great work!
by evanlezar
Wed May 16, 2012 12:24 pm
Forum: User discussion
Topic: Incorrect LAPACK/BLAS uppercase symbols
Replies: 4
Views: 2465

Incorrect LAPACK/BLAS uppercase symbols

It seems that the external lapack and blas symbols where all translated from the complex double precision case, and as such, the names mangled using the FORTRAN_NAME macro in magma_[csd]lapack.h are not consistent. For example, magma_slapack.h contains #define blasf77_saxpy FORTRAN_NAME( saxpy, ZAXP...
by evanlezar
Sat Feb 18, 2012 2:25 am
Forum: User discussion
Topic: Magma bug reports
Replies: 2
Views: 1455

Re: Magma bug reports

Thanks I will do so.
by evanlezar
Mon Feb 13, 2012 10:11 am
Forum: User discussion
Topic: Magma bug reports
Replies: 2
Views: 1455

Magma bug reports

Hi, I just have a quick question. If we come across a bug in MAGMA, what is the best way to report it? Is it still, as was the case in the past, best to email one of the developers with a description and a possible fix, or is there a more formal issue tracker? Thanks for the great work! Keep it up. ...
by evanlezar
Sat Mar 12, 2011 9:52 am
Forum: User discussion
Topic: Segmentation violation when running MAGMA rc4 tests
Replies: 1
Views: 2493

Re: Segmentation violation when running MAGMA rc4 tests

I seem to have solved the problem.

The issue is that my make.inc had CC=gcc -shared and as such the executables were being compiled as shared libraries. Setting CC=gcc in the testing Makefile and including some of the other object files for the compilation solved the problem.
by evanlezar
Sat Mar 12, 2011 9:05 am
Forum: User discussion
Topic: Segmentation violation when running MAGMA rc4 tests
Replies: 1
Views: 2493

Segmentation violation when running MAGMA rc4 tests

I have been trying to get MAGMA 1.0.0-rc4 running and have run into a problem. Every one of the testing programs causes a segmentation fault. Furthermore, this segmentation fault is thrown before any of the code in the main routine is executed. The gdb output for the testing_sgemm program is shown b...