Conversation
…rary compilation not yet available.
Conflicts: configure
This reverts commit ffc7c5e.
MODIFIED * include/driver/version.h dipoles/DIPOLE_overlaps.F
NEW *
Bugs:
- workaround for issue #729 to fix compilation
Patch sent by: Davide Sangalli <davide.sangalli@gmail.com>
Conflicts: config/report.in include/driver/version.h src/modules/.objects
src/modules/mod_magma2_cfortran.F
Outdated
| complex(c_float_complex), target :: A(lda,*) | ||
| real(c_double), target :: w(*) | ||
| complex(c_float_complex), target :: work(*) | ||
| real(c_double), target :: rwork(*) |
There was a problem hiding this comment.
In the MAGMA Fortran stubs w and rwork are float, not double. Any reason they are double in here?
There was a problem hiding this comment.
Hi,
you are right, this looks like a mistake... [ I haven't written the modules myself, but generated from magma-2.7.1 using a codegen.py script (only the z version is there)... Here the point is that the script is no longer distributed with this version of magma and I've downloaded a previous version... ]
nevertheless, you are right and this needs to be fixed !
I've attempted a fix here: 20369f3
|
I was trying to test this branch and add our part of the code over it.
|
Blanca Mellado Pinto)
|
Concerning these last comments:
|
| cat > conftest.$ac_ext <<_ACEOF | ||
| program main | ||
|
|
||
| use magma |
There was a problem hiding this comment.
Is this test needed? I think MAGMA Fortran stubs are not needed since the ones defined in Yambo are the ones that are used. In theory you could link a MAGMA library compiled without Fortran.
There was a problem hiding this comment.
So I was giving it another try.
- I am also using the nvidia sdk
- Apparently there is some conflict, at least in my case, when you have the Fortran stubs from MAGMA and also from Yambo. I was able to compile it using a MAGMA library without Fortran stubs. However I needed to comment this test in the configure to be able to do so. Have you tested this with a MAGMA library with Fortran stubs?
There was a problem hiding this comment.
Hi Blanca, thanks for reporting... we'll run the tests ASAP and let you know
|
@blmelp Was there anything else we should consider on your side and related to magma, or everything was in the valencia_magma branch? |
|
Everything from our side is on the valencia_magma branch, but using the MAGMA Fortran stubs. I tried to run a test with our code over the Yambo Fortran stubs for MAGMA (on a local branch), and ran into some problems that we discussed in this thread. |
4 participants
No description provided.