Hi Massimiliano,
Sorry for taking a long time to reply here. I've been at a GRC which tends
to leave one no time to actually check email - something that is actually
good for the soul occasionally I think.
So, the very short answer is that:
1) AMBER 11 does not support CUDA > 4.2
2) AMBER 11 does not support Kepler (aka K10, K20, GTX6XX, GTX7XX) GPUs
Both these will require AMBER 12 I'm afraid. The changes needed which
included an entirely new fixed precision model (SPFP) were such that it
was not practical to back port this to AMBER 11.
As for the failed bug fixes. I am not sure why it is not applying cleanly.
They should work. Have you tried starting from a completely clean AMBER 11
tar file? From what I read from the error messages it looks like it is
complaining that the patches have already been applied which suggests that
either the first patch attempt you did worked or the source tree you have
is messed up in some way. I would try reverting to the original tar file
and trying again if you can.
So my suggestion if you intend to use AMBER 11 is to:
1) Install CUDA 4.2
2) Untar from original tar files
3) Patch AMBER Tools first using the instructions here (make sure to
scroll down to v 1.5)
http://ambermd.org/bugfixesat.html
4) Patch AMBER 11 with:
Download bugfix.all.tar.bz2
<
http://ambermd.org/bugfixes/11.0/bugfix.all.tar.bz2>.
Download apply_bugfix.x
<
http://ambermd.org/bugfixes/11.0/apply_bugfix.x>.
Be sure that your AMBERHOME environment variable points to the amber11
directory that contains your Amber11 distribution
Patch your system by typing the following (in whatever directory you
downloaded these files to):chmod 700 apply_bugfix.x
./apply_bugfix.x /path/to/bugfix.all.tar.bz2
This should patch your source files with the latest bugfixes. Any
bugfixes that appear to have already been applied will be skipped.
This should hopefully apply cleanly and then you can build. Note as far as
I know this will only work for GPUs with SM2.0 or less (I.e. M2090, GTX580
and lower) so it won't be any use on your K20s. Your only option there is
to switch to AMBER 12.
Hope that helps.
All the best
Ross
On 7/25/13 7:55 AM, "Massimiliano Porrini" <m.porrini.iecb.u-bordeaux.fr>
wrote:
>Hi all,
>
>I have been trying to install Amber11 and AT1.5 on our GPU
>workstation with one Kepler K20c and Ubuntu 13.04 OS .
>The CUDA toolkit 5 and the Nvidia drivers were installed (not by me)
>from the Ubuntu repositories (using the apt-get utility).
>
>The issue is that I am experiencing problems with some bugfixes of
>Amber11,
>specifically with bugfix.19, bugfix.20 and bugfix.22 .
>
>This is what I get when I try to patch Amber11, both using the tarball
>file (bugfix.all.tar.bz2) and the individual ones (19, 20 and 22,
>as shown below):
>
>
>patch -p0 -N < bugfix.19
>
>patching file src/pmemd/src/cuda/gpu.cpp
>Hunk #1 FAILED at 442.
>Hunk #2 FAILED at 3033.
>Hunk #3 FAILED at 3068.
>Hunk #4 FAILED at 3859.
>4 out of 4 hunks FAILED -- saving rejects to file
>src/pmemd/src/cuda/gpu.cpp.rej
>patching file src/pmemd/src/cuda/gputypes.h
>Reversed (or previously applied) patch detected! Skipping patch.
>1 out of 1 hunk ignored -- saving rejects to file
>src/pmemd/src/cuda/gputypes.h.rej
>patching file src/pmemd/src/cuda/kBNL.h
>Hunk #1 FAILED at 42.
>Hunk #2 FAILED at 58.
>Hunk #3 FAILED at 115.
>Hunk #4 FAILED at 372.
>Hunk #5 FAILED at 429.
>5 out of 5 hunks FAILED -- saving rejects to file
>src/pmemd/src/cuda/kBNL.h.rej
>patching file src/pmemd/src/cuda/gputypes.h
>Reversed (or previously applied) patch detected! Skipping patch.
>1 out of 1 hunk ignored -- saving rejects to file
>src/pmemd/src/cuda/gputypes.h.rej
>
>
>patch -p0 -N < bugfix.20
>
>patching file src/pmemd/src/cuda/cuda_info.fpp
>Reversed (or previously applied) patch detected! Skipping patch.
>1 out of 1 hunk ignored -- saving rejects to file
>src/pmemd/src/cuda/cuda_info.fpp.rej
>patching file src/pmemd/src/cuda/gpu.cpp
>Hunk #1 FAILED at 441.
>Hunk #2 FAILED at 2511.
>Hunk #3 FAILED at 3070.
>Hunk #4 FAILED at 3765.
>Hunk #5 FAILED at 3821.
>Hunk #6 FAILED at 5372.
>Hunk #7 FAILED at 5618.
>Hunk #8 FAILED at 6774.
>8 out of 8 hunks FAILED -- saving rejects to file
>src/pmemd/src/cuda/gpu.cpp.rej
>patching file src/pmemd/src/cuda/gputypes.h
>Reversed (or previously applied) patch detected! Skipping patch.
>1 out of 1 hunk ignored -- saving rejects to file
>src/pmemd/src/cuda/gputypes.h.rej
>patching file src/pmemd/src/cuda/kBNL.h
>Reversed (or previously applied) patch detected! Skipping patch.
>12 out of 12 hunks ignored -- saving rejects to file
>src/pmemd/src/cuda/kBNL.h.rej
>patching file src/pmemd/src/cuda/kNeighborList.cu
>Reversed (or previously applied) patch detected! Skipping patch.
>4 out of 4 hunks ignored -- saving rejects to file
>src/pmemd/src/cuda/kNeighborList.cu.rej
>
>
>patch -p0 -N < bugfix.22
>
>patching file src/pmemd/src/master_setup.fpp
>Reversed (or previously applied) patch detected! Skipping patch.
>1 out of 1 hunk ignored -- saving rejects to file
>src/pmemd/src/master_setup.fpp.rej
>patching file src/pmemd/src/runfiles.fpp
>Hunk #1 FAILED at 269.
>1 out of 1 hunk FAILED -- saving rejects to file
>src/pmemd/src/runfiles.fpp.rej
>
>
>>From what I read in amber mailing list archives I should do a "fresh"
>installation
>of Amebr11 and AT1.5, and this is what I have done twice,
>but the above failures remain.
>
>By the way, in one of my attempts, I did not care about these
>"failure" messages and I continued installing. To bottom line:
>I was able to build Amber11 and AT1.5 both in serial and parallel,
>except the CUDA version of Amber11 (obviously..!).
>
>In the mailing list archives, I also read that the best thing to do would
>be upgrading
>to Amber12, but I wonder if someone has the solution to these issues,
>as for the moment I'd urgently need to install Amber11.
>
>Does anybody have any suggestion on how to fix these CUDA
>Amber11 patching failures?
>
>Many thanks in advance,
>Max
>
>
>--
>Dr Massimiliano Porrini
>Institut Européen de Chimie et Biologie (IECB)
>CNRS UMR 5248 Chimie et Biologie des Membranes et des Nano-objets (CBMN)
>2, rue Robert Escarpit
>33607 Pessac Cedex
>FRANCE
>
>Tel : 33 (0)5 40 00 30 31
>Fax : 33 (0)5 40 00 22 00
>http://www.iecb.u-bordeaux.fr/
>Emails: m.porrini.iecb.u-bordeaux.fr
> M.Porrini.ed.ac.uk
> maxp.iesl.forth.gr
> mozz76.gmail.com
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jul 30 2013 - 20:00:03 PDT
