It appears that if you installed ABySS with MPI support AND did not configure MPI then you could still use the np option to use ABYSS-P on one machine. ABYSS-P was not generated since you didn't install with mpi support. This means you can only use the threading parameter (j=20) and cannot run abyss-pe with the np parameter, or you must install with MPI support.
Cheers,
Tony
________________________________________
From: abyss...@googlegroups.com [abyss...@googlegroups.com] On Behalf Of Dong [dong...@gmail.com]
Sent: Wednesday, September 07, 2011 10:10 AM
To: ABySS
Subject: run abyss-pe on a single machine with multiple cores
Hi Dong,
I’m actually not too sure about this. Shaun or someone else in the ABySS users group may be able to help you find a suitable value for the s parameter.
Cheers,
Tony
From: dong he
[mailto:dong...@gmail.com]
Sent: Friday, September 09, 2011 5:59 AM
To: Tony Raymond
Subject: Re: run abyss-pe on a single machine with multiple cores
Hi Tony,
Yes, it was tmp space problem. Thank you! By the way, I have another question
need your suggestion:
I have four libraries with different fragment sizes, these libraries can be
assembled together, but how to specify the s parameter?
Thank
Dong
On Thu, Sep 8, 2011 at 12:16 PM, Tony Raymond <tray...@bcgsc.ca> wrote:
Hi Dong
Looks like this is your problem:
sort: write failed: /tmp/sortuatupN: No space left on device
If you don't have access to that directory to clear some
space or it is too small you should be able to change the TMPDIR environment
variable to change the directory sort uses for temporary files.
Cheers,
Tony
________________________________________
From: dong he [dong...@gmail.com]
Sent: Thursday, September 08, 2011 8:50 AM
To: Tony Raymond
Subject: Re: run abyss-pe on a single machine with multiple cores
Hi Tony,
I re-ran my job, it seems
working well, but still got error:
--------------------------------------------------------------------------------
-----
----
sort: write failed: /tmp/sortuatupN: No space left on device
gunzip -c testnohup-3.sam.gz \
|DistanceEst -j20 -k55 -s100 -n5 -o testnohup-3.dist testnohup-3.hist
warning: the seed-length should be at least twice k: k=55, s=100
DistanceEst: error: no @SQ records in the SAM header
make: *** [testnohup-3.dist] Error 1
make: *** Deleting file `testnohup-3.dist'
-----------------------------------------------------------------------------------------------
could you let me know what's wrong with it?
Thanks!
Dong
On Wed, Sep 7, 2011 at 2:22 PM, Tony Raymond <tray...@bcgsc.ca<mailto:tray...@bcgsc.ca>> wrote:
> Hi Dong,
>
> No problem! With your current installation, the single end assembly
(ABYSS) would run on one core. Most of the paired end assembly is threaded and
that portion will use 20 cores. You will need to reinstall with MPI support if
you wish to have the whole assembly run in parallel. Also, if you do reinstall
you will only need to specify the np parameter since the default for j=np.
>
> Cheers,
> Tony
>
> -----Original Message-----
> From: dong he [mailto:dong...@gmail.com<mailto:dong...@gmail.com>]
> Sent: Wednesday, September 07, 2011 12:13 PM
> To: Tony Raymond
> Subject: Re: run abyss-pe on a single machine with multiple cores
>
> Hi Tony,
>
> Thank you for your email. I will try it, and if I use j=20 on my
> current installation, would abyss-pe behave same as that when np=20
> is used on ABySS with MPI support? it means if abyss-pe would run on
> 20 cpus when j=20?
>
> Thanks again!
>
> Dong
>
>
>
> On Wed, Sep 7, 2011 at 1:44 PM, Tony Raymond <tray...@bcgsc.ca<mailto:tray...@bcgsc.ca>> wrote:
>> Hi Dong,
>>
>> It appears that if you installed ABySS with MPI support AND did not
configure MPI then you could still use the np option to use ABYSS-P on one
machine. ABYSS-P was not generated since you didn't install with mpi
support. This means you can only use the threading parameter (j=20) and cannot
run abyss-pe with the np parameter, or you must install with MPI support.
>>
>> Cheers,
>> Tony
>> ________________________________________
As a reasonable default, I'd suggest setting s=200.
Cheers,
Shaun
On Fri, 2011-09-09 at 16:07 -0700, Tony Raymond wrote:
> Hi Dong,
>
>
>
> IοΏ½m actually not too sure about this. Shaun or someone else in the