Compiling for use on CPU

65 views
Skip to first unread message

Stanley Fujimoto

unread,
Feb 18, 2015, 4:22:00 AM2/18/15
to nvbio...@googlegroups.com
Is it possible to compile the included nvbio programs to run solely on the CPU? Specifically, nvBowtie and nvBWT but the others as well if possible.

I've been so imporessed with their performance and would like to see how they perform just on a CPU for comparison and portability of code utilizing nvbio.

Perhaps there is a comprehensive list of arguments that I can pass to cmake somewhere?

Thanks!

Jacopo Pantaleoni

unread,
Feb 18, 2015, 7:40:44 AM2/18/15
to Stanley Fujimoto, nvbio...@googlegroups.com

Dear Stanley,

The applications, in particular nvBowtie and nvBWT, are GPU only.

There's no specific reason for it, besides they were started when nvbio was still mostly GPU only.
However, we might build the CPU counterpart in the future - it's mostly about rewriting some kernels as for_each statements and switching device_tag's to host_tag's.

On the other hand those are apps for which there is a CPU version (e.g. bowtie2) - so we find it a little less important.

But converting the example programs (e.g. fmmap) to the host should be trivial and give a fairly good idea of how fast the CPU backend is.

best,
-jacopo

--
You received this message because you are subscribed to the Google Groups "nvbio-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nvbio-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Stanley Fujimoto

unread,
Feb 19, 2015, 12:32:28 PM2/19/15
to nvbio...@googlegroups.com, sfuj...@gmail.com
cool, thanks for the response! i really appreciate it.

Jacopo Pantaleoni

unread,
Feb 23, 2015, 2:57:24 PM2/23/15
to nvbio...@googlegroups.com, sfuj...@gmail.com
Dear Stanley,

as of today NVBIO contains a new app, nvLigther, which is both cross-platform as well as multi-gpu.

The source code should give you an idea of how easy it is to run essentially the same exact code
on both platforms.

best,
-jacopo

M. Stanley Fujimoto

unread,
Feb 23, 2015, 6:10:58 PM2/23/15
to Jacopo Pantaleoni, nvbio...@googlegroups.com
this is great! thanks again.
--
Stanley Fujimoto
Computational Science Laboratory
Brigham Young University
sfuj...@gmail.com
Reply all
Reply to author
Forward
0 new messages