Topic
  • 4 replies
  • Latest Post - ‏2014-05-08T17:37:08Z by Bill_Buros
chwilk
chwilk
9 Posts

Pinned topic State of genomics/bioinformatics software support on PowerLinux

‏2014-05-07T17:10:30Z | porting software

I have some users on a Power7 cluster who are trying to run the latest versions of several genomic sequencing software packages, namely Tophat/Bowtie and bwa. As we're trying to compile these packages, I've found that the most recent versions have been massaged by Intel developers to include lots of SSE assembly code with no clear #ifdef syntax to get around their optimizations.

Is somebody at IBM keeping up with codes like these who could provide some upstream help to the developer community to keep their codes from becoming tied to closely to one processor platform? It seems a step backwards to me, because older versions compile fine.

Offending codes:

  • Bill_Buros
    Bill_Buros
    190 Posts

    Re: State of genomics/bioinformatics software support on PowerLinux

    ‏2014-05-07T22:14:15Z  

    Hey Chandler, we do see a marked tendency for some codes to be very x86 centric and specific.   With OpenPOWER Foundation emerging, we are hopeful that we can begin instilling some cross-platform focus on codes like this.

    We do have the IBM SDK available which can help identify the common code sequences and offer Power specific alternatives.   Is there someone there who might be available or willing to start working through the source with the tools?

  • chwilk
    chwilk
    9 Posts

    Re: State of genomics/bioinformatics software support on PowerLinux

    ‏2014-05-07T22:27:47Z  

    Hey Chandler, we do see a marked tendency for some codes to be very x86 centric and specific.   With OpenPOWER Foundation emerging, we are hopeful that we can begin instilling some cross-platform focus on codes like this.

    We do have the IBM SDK available which can help identify the common code sequences and offer Power specific alternatives.   Is there someone there who might be available or willing to start working through the source with the tools?

    Hi Bill,

    I'm not sure if my customers for these packages have any developers among them. I was somewhat hoping that we might find somebody within IBM who could advocate portability in the upstream, much the same as Intel has already advocated for their own optimizations. I remember something from the old days of the PowerMac Beowulf cluster at Virginia Tech about a special "Bio Team" of IBM developers, engineers and scientists who specialized in the application domain. 

  • Bill_Buros
    Bill_Buros
    190 Posts

    Re: State of genomics/bioinformatics software support on PowerLinux

    ‏2014-05-07T22:57:40Z  
    • chwilk
    • ‏2014-05-07T22:27:47Z

    Hi Bill,

    I'm not sure if my customers for these packages have any developers among them. I was somewhat hoping that we might find somebody within IBM who could advocate portability in the upstream, much the same as Intel has already advocated for their own optimizations. I remember something from the old days of the PowerMac Beowulf cluster at Virginia Tech about a special "Bio Team" of IBM developers, engineers and scientists who specialized in the application domain. 

    A challenge for all of us.    :- )

    It's easy to advocate portability, it's sometimes harder to make something happen in specific domain knowledge realms.   

    I'll go chase down our buds in the HPC realm here @ IBM and see what help can be applied!  

  • Bill_Buros
    Bill_Buros
    190 Posts

    Re: State of genomics/bioinformatics software support on PowerLinux

    ‏2014-05-08T17:37:08Z  
    • chwilk
    • ‏2014-05-07T22:27:47Z

    Hi Bill,

    I'm not sure if my customers for these packages have any developers among them. I was somewhat hoping that we might find somebody within IBM who could advocate portability in the upstream, much the same as Intel has already advocated for their own optimizations. I remember something from the old days of the PowerMac Beowulf cluster at Virginia Tech about a special "Bio Team" of IBM developers, engineers and scientists who specialized in the application domain. 

    Chandler, we have found several teams inside IBM who have been updating / modifying / optimizing those codes.   We're working to find out how to get those improvements to the community maintainers.    :- )