Difference between revisions of "PAID Project"

From SCECpedia
Jump to navigationJump to search
 
(6 intermediate revisions by the same user not shown)
Line 5: Line 5:
 
*[http://hypocenter.usc.edu/research/PAID_Project/Oct5_SCEC.pptx SCEC PAID Project Summary (ppt)]
 
*[http://hypocenter.usc.edu/research/PAID_Project/Oct5_SCEC.pptx SCEC PAID Project Summary (ppt)]
 
*[http://hypocenter.usc.edu/research/PAID_Project/future_performance_comparisons.docx.pdf PAID Performance Definitions (pdf)]
 
*[http://hypocenter.usc.edu/research/PAID_Project/future_performance_comparisons.docx.pdf PAID Performance Definitions (pdf)]
 +
 +
== Project Progress ==
 +
The SCEC PAID project funded by NCSA through Blue Waters project supports a GPU tuning effort, and will end in summer 2016. Recently,  plasticity and Q(f) have been merged to the baseline GPU code. NVIDIA Peng Wang and UIUC team help tune the code now, as shown in the Jan 2016 two slides, getting very good results. We plan soon to merge multi-streaming feature also into the new code, together we expected 30% or better performance gain overall in terms of wall clock time, will benefit High-F runs directly, with potential to be integrated to the  current SGT code to benefit CyberShake as well.
  
 
== Monthly Progress Reporting ==
 
== Monthly Progress Reporting ==
NCSA holds quarterly PAID progress telecoms. Each PAID project reports their progress to the group.
+
NCSA holds monthly PAID progress telecoms. Each PAID project reports their progress to the group.
*[ TBD Slides from Jan 26 Telecon]
+
*[http://hypocenter.usc.edu/research/PAID_Project/NVIDIA_PAID_Cui_Jan27.pptx Yifeng Cui slides from 27 Jan 2016 Telecon (pptx)]
*[http://hypocenter.usc.edu/research/PAID_Project/NCSA BW-PAID Presentation.Oct.7.2015.pptx Slides from Oct 2015 Telecon]
+
*[http://hypocenter.usc.edu/research/PAID_Project/NCSA_BW-PAID_Presentation.Oct.7.2015.pptx Yifeng Cui Slides from Oct 2015 Telecon (pptx)]
  
 
== Related Entries ==
 
== Related Entries ==

Latest revision as of 21:50, 27 January 2016

NCSA Blue Waters provides PAID projects with support to improve performance of scientific codes on current and future supercomputers. Yifeng Cui and Tom Jordan have received a PAID award to support software performance improvements.

SCEC PAID Proposal

Project Progress

The SCEC PAID project funded by NCSA through Blue Waters project supports a GPU tuning effort, and will end in summer 2016. Recently, plasticity and Q(f) have been merged to the baseline GPU code. NVIDIA Peng Wang and UIUC team help tune the code now, as shown in the Jan 2016 two slides, getting very good results. We plan soon to merge multi-streaming feature also into the new code, together we expected 30% or better performance gain overall in terms of wall clock time, will benefit High-F runs directly, with potential to be integrated to the current SGT code to benefit CyberShake as well.

Monthly Progress Reporting

NCSA holds monthly PAID progress telecoms. Each PAID project reports their progress to the group.

Related Entries