MVAPICH MPI on LC Clusters

Oct 30, 2018
 

Date / Time Oct 30, 2018   9:00am - 5:00pm
Location

Building 453  Room 1012  (Black Diamond Room)
Note: This is a Property Protection Area. Foreign national temporary escorted building access procedures apply.

Title MVAPICH MPI on LC Clusters
Presenters Hari Subramoni and Mark Arnold
Description

Hari Subramoni and Mark Arnold from the MVAPICH development team will on site to discuss how to get the most out of MVAPICH MPI on LC clusters, including both TOSS and CORAL systems.  Drop in for any and all sessions you’d like.  After each presentation, time is scheduled for Q&A, so come ready with your questions and feedback.

Agenda:

9am-10am: Programming, Optimizing and Tuning your Message Passing Interface (MPI) Application with MVAPICH2, non-blocking collectives and profiling with MPI_T

10am-10:30am: break w/ Q&A

10:30am-11:30am: Taking Advantage of Enhanced Support in MVAPICH2-X to Obtain Best Application Performance

11:30am-12pm: break w/ Q&A

12pm-2pm: break for lunch

2pm-3pm: MVAPICH2-GDR: High-Performance CUDA-Aware MPI Library for GPU-Enabled Clusters

3pm-4pm break w/ Q&A

4pm-4:45pm: Challenges in Designing MPI Runtimes and Deep Learning Frameworks for Modern Multi-GPU Clusters

4:45pm-5pm Q&A

What to Bring This is a presentation-only workshop - no hands-on. Attendees do not need to bring anything.
Fee No cost
Level/Prerequisites Familiarity with MPI and using LC clusters would be beneficial.
Registration

No registration is required

Questions? Please call or send e-mail to Adam Moody (925-422-9006 / moody20@llnl.gov).

 

UCRL-MI-131558