Computer Science-Adaptive Two-level Thread Management.ppt

Computer Science-Adaptive Two-level Thread Management Adaptive Two-level Thread Management for MPI Execution on Multiprogrammed Shared Memory Machines
Kai Shen, Hong Tang, and Tao Yang

Department puter Science
University of California, Santa Barbara
2017/12/21
1
Shen, Tang, and Yang @ puting'99
MPI-Based putation on Shared Memory Machines
Shared Memory Machines (SMMs) or SMM Clusters e popular for high puting.
MPI is a portable high performance parallel programming model.
 MPI on SMMs
Threads are easy to program. But MPI is still used on SMMs:
Better portability for running on other platforms (. SMM clusters);
Good data locality due to data partitioning.
2017/12/21
2
Shen, Tang, and Yang @ puting'99
Scheduling for Parallel Jobs in Multiprogrammed SMMs
Gang-scheduling
Good for parallel programs which synchronize frequently;
Affect resource utilization (Processor-fragmentation; not enough parallelism to use allocated resource).
Space/time Sharing
Time bined with dynamic partitioning;
High throughput. Popular in current OS (., IRIX )
Impact on MPI program execution
Not all MPI nodes are scheduled simultaneously;
The number of available processors for each application may change dynamically.
Optimization is needed for fast MPI execution on SMMs.
2017/12/21
3
Shen, Tang, and Yang @ puting'99
Techniques Studied
Thread-Based MPI execution [PPoPP’99]
Compile-time transformation for thread-safe MPI execution
Fast context switch and synchronization
munication through address sharing
Two-level thread management for multiprogrammed environments
Even faster context switch/synchronization
Use scheduling information to guide synchronization
Our prototype system: TMPI
2017/12/21
4
Shen, Tang, and Yang @ puting'99
Impact of synchronization on coarse-grain parallel programs
Running munication-infrequent MPI program (SWEEP3D) on 8 SGI Origin 2000 processors with multiprogramming degree 3.
Synchronization costs 43%-84% of total time.
Execution time breakdown for TMPI and SGI MPI:
2017