UPC group display case slide

Download Report

Transcript UPC group display case slide

Unified Parallel C Group
PI: Dr. Alan D. George
Group Leader: Hung-Hsun Su
Drastic Increase
in Complexity
Sequential Programming
(C, Java, …)
Parallel Programming
(UPC, SHMEM)
Research
SHMEM
• Benchmarking and analysis of current
compilers
• SHMEM = single-ended, asynchronous data
passing library based on a hybrid message
passing/shared memory programming model
• Creation of a test suite of UPC
applications for evaluating current and
future implementations of UPC on SANbased clusters
4
4.5
3.5
4
3
3.5
2.5
3
Speedup
Execution Time (sec)
• UPC = Parallel extension of ANSI C standard
based on shared-memory programming model
2
1.5
2.5
2
1.5
1
1
0.5
0.5
0
1
10
Block Size
2 Processors
100
1000
0
UPC Shared Pointer
4 Processors
1 Processor
UPC Shared/Local Pointer
2 Processors
4 Processors
UPC Local Pointer
Unified Parallel C Group
PI: Dr. Alan D. George
Without Tool
With Tool
Group Leader: Hung-Hsun Su
Development of tools used for performance analysis,
identifying bottlenecks and enable source code and
compilation optimization
Tools
UPC/SHMEM
Program
UPC/SHMEM
Program
Compile & Execute
Performance
Analysis Tool
Poor Performance
Optimized
Program
Parallel Programming
Models
SHMEM
Compile & Execute
Improved
Performance
Optimized
Performance
• Bring performance analysis to UPC and SHMEM tools
• Bring UPC and SHMEM support to performance tools
• Determine most feasible approach and pursue
• Focus on key issues at multiple levels; language,
mapping, architecture