Lecture 8: Perf. Analysis & Visualization

[Pages:24]High Performance Computing Systems (CMSC714)

Lecture 8: Perf. Analysis & Visualization

Abhinav Bhatele, Department of Computer Science

Summary of last lecture

? Task-based programming models and Charm++ ? Key principles:

? Over-decomposition, virtualization ? Message-driven execution

? Automatic load balancing, checkpointing, fault tolerance

Abhinav Bhatele (CMSC714)

LIVE RECORDING 2

Tracing tools

? Record all the events in the program with timestamps ? Events: function calls, MPI events, etc.

Vampir visualization:

Abhinav Bhatele (CMSC714)

LIVE RECORDING 3

Tracing tools

? Record all the events in the program with timestamps ? Events: function calls, MPI events, etc.

Vampir visualization:

Abhinav Bhatele (CMSC714)

LIVE RECORDING 3

Tracing tools

? Record all the events in the program with timestamps ? Events: function calls, MPI events, etc.

Vampir visualization:

Abhinav Bhatele (CMSC714)

LIVE RECORDING 3

MPI trace visualization

Vampir

Abhinav Bhatele (CMSC714)

Jumpshot

LIVE RECORDING 4

Projections Performance Analysis Tool

? For Charm++/Adaptive MPI programs ? Instrumentation library

? Records data at the granularity of chares (Charm++ objects)

? Java-based GUI

Abhinav Bhatele (CMSC714)

LIVE RECORDING 5

Time Profile



Figure 3. Time profile for ApoA1 on 1k processors of Blue Gene/L (with PME) in Projections

Abhinav Bhatele (CMSC714)

LIVE RECORDING 6

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download