Dynamic Instrumentation of Large-Scale MPI and OpenMP Applications .


37 views
Uploaded on:
Category: Home / Real Estate
Description
04/16/2003. NC State University. 2. The Basic Idea ... Instrumentation expected to gather execution information. Static (incorporate or connection time) instrumentation Dynamic (run-time) instrumentation . Plausibility study for element instrumentation and progressively controlled static instrumentation. .
Transcripts
Slide 1

Dynamic Instrumentation of Large-Scale MPI and OpenMP Applications Christian Thiffault, Michael Voss, Steven T. Healey, Seon Wook Kim Presented By: Jaydeep Marathe NC State University

Slide 2

The Basic Idea .. Instrumentation expected to gather execution information. Static (assemble or connection time) instrumentation Dynamic (run-time) instrumentation Feasibility consider for element instrumentation and powerfully controlled static instrumentation . Paper attests: " Dynamic strategies superior to static instrumentation as for execution overhead & follow information measure " But that is still far from being obviously true, as we might see ... NC State University

Slide 3

Base Trampoline Mini-Trampoline Save Context Instrumentation Code Pre-Instrument Log_Event(… ); Start_Timer(); … .. … ... Reestablish Context Relocated Instruction Save Context Post-Instrument Restore Context Dynamic Instrumentation (with DPCL) Executing Program JMP _Test Probe Point NC State University

Slide 4

List of Active Points Yes ! Instrumentation Code Log_Event(… ); Start_Timer(); … .. … ... Dynamic Control of Static Instrumentation Library Compiler-Instrumented Program Conf_sync() { return; } … ... … ... Conf_sync() … ... Perused Active List … ... … ... Point Active ? Conf_sync() … ... … ... … ... Conf_sync() … ... Instrument order time , choose actuation at run-time . NC State University

Slide 5

Instrumentation Tool : Dynprof For blended (?) OpenMP + MPI programs GuideTrace OpenMP Library KAI Guide Compiler Application Source Dynprof VampirTrace Library Executable MPI Library Applicable just to the KAI OpenMP compiler Instrumentation : VampirTrace call to log follow Trace Data NC State University

Slide 6

Experiments .. Analyze execution times for static versus dynamic instrumentation. 4 ASCI benchmarks - Smg98 , Sppm , Sweep3d (MPI), Umt98 (OpenMP) IBM Power3 144-hub framework, every hub has 8 Power3 processors. Measure execution times with 5 situations: FULL : Full Static Instrumentation +Trace logging FULL-OFF : Full Static Instrumentation, no follow logging SUBSET : Full Static Instrumentation, just chose capacities dynamic. None : No instrumentation (Baseline) Dynamic : Dynamically embedded instrumentation. NC State University

Slide 7

Results: Smg98 NC State University

Slide 8

Results: Smg98 (contd) Static approach ( SUBSET ), more costly than element inclusion ( DYNAMIC ). WHY ?? - No clarification in paper . NC State University

Slide 9

Results: Smg98  Another understanding Observe values for FULL, FULL-OFF,SUBSET . Close !! Compiler instrumentation wastefulness  skews comes about !  dynamic approach looks better. NC State University

Slide 10

Time to make and instrument Significant cost, however just happens at startup. May altogether irritate application , when instrumented amid run-time . NC State University

Slide 11

Summary and Conclusions Tool to instrument MPI and OpenMP programs ( Dynprof ) Trace logging/perception through existing VampirTrace library. Look at expenses of static , dynamic and powerfully controlled static instrumentation. Affirms that powerfully embedded instrumentation is less costly, when contrasted with static instrumentation. NC State University

Recommended
View more...