Execution Speed of NairnMPM

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Check out and compile revision 188
2. Check out and compile revision 189
3. Compare speeds of some calculations

What is the expected output? What do you see instead?

The two codes should be identical and run in about the same speed. The changes 
between r188 and r189 did not change any algorithms, but did reorganize the 
code into tasks. On most machines, the results and the speed are identical. On 
the Carbon cluster at OSU, however, r188 is about twice as slow and r189.

Benchmark calculations have been done with an OSB compaction run that takes 
about an hour on most machines. On Carbon cluster, r188 take about 5000 seconds 
and r189 takes about 10000 seconds. The slow down is not uniform. It slows the 
most in the tasks to get grid forces (2X slower), update momenta (2X slower), 
and update particles (10X slower)

Original issue reported on code.google.com by johnanairn@gmail.com on 1 Dec 2010 at 11:44

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

I noticed a small calculation ran three times faster (in r271) when compiled in 
Gnu 4.0.3 then it did when compiled with Gnu 4.3.1. I reran the large OSB 
calculations, however, and the Gnu version did not affect the execution speeds 
between r188 and r189 (still a factor of two).

Original comment by johnanairn@gmail.com on 1 Dec 2010 at 11:49

• Added labels: Type-Speed
• Removed labels: NairnMPM, Type-Other