search

Noahs-ARK / semafor

http://www.ark.cs.cmu.edu/SEMAFOR
GNU General Public License v3.0
96 stars 46 forks source link

runMalt.sh tweak #12

Closed Liontooth closed 9 years ago

Liontooth commented 9 years ago

Should bin/runMalt.sh have a path for java on line 71?

time ${JAVA_HOME_BIN}/java -Xmx2g -jar maltparser-1.7.2.jar ...

BTW we're experimenting with -XX:+UseNUMA -- would you expect a speed-up? We're not seeing any improvements in median speed, but the occasional slower processing runs vanish. I don't know if there's a simple way to test for a NUMA-capable machine.

Cheers, David

sammthomson commented 9 years ago

On Wed, Mar 11, 2015 at 1:56 AM, David Liontooth notifications@github.com wrote:

Should bin/runMalt.sh have a path for java on line 71?

time ${JAVA_HOME_BIN}/java -Xmx2g -jar maltparser-1.7.2.jar

Yes, that's better.

BTW we're getting improved results if we use -XX:UseNUMA -- I don't know if there's a cheap check for NUMA-capable systems.

Oh, interesting! Improved results meaning faster runtime, right? I don't know if there's a general way to check either. If not, we could make it a user-controlled setting in config.sh, or just mention it in the readme.

Cheers, David

— Reply to this email directly or view it on GitHub https://github.com/sammthomson/semafor/issues/12.

Liontooth commented 9 years ago

It's a moderately faster runtime -- what we're seeing is a regularization of runtimes. Without NUMA we get spikes of really slow runs that take twice as long. With NUMA, these disappear, while the mode stays the same. This suggests NUMA improves memory management.

Note the correct syntax is -XX:+UseNUMA.

sammthomson commented 9 years ago

Fixed in 102e0c2d09618bbaf7ab3f924e0a8736d23e32dd and 20a58b732e181d3087da18aaf902dac9a2cdd36a (I added a note in the README). Thanks!