[armedbear-devel] Maxima profiling results

Erik Huelsmann ehuels at gmail.com
Thu Jul 30 10:41:13 UTC 2009


Today I changed the profiler in ABCL to determine hot spots in our
code, by looking only at the 8 stack frames at the head of the stack.
The outcome of running the test suite with that changed profiler is
this:


 10.0   271135 TMS
 10.2   275152 TIMESIN
 10.7   289952 SIMPTIMES
 11.1   300877 GREAT
 12.3   332154 KINDP
 13.7   371011 MEVAL1
 14.1   382393 DMARK
 15.8   427010 MEVAL
 26.1   706542 SIMPLIFYA
 74.3  2009777 DLSF
 80.5  2177991 DLS
 92.2  2493681 DGRF
100.0  2705263 DGR

Where DGR is the hottest spot and the others are percentages of that.
So, it looks like it that DGR is in the 8 most recent stack frames the
most often of all functions used. Note this is based on sampling at a
rate of 1ms intervals, not exact measurement.

Robert,

I'm hoping together we can explain the outcomes?


Bye,


Erik.




More information about the armedbear-devel mailing list