search

dwaaan / pyrit

Automatically exported from code.google.com/p/pyrit
0 stars 0 forks source link

issue 208 should not be merged with issue 173. Please address this performance graphing error. #219

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago 
You merged http://code.google.com/p/pyrit/issues/detail?id=208
with http://code.google.com/p/pyrit/issues/detail?id=173
I still don't feel these are the same topic.  You refused to ack the straight 
forward point I made on the perf graphs being incorrect as of the change in 
PMK/s method of calc I mention in id 208.  Please address this.

Original issue reported on code.google.com by codytu...@gmail.com on 6 Dec 2010 at 11:49

GoogleCodeExporter commented 8 years ago 
I'll try to explain again.

Total performance is SUSTAINED PERFORMANCE for whole system. This includes cpu 
data preparation, data transfer to/from gpu, gpu computations. 
Numbers on the right side of the core are PEAK PERFORMANCE for this core ( 
opencl core or cuda core ). This means that if the cpu would prepare data with 
infinite speed, data transfer would have infinite speed you would achieve those 
values for sustained performance.

But in real life PEAK PERFORMANCE >>> SUSTAINED PERFORMANCE. 

There was no change in the method of computing overall performance. What you 
see is not an error in performance calculations. Simply due to problems from 
issue 173 it's much more visible in your system that PEAK >>> SUSTAINED.

> Adding all of the PMKs/s should be 60672.2, not 28389.98
I'll write once more to be perfectly clear. PEAK PERFORMANCE doesn't add up to 
SUSTAINED PERFORMANCE.

In your system CPU is bottleneck ( issue 173 ). Use latest svn version and 
disable all CPU cores ( using limit_ncpus=4 in ~/.pyrit/config ).

Original comment by hazema...@gmail.com on 10 Dec 2010 at 5:58