Friday, March 18, 2016

lmbench results

I downloaded and ran lmbench on the Raspberry Pi 3, and also on the Pentium III as a comparison.  There is a page with the summary and some graphs.  The one of most interest is the memory read latency benchmark.


LMBench is pretty stale, originally released (pdf) back in 1996, it was quite sophisticated in divining the actual hardware architecture of many diverse processors.  I like the graphs.  They have a solid engineering feel to them to this day.

What I did on the rpi to run this:

mkdir src
cd src
wget http://www.bitmover.com/lmbench/lmbench3.tar.gz
tar zxvf lmbench3.tar.gz
cd lmbench3
mkdir SCCS
touch SCCS/s.ChangeSet
cd src
make results
cd ../results/
sudo echo true > /usr/local/bin/bk
sudo chmod +x /usr/local/bin/bk
mkdir HTML
sudo aptitude install ghostscript netpbm imagemagick
make html


The HTML directory has the summary and some gif files.  I had to edit the Makefile and comment out the gs PS.6, PS.7 and PS.8 lines to get make to run to completion.
Bitrot setting in after 11 years.

-r--r--r-- ob/ob          9474 2005-08-22 20:19 lmbench3/results/Makefile

#gs -sOutputFile=HTML/mem-unscaled%02d.$(IMAGE) -sDEVICE=$(IMAGE) -q -dNOPAUSE PS/PS.6 < /dev/null
#gs -sOutputFile=HTML/bwfile-unscaled%02d.$(IMAGE) -sDEVICE=$(IMAGE) -q -dNOPAUSE PS/PS.7 < /dev/null
#gs -sOutputFile=HTML/bwmem-unscaled%02d.$(IMAGE) -sDEVICE=$(IMAGE) -q -dNOPAUSE PS/PS.8 < /dev/null

No comments:

Post a Comment