FreeBSD benchmarks (56K warning)

sossego said:
How would this compare to, say, OpenBSD or NetBSD or to other architectures?
Click to expand...

The runs above are almost exclusively userlevel. The OS really doesn't matter. It is only to compare different Intel CPUs, caches, memory speed and the like.

I also have parallel versions of those but even two different FreeBSD kernels show different results. Multithreaded performance comparison of OSes will require more effort.
 
I'm aware that the 750 is an older PPC; but, can the test suite be built on such architecture and be run to compare the results?
IMO, running tests on a single architecture or a branch thereof- read x86(_64)- cannot truly measure the performance of a system.
Or...
If you have the source code available, let me build it on the G3 and I will post the same results here. If a test fails, I will also report that.
 
It's not "source code". It is a harddrive with a script that runs a bunch of stuff that came out of ports. Some things are in /compat/linux (such as the Linux kernel compile).

You will have a lot of work in front of you reproduce all this.

And after that you have things you can't compare. All the compile benchmarks have to be turned into cross-compile benchmarks (cross to x86) and you must use the same compiler versions. The video encoding must use, well, that's difficult because I let mplayer optimize itself to the processor at hand, but it is a mplayer version right out of the stone age.
 
My apologies for not knowing it is a script.
I'll take my chances if the script is available.

Not enough memory or CPU speed to cross compile. Mozilla/firefox version that I have on the G3 probably is not the same version that you have.
Linux compatibility is non-existent for FreeBSD on PPC. For cross compiling, see the memory and cpu statement earlier.

My machine is older- the G3- and much more limited.

Anything x86 dependent will not compile. I'm using SNAP 9.0.
You can read the PPC problem lists for what doesn't compile.
Kernel doesn't always build. Watch dog is enabled. All possible instruction set commands are not enabled. OpenBSD is more complete on that.


If I use the "default" setting for C on any non x86 processor, nothing will work.

Comparing one PPC to seven x86_64 processors doesn't seem that difficult, or even that much of a request.


I'll ask again when I do system rebuild on the G3.
 
The "source code" for the main benchmark runner is here:
http://www.cons.org/cracauer/crabench/crabench

Which benchmark specifically do you think will be useful for comparing platforms? I can only think of the crosscompiler setup, python and php. But the latter two just run the official php and python useless microbenchmarks anyway.
 
You must log in or register to reply here.
Back
Top