[ale] linux and apples

[Chris Ricker]

On Thu, 25 Sep 2003, Jonathan Rickman wrote:

> On Thursday 25 September 2003 10:42, Stephen Turner wrote:
> > linux works on a mac so its really comparing similar apps :D or even
> > benchmarking it somehow? does linux have a benchmark tool? :-p

lmbench is the usual tool for these sorts of things -- fairly low-level 
system call benchmarker, basically

> 
> Linux also runs on sparc, but it's not really fair to compare it to 
> Solaris. Solaris has the home field advantage. The Intel platform is 
> essentially the only platform where Linux does not have any quirky 
> proprietary barriers to overcome. I expect this to change on the G5 since 
> it is an IBM processor that will see plenty of Linux installs on their 
> pSeries line. 

I disagree. PCs are the worst platform for quirky proprietary barriers.  
You've got 3000 different vendors, all subtly incompatible with what few
standards exist to define your hardware, and none of the vendors want to
disclose any info about their hardware (with few exceptions). Sun, in
contrast, is essentially the only vendor in the sparc / ultrasparc space,
and they've been reasonably open with Linux developers when they've needed
docs (not with OpenBSD, but that seems as much due to how Theo demanded as
much as anything else, as usual ;-). There are lots of things in sparc-space
which aren't supported by Linux, but it's more a matter of cost / hardware
availability and developer interest than anything else.

You'll also find that if you benchmark Solaris vs Linux on identical
Ultrasparc gear, Linux trounces Solaris in many ways. Look at something as
basic as fork() -- last time I benched it, the cost of forking a process on
low-end Ultrasparcs is about 4500 microseconds running Solaris 9, and about
500 microseconds running Linux 2.4 on the exact same machine. As Dave Miller 
likes to point out, stracing a fork() on Linux is quicker than fork() on 
Solaris! ;-)

later,
chris