Re: [RFC] LZO de/compression support - take 4

[Richard Purdie <richard@openedhand.com>]

On Fri, 2007-05-25 at 18:07 +0530, Satyam Sharma wrote:
> On 5/25/07, Richard Purdie <richard@openedhand.com> wrote:
> > On Fri, 2007-05-25 at 17:15 +0530, Nitin Gupta wrote:
> > > Richard, can you please provide perf. results for this patch also?
> > > Also, can you please mail back latest version of your LZO patch? In
> > > meantime, I will try to include benchmarking support to the
> > > 'compress-test' module.
> >
> > This version is 15% slower at decompression and about equal on
> > compression.
> 
> I hope you tested your _safe variant against this, Nitin has done away
> with the _unsafe version in this patch. 

I am.

> Also, are you using your crypto
> lzo-support + tcrypt changes ( http://lkml.org/lkml/2007/5/1/303 ) to
> benchmark these?

No, I'm putting the code into userspace and testing it there. Its not
difficult since there is little support needed from the kernel for
either set of code.

The tester runs each one in turn under the same circumstances from the
same binary and gives me a speed for each. Multiple runs return
consistent values and I'm being careful to make sure its an otherwise
idle machine and processes enough data to make sure there is no nasty
cache effects.

> > I am however still strongly of the opinion that we should just use the
> > version in -mm (which is my latest version).
> 
> Right, if the difference is anything >10%, code cleanup does lose
> its attractiveness.

Agreed, and I still have the security and maintainability concerns. Add
them all together and its more unattractive.

Cheers,

Richard