[Zlib-devel] [7/8][RFC V3 Patch] Blackfin implementation

[Jan Seiffert]

2011/7/5 Mike Frysinger <vapier at gentoo.org>:

Hmm, was this mail a little bit stuck in some mail queue?

> On Monday, May 09, 2011 13:42:05 Jan Seiffert wrote:
>> I am doing them with caches turned on and with a blocksize of 160kb.
>> Which is larger then the cache. And still see an speedup factor of
>> 1.79. Or are the caches in some way prefetching in the background?
>> If i limit the size to 16kb i see a speedup factor of 2.25.
>
> the data caches dont prefetch.  i think you're seeing the 2-way aspect of
> Blackfin data caches ... while there is 32KiB of cache most the time, being a
> 2-way means things fall off a little after 16KiB.
>
> is this a tunable you can set somehow at build time ?  Blackfin systems would
> prefer 16KiB on cached setups ...
>

Oh, no, sorry, little misunderstanding.
With blocksize i mean the whole Data size under test.
This size is given by the application using zlib.
Normally i test with a Data size of 160kb, emulating an application
feeding 160kb at once into zlib.

And thats when i see a speedup of 1.79.
Which i, it seems, should _not_ see because in that case we are bound
by the memory interface because 160kb are way larger than the cache,
and the cache has no intelligence what so ever.

Exactly because the cache is 16kb i also tested with 16kb data, to get
the speed when the cache does work.
Thats when a 2.25 speedup is seen.

So, there is no tunable.

[snip]
> -mike
>
Greetings
Jan


-- 
Murphy's Law of Combat
Rule #3: "Never forget that your weapon was manufactured by the
lowest bidder"