[tremor] Tremor on x86 vs ARM, performance highlights: CPU use 120x different

[slothead]

> > I get 33% CPU use for the decoder on an ARM 200 mhz.
>
> For what's it worth, the AMD x86 test platform I build everything on is 2 CPU seconds per
> HOUR (for a GHz) instead of the ARM's 20 CPU seconds per minute (1200 CPU seconds per
> hour for 200 MHz).  That's tremor.  Scaled by equal MHz, for grins, and my StrongARM is
> 120x slower than my AMD x86.

Before you get the idea I think the machine is that much faster, realize that in debug builds it
hovers around 30% on the same x86!  Now that is weird, I thought.

To dilb:

I'm talking about the original tremor PLUS the Nicolas 'Honey I shrunk tremor in half' Pitre patch
from a week back.  The new stuff doesn't fit around the shrunken code.  I don't have your asm mdct
code which may account for the difference (33% vs 18%).  To double-check the CPU use I run a console
test version outputing to null and sure enough, same results (20 seconds time per minute of
CM).  -q10 uses more CPU than -q5 for me, about 20% more?, but that may simply be a result of all
the extra memory handling.  I don't really know.  Like most that aren't intimate with its
construction, I'm happy it works at all.

<p>I'd sure like to see the NP half-size, smaller even, version get the same treatment the base tremor
is getting, but I'm happy with what's running now too so no biggie if not.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'tremor-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.