[Tremor] LOW_ACCURACY mode and MDCT improvements

[Mike Giacomelli]

----------------------------------------
> Date: Tue, 23 Mar 2010 13:07:39 -0400
> From: dkg at fifthhorseman.net
> To: giac2000 at hotmail.com; tremor at xiph.org
> Subject: Re: [Tremor] LOW_ACCURACY mode and MDCT improvements
>
> [back on-list, i hope that's OK]
>
> On 03/23/2010 01:00 PM, Mike Giacomelli wrote:
>> Is _ARM_ASM defined? Which -O2 right?
>
> i'm using -O2.
>
> i don't see any mention of _ARM_ASM in either the source or the build
> logs. Where should i be looking for it? Where should i be setting it?
>
>> FWIW I saw a speed up on x86-64, ARM922 and Cortex A8. It was smaller then in rockbox
>> but still several MHz in all cases. I'm not sure why Rockbox seemed to benefit more, its
>> possible we programmed towards our specific gcc version (4.0.3) a little too much and that
>> tweaking the code could further improve things.
>>
>> Perhaps I should also port our ASM version of the current MDCT and see if that works better.
>
> I'm up for trying other changes to see if i can coax a speedup out of
> it, if you can point me in the right direction.
>
> --dkg
>

I asked on rockbox IRC:

[20:55] <stripwax> saratoga - hm, on a default compile (on an arm platform -  here, nslu2) it doesn't seem to define _ARM_ASSEM_ .  That might just be me though.  But that's not a particularly exotic platform (i think, anyway)


So it seems you will have to either fix the makefile for your target or manually define that in the make flags. 		 	   		  
_________________________________________________________________
Hotmail has tools for the New Busy. Search, chat and e-mail from your inbox.
http://www.windowslive.com/campaign/thenewbusy?ocid=PID27925::T:WLMTAGL:ON:WL:en-US:WM_HMP:032010_1