[Tremor] Help required to compute mips for tremor (int ogg-vorbis

[Madhav]

Hi Ripinder,Thanks for the reply, My doubt about this MIPS is clear. The tremor codeon arm for 1st frame takes 60 MIPS and then onward it takes 6 MIPS as worst.These figures seems to be quite good for arm platform if we not considerthe 1st frame cycles. Anyway I have done the detail profiling for 1st frame & identified 3 functionsin _make_decode_ready ( ) I.e. init routine as most cycles taking modules:1. _make_words ( ).2. quick_sort ( ).3. _book_unquantize ( ).I have optimized _make_words ( ) function for 5 odd MIPS. Now I am looking for other functions. Out of this quick sort is called a c library. I want the csource code for this standard library. I am not able to check its functionality.Can you help me in this?With Best Regards,Madhav Kulkarni=======================================================================Well,

Since these cycles need to be spent only once during the start of a
new song, I suppose we should not include them in the worst case
calculation. However if one need to reduce the initial cycle delay,
you may reduce these cycles also. The only thing to be noted here is
that this is a static initialization and so far the real time process
may not be started. Only after the first call should the real-time
processing be started and thereafter one can-not miss the realtime
requirements (worst case MIPS).

Ripinder


On Tue, 8 Mar 2005 18:32:11 -0800, Madhav <madhav at mtekvision.com> wrote:
>  Hi,I understand about this MIPS calculation but I surprise that evenif we
> make the 1st call static these cycles will contribute as the worst case
> figure.So I started optimizing some intitialization routines to reduce the
> 1st frame MIPS.Is it the correct step?With regards,Madhav
> Kulkarni======================================================================================Hi,

--------------------------------------------------------------------------------


  a.. Previous message: [Tremor] Help required to compute mips for tremor (int ogg-vorbis 
  b.. Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 

--------------------------------------------------------------------------------
More information about the Tremor mailing list
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/tremor/attachments/20050311/3376c54e/attachment.html