[FFmpeg-devel] [PATCH 2/2] Move MDCT function pointers to MDCTContext

[Måns Rullgård]

M?ns Rullg?rd <mans at mansr.com> writes:

> Michael Niedermayer <michaelni at gmx.at> writes:
>
>> On Thu, Sep 17, 2009 at 12:51:06AM +0100, M?ns Rullg?rd wrote:
>>> Michael Niedermayer <michaelni at gmx.at> writes:
>>> 
>>> > On Thu, Sep 17, 2009 at 12:33:11AM +0100, M?ns Rullg?rd wrote:
>>> >> Mans Rullgard <mans at mansr.com> writes:
>>> >> 
>>> >> > ---
>>> >> >  libavcodec/arm/fft_init_arm.c |    6 +++---
>>> >> >  libavcodec/dsputil.h          |   12 ++++++------
>>> >> >  libavcodec/mdct.c             |    6 +++---
>>> >> >  libavcodec/x86/fft.c          |   12 ++++++------
>>> >> >  4 files changed, 18 insertions(+), 18 deletions(-)
>>> >> 
>>> >> Ping.  What about this, assuming the first patch is OK?
>>> >
>>> > Its the first patch iam uncertain about
>>> > it does " 41 insertions(+), 14 deletions(-)"
>>> > which is it makes the code bigger and iam not sure what that is good
>>> > for ..
>>> 
>>> I thought it made sense in preparation for the patch I just sent,
>>> which will then be followed by another patch making the NEON MDCT a
>>> few % faster.  Feel free to suggest a better way.
>>
>> what is the problem with just not spliting it?
>
> fft_init() works (rightly) on FFTContext, and I need to change things
> in MDCTContext.  I can't make the arch init calls from mdct_init since
> fft can be used without mdct.  Besides, having the mdct function
> pointers in FFTContext seems like a mistake to begin with.
>
> What I ultimately need is for the cos/sin tables in MDCTContext to be
> interleaved as per another patch.

Loren suggested merging the two structs instead.  The attached patch
does that.  Is this more palatable?

-- 
M?ns Rullg?rd
mans at mansr.com