[XviD-devel] [CVS commits] head -- Stable merge + decoder optims

[Michael Militzer]

Hi,

I suppose the 64 bit change affects timecodes (as these are the only ones
where we use 64 bit variables in the code (apart from 2pass code) afaik).
I admit I haven't looked at the actual code changes yet but I remember
there was some reason for switching to 64 bit originally. Ok, I again 
assume you only _not_ use 64 bit timecodes for the calculation of B-direct
vectors (and this may be ok), but I'd like to have Christoph a look at 
this because I believe he originally wrote the timecode support and also
originally introduced the 64 bit arithmetics. So he should know why he
initially did this ;-))

bye,
Michael

Quoting Edouard Gomez <ed.gomez at free.fr>:

> As the cvs is up again, here are some commits:
> 
> 2004-05-21 14:32:41 GMT		patch-28
> 
>     Summary:
>       Don't set edges twice on a frame. 
>     Revision:
>       xvidcore--head--0.0--patch-28
> 
>     From ed.gomez:
>      * A similar optimization has been done for encoder long ago,
>        dunno why this hasn't been "ported" to decoder. This speeds
>        up quite a lot the decoder for no effort (~7%).
> 
>     modified files:
>      src/decoder.c src/decoder.h
> 
> 
> 2004-05-21 14:25:19 GMT		patch-27
> 
>     Summary:
>       No 64 bit arithmetic in critical path. 
>     Revision:
>       xvidcore--head--0.0--patch-27
> 
>     From ed.gomez:
>      * No 64 bit arithmetic in critical paths (direct blocks in bvops), it's
>        way too slow (__divdi3 GNU/Linux ABI for 64bit division was taking
>        up to 5% cycles)
> 
>     modified files:
>      src/decoder.c
> 
> 
> 2004-05-15 22:20:11 GMT		patch-26
> 
>     Summary:
>       Merged stable tree changes
>     Revision:
>       xvidcore--head--0.0--patch-26
> 
>     Merged stable tree changes
>     
>     Patches applied:
>     
>      * ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-23
>        Some very light Unix build system changes 
>     
>      * ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-24
>        Possible VOL header corruption. 
>     
>      * ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-25
>        DC prediction fix. 
>     
>      * ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-26
>        Small mismatch in hint<->widget in VFW
>     
>      * ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-27
>        Marking 1.0.0 final
>     
> 
>     modified files:
>      ChangeLog build/generic/Makefile build/generic/bootstrap.sh
>      build/generic/configure.in src/bitstream/bitstream.c
>      src/decoder.c src/motion/estimation_rd_based.c
>      src/prediction/mbprediction.c src/prediction/mbprediction.h
>      src/xvid.h vfw/src/resource.rc
> 
>     new patches:
>      ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-23
>      ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-24
>      ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-25
>      ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-26
>      ed.gomez at free.fr--2004-1/xvidcore--stable--1.0--patch-27
> 
> -- 
> Edouard Gomez
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel
>