October 2008 - ffmpeg-devel - lists.ffmpeg.org

[PATCH]Fix vorbis decoder memlead on corrupt input
by cehoyos＠rainbow.studorg.tuwien.ac.at 01 Oct '08

01 Oct '08

Hi! I just found this patch in my local tree. AFAIR, it fixes a memleak on currupt input (which I cannot find atm). Is it ok? Carl Eugen

3 2

[PATCH 0/3] A couple more fixes (not just for Solaris)
by flameeyes＠gmail.com 01 Oct '08

01 Oct '08

I'm sending three more (little) patches I prepared today while trying to get FFmpeg to build under OpenSolaris (and I think I finally accomplished it!), just one is strictly Solaris-related: forcing PIC on when building the libraries (otherwise there will be problems with linking, using GNU ld). The other two patches are quite Solaris-independent: one is a trivial reduction of warnings by fixing two format string sin ffserver.c, while the other, needed for OpenSolaris to work, adds the extra library object flags (that is, if needed, PIC flags) before checking for %ebx (to avoid false positives). HTH, Diego --- Diego 'Flameeyes' Petten? (3): Remove some format string warnings from ffserver.c Use LIBOBJFLAGS when checking if %ebx is available. Always create PIC objects for shared libraries under Solaris. configure | 1 + ffserver.c | 6 +++--- 2 files changed, 4 insertions(+), 3 deletions(-)

4 9

[PATCH] vp3.c: Don't compile a debug loop
by lessen42＠gmail.com 01 Oct '08

01 Oct '08

Hi, gcc 4.0.1 isn't smart enough to optimize out this loop when debug_idct is defined to nothing. This forces the loop to not be compiled when it does nothing, for an overall speedup of about 1%.

3 11

[PATCH] vp3: Reduce size of filter_limit tables
by lessen42＠gmail.com 01 Oct '08

01 Oct '08

Hi, The filter limit tables only use 7 bits per element in the Theora spec and the largest VP3 value is 30. This uses uint8_t for these tables rather than uint32_t.

2 2

[PATCH] Fix pkg-config -uninstalled.pc when static libraries are disabled.
by flameeyes＠gmail.com 01 Oct '08

01 Oct '08

The -uninstalled.pc pkg-config files rely on the presence of the static version of the libraries when linking to it, but they might be disabled. When the static version of the libraries is disabled, use the shared version by just giving the search path and the library name, so that it still works. --- configure | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/configure b/configure index 8b54550..0fb0721 100755 --- a/configure +++ b/configure @@ -2261,7 +2261,8 @@ Description: $comment Version: $version Requires: $requires Conflicts: -Libs: \${libdir}/${LIBPREF}${shortname}${LIBSUF} $libs +Libs: $(enabled static && echo "\${libdir}/${LIBPREF}${shortname}${LIBSUF} $libs") $(disabled static && echo "-L\${libdir} -l${shortname}") +Libs.private: $(disabled static && echo "$libs") Cflags: -I\${includedir} EOF }

2 2

The fate of Fate
by stefano.sabatini-lala＠poste.it 01 Oct '08

01 Oct '08

Hi all, today I tried to acces to the FATE page at http://fate.multimedia.cx/ and with my great surprire I read: "Closed due to lack of interest." Looks like a joke, I always found fate greatly useful, sad if it has been really closed. Can someone cast a shed of light on the issue (Mike)? Regards. -- FFmpeg = Frenzy & Frightening Minimalistic Powered Experimenting Gospel

11 29

[PATCH] ffvorbis, better L1 cache use and simplification of code
by siarhei.siamashka＠gmail.com 01 Oct '08

01 Oct '08

Hi, Interleaved forward/backward channels processing in order to increase chances of stepping on already cached data for the cores with extremely small data cache. Ensure that IMDCT per-rotation does not introduce cache write misses (write misses on random memory accesses are bad for ARM cores with no write-allocate cache as they prevent combining data in the write buffer). ARM11 with 32K of L1 data cache (no L2) shows performance improvement in the range 0.5-1% which is not so bad considering that IMDCT/IFFT and also many other important dsputil functions are not assembly optimized for it yet. According to cachegrind simulation, there might be also some improvement for x86 cores with 16K of L1 data cache (decrease of the number of cache misses is most visible in this configuration). -- Best regards, Siarhei Siamashka

3 4

Status of frame-based multithreading
by 2manybills＠gmail.com 01 Oct '08

01 Oct '08

Hi, everybody. What's the current status/activity on Alexander Strange's frame-based multithreading work? Is it going to be committed? I'm a bit rusty in C/C++ but I've been pocking around ffmpeg code to get acquainted with it and see if I can be useful somehow. Thanks.

1 0