On Wed, Feb 25, 2009 at 12:15 PM, <mplayer-users-request@xxxxxxxxxxxx>wrote: > Message: 4 > Date: Wed, 25 Feb 2009 15:59:54 +0000 (UTC) > From: Carl Eugen Hoyos <cehoyos@xxxxxxxx> > Subject: Re: debug compiles > To: mplayer-users@xxxxxxxxxxxx > Message-ID: <loom.20090225T155814-897@xxxxxxxxxxxxxx> > Content-Type: text/plain; charset=us-ascii > > Greg Bryant <bryantgtx <at> gmail.com> writes: > > > First, is there a way to disable optimizations for compiling? The > default > > made by configure has optimizations at 4, with enable-debug it goes to > -O1, > > but that still inlines code and takes away local vbls, which make my head > > hurt. If I remove optimization completely, I get the following error: > > In file included from liba52/imdct.c:725: > > liba52/imdct_3dnow.h: In function 'fft_asmb_3dnow': > > liba52/imdct_3dnow.h:257: error: can't find a register in class > > 'GENERAL_REGS' while reloading 'asm' > > liba52/imdct_3dnow.h:257: error: 'asm' operand has impossible constraints > > I believe that this is a known limitation on builds with -O0. > Fortunately, you can simply disable building liba52: > ./configure --disable-liba52-internal > > Carl Eugen Thanks Carl. I hadn't worked through all the options, so I tried that, unfortunately it just ended up moving the same problem to a different module - it seems to be an optimization issue in general between assembly and c++. It tried it with --disable-ass, with no better luck. I finally noticed that it was x264 that went to -O1 with --enable-debug, mplayer/mencoder went to -O2. I tried mplayer with -O1, and it seems to have a much better level of sanity - I can see all the buffers and even made it into the x264 library code, so I'm good to go for now. Back later with more info to seek help for dropped frames from mencoder. _______________________________________________ MPlayer-users mailing list MPlayer-users@xxxxxxxxxxxx https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
