Date: Tue, 6 Jul 1999 12:11:45 +0200 (CEST) From: Niteshadow To: pgcc AT delorie DOT com Subject: Re: -O2 versus -O1 In-Reply-To: <19990705163430.C3137@cerebro.laendle> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Reply-To: pgcc AT delorie DOT com X-Mailing-List: pgcc AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk On Mon, 5 Jul 1999, Marc Lehmann wrote: > On Fri, Jul 02, 1999 at 07:51:22PM +0200, Niteshadow wrote: > > I just downloaded the bladeenc 0.81 sources and tried to tweak it > > to the max. Well... tried... for some reason when I changed the original > > - -O1 in the make rule for codec.o to -O2, the bladeenc started to create > > MP3s with DIFFERENT data (compared using md5sum). I tried it only at work > > on a Celeron 300A/128MB Ram/PGCC 1.1.3/Linux Mandrake 6.0/kernel 2.2.10 > > Options like -mpentiumpro, -malign-double, -march=pentiumpro did > > not cause this to happen. Could anyone give me a clue about what could > > cause this kind of ``feature'' ??? > > Dumb question: were the resulting streams correct? Rounding differences are > something you might expect (doesn't bladeenc use -ffast-math by default?) I do not know, whether the streams where correct. I was just trying to find out, whether the documentation was correct (it said something about gcc corrupting the output stream), since I thought this could not happen, if you use the _right_ optimisations. I am walking on thin ice here.. I do not think, that specifying -O2 should introduce rounding errors, but I am not all-too familiar with the way compiler does all optimisations. Yet I think that as long as one does NOT specify -ffast-math (which bladeenc doesn't), there should not be any round-ff errors introduced by any of the optimisation options. The implicit compiler-options for bladeenc are: for the codec engine: -O1 for the rest of code: -O2 -malign-loops=2 -malign-jumps=2 -funroll-all-loops > > -- > -----==- | > ----==-- _ | > ---==---(_)__ __ ____ __ Marc Lehmann +-- > --==---/ / _ \/ // /\ \/ / pcg AT goof DOT com |e| > -=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE --+ > The choice of a GNU generation | > | > ------------------------------------------------------------------------------ From : Robert 'Niteshadow' Varga email : niteshadow AT hq DOT alert DOT sk, varga AT ibl DOT sk PGPkey: http://hq.alert.sk/~niteshadow/pgpkey.txt phone : +421 903 511 985 (mobile, always), +421 863 432 49 75 (home, rarely) ------------------------------------------------------------------------------ We are Linux. Resistance means you didn't get the point.