X-POP3-Rcpt: mlehmann AT universe DOT sgh-net DOT de Message-ID: <19980125020825.22422@cerebro.laendle> 25 Jan 1998 02:08:25 +0100 From: Marc Lehmann To: Dustin Marquess Cc: beastium Subject: Re: -malign-double & zgv References: <199801240204 DOT VAA29208 AT ppp-x9-16.> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.88 In-Reply-To: ; from Dustin Marquess on Sat, Jan 24, 1998 at 10:46:45AM -0600 X-Operating-System: Linux version 2.1.79 (root AT cerebro) (gcc version pgcc-2.91.03 971225 (gcc-2.8.0)) Status: RO Content-Length: 1701 Lines: 35 On Sat, Jan 24, 1998 at 10:46:45AM -0600, Dustin Marquess wrote: > On Fri, 23 Jan 1998, Dinesh Somasekhar wrote: > > > Hi, > > > > You were right, -mamdk6 seems to the best for the cyrix > > 6x86. Surprisingly -mamdk6 and -mamdk6 -fno-risc bench > > identically. The -mamdk6 opt fixes core dumps (SEGV). Almost every > > one of the results is better except for FP emulation, and the float > > benchmarks. Maybe you should just set -mcyrix to -mamdk6 :-). > > > > f) pgcc -mamdk6 -O6 -malign-double -malign-loops=4 -malign-jumps=4 > > -malign-functions=4 -Wall -Wno-unused > > Hmmm... When compiling zgv on my 6x86/200, zgv won't run right > when compiled with -malign-double. It doesn't seem to mind the other > switches though. -malign-double breaks the abi... libjpeg for example is sensitive to this, so you have to recompile libjpeg with -malign-double, too (and ALL applications using it as a shared library), or alternatively get rid of the -malign-double, the speed shouldn't be affected (in zgv). In my system, I compiled applications myself, so I normally can specify this flag safely. Recently, though, I recompiled xv, gimp & the hdf libraries without -malign-double... how do I shoot myself into the foot? -----==- | ----==-- _ | ---==---(_)__ __ ____ __ Marc Lehmann +-- --==---/ / _ \/ // /\ \/ / pcg AT goof DOT com |e| -=====/_/_//_/\_,_/ /_/\_\ --+ The choice of a GNU generation | |