delorie.com/archives/browse.cgi | search |
From: | "John S. Fine" <johnfine AT erols DOT com> |
Newsgroups: | comp.os.msdos.djgpp |
Subject: | Re: Optimizations |
Date: | Sun, 20 Sep 1998 13:01:34 -0400 |
Lines: | 30 |
Message-ID: | <360534EE.4241@erols.com> |
References: | <Pine DOT SUN DOT 3 DOT 91 DOT 980920103343 DOT 406G-100000 AT is> |
Reply-To: | johnfine AT erols DOT com |
NNTP-Posting-Host: | 207-172-240-213.s22.as4.bsd.erols.com |
Mime-Version: | 1.0 |
To: | djgpp AT delorie DOT com |
DJ-Gateway: | from newsgroup comp.os.msdos.djgpp |
Eli Zaretskii wrote: > > I saw nothing relevant and nothing that addressed the > > NOPs inserted for alignment. > > How about -m386 and -m486? In 2.7.2.1 -m386 had no effect (I assume it is the default). -m486 only made things worse. It changed the alignment from 2 to 4. (I finally noticed the -malign switches. I had been looking under optimization switches rather than under Intel 386 switches). It changed every "leave" to "mov %ebp,%esp pop %ebp". It changed every push of a memory location to a mov to a register followed by a push of the register (even when that use of a register made nearby code worse for lack of a register. I changed to 2.8.1 (I may be looking for advice on changing back). It generated slightly worse code. The -m386 -m486 and even -mpentium switches now make no difference at all. I must be doing something wrong, but I can't guess what it is. -- http://www.erols.com/johnfine/ http://www.geocities.com/SiliconValley/Peaks/8600/
webmaster | delorie software privacy |
Copyright © 2019 by DJ Delorie | Updated Jul 2019 |