From: James MacDonald Newsgroups: comp.os.msdos.djgpp Subject: Re: DJGPP on a 386 Date: Sat, 14 Jun 1997 13:46:03 +0100 Organization: Trills and Technologies Distribution: world Message-ID: References: <33a1d6f2 DOT 4203083 AT news DOT cis DOT yale DOT edu> NNTP-Posting-Host: netbook.demon.co.uk MIME-Version: 1.0 Lines: 29 To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp Precedence: bulk In article <33a1d6f2 DOT 4203083 AT news DOT cis DOT yale DOT edu>, mapson scribbled : >I have DJGPP on a pentium at work, a 486 at home, and yesterday I put >it on a spare 10-year old 386 with just under 900k memory. ^^^^^^^^^^^^ Come on! GCC needs much more memory than that. So CWSDPMI has to allocate a swapfile. It takes much more time to access this than real memory, therefore the slowdown. >Just curious, what is it about this 386 system that makes it so much >slower than just the slower clock speed would indicate? The lack of >memory? It certainly runs the compiled programs at the expected rate. Sure, but I have a network of 4 machines (IBM PS/2 16Mb RAM, Acorn A3000 8Mb RAM, P75 24Mb RAM, P166 64Mb RAM) and the time that it takes to compile on the 486 is huge, as is the GCC port on the Acorn. If I compile, link and run hello.c on my P75 it takes about 10 seconds, and they'll run at the same speed on the 486. The memory bottleneck is the killer. 'Nuff said. > -- Revised anti-spam in use : remove X to reply - 'Xnetbook' becomes 'netbook' Anti-spam thermonuclear warheads cheap at only $300!