From: jenkinsj AT cuug DOT ab DOT ca (Jim Jenkins) Subject: DPMI problems under version 1.12 To: djgpp AT sun DOT soe DOT clarkson DOT edu Date: Fri, 12 Aug 1994 02:51:45 -0600 (MDT) -----BEGIN PGP SIGNED MESSAGE----- I have been having trouble getting the new 1.12 version of DJGPP to run at a reasonable speed under DPMI. I know this is a FAQ, so I looked in the FAQ list, and it suggested I do several things, some of which did help. Unfortunately, I could not solve the problem. For those who might be able to, however, here are what I believe to be the relevant details: I have a 386 computer with a 387 and 8 MB of memory installed, running MS-DOS 5.0. In my AUTOEXEC.BAT file, I have installed SmartDrive and Doskey: ...stuff... c:\dos\doskey.com /insert c:\windows\smartdrv.exe a- b- c 1024 512 ...more stuff... Many environmentals are defined here and there: COMSPEC=C:\COMMAND.COM PROMPT=$p$g BLASTER=A220 I5 D1 T3 DIRCMD=/ogn TEMP=c:/temp TMP=c:/temp TMPDIR=c:/temp TZ=MST7MDT DJGPP=c:/lang/djgpp/djgpp.env (I haven't changed this file at all) GO32TMP=c:/lang/djgpp/tmp PATH=c:\batch;c:\dos;c:\util;c:\lang\tc;c:/lang/djgpp/bin GO32=driver c:/lang/djgpp/drivers/trident.grd gw 640 gh 480 As a contrast, my CONFIG.SYS file looks quite bare: buffers=30 files=40 device=c:\windows\himem.sys dos=high shell=c:\command.com /e:1024 /p stacks=9,256 This is what go32.exe reports about my system: go32 version 1.12 Copyright (C) 1994 DJ Delorie Lowest version I can run is 1.08 go32.exe usage: go32 [-d {debugger}] [{program} [{options} . . . ]] go32.exe build time was Thu Aug 04 19:52:27 1994 XMS memory available: 6076 Kb Swap space available: 25728 Kb Nevertheless, when I set the topline and/or nodpmi flags in the GO32 environmental, it takes many seconds (>30) even to compile and link a program as simple as hello.c. The topline display shows that the XMS memory is eaten at a ridiculous rate by the various programs called by the gcc loader, and 120-170 Kb of swap space is used. This seems very strange to me, since if I configure my system to use emm386.exe and VCPI instead of DPMI, gcc and the programs it spawns don't usually eat all of available RAM when doing a typical compile. On the other hand, if I do that, I can't spawn the real-mode version of GNU Make (version 3.71, I believe) from GNU Emacs 19.24, which I can do from DPMI. It's very slow, of course, but it works. It worked much better in version 1.11.maint5, and I request help in figuring out why. Yours, perplexed, Jim Jenkins -----BEGIN PGP SIGNATURE----- Version: 2.6 iQCVAgUBLks4FXEx0BaTGk6ZAQHWIAP9GyPh809vNafAtkM4jANE8owLTchY120E 3NiuLjrzpSZ2t2YSUcUffIqk14/NEYbQMyDhe0hNcHnzJR0hdfxNCIVC+i/7ARGF py2McgJKf1oNdEgeP/sFU9iLfSzeV8HVm8cdazXzKcEl+5DAQnLcu5Accew7k91r CPDXYxfIylY= =5rk3 -----END PGP SIGNATURE-----