Tom's Guide | Tom's Hardware | Tom's Games
![]() |
![]() |
![]() |
Dear Friend,
I have an executable MS-DOS program related with finite element modelling. It was originally compiled using the Lahey F77L-EM/32 Fortran compiler. I have some problems
related with the program which I have written below:1- I can't operate the program using MS-Dos prompt under windows 98SE. It gives errors during the solution time.
2- While using it under Ms-Dos 7.1, the solution time can reach many many hours for big sized problems.In its manual it is written that the program is compatible with the QEMM expanded memory manager program and can be run under the DESQView-386 multitasking environment. I don't know much about QEMM and Deskview programs.
I want to learn if I can manage to shorten the solution time using my computer's capacity (Celeron 1300, 384MB Extended memory).I am writing the system requirement of the program below. If you can help me, I'll be very pleased.
Best regards...
System Requirement:
80386/80486 based IBM-AT/PS2 or 100% compatible computer. If the CPU
is 80386, the machine must have the 80387 math coprocessor.MS-DOS 3.30 or later.
At least 1MB extended memory available excluding all device drivers and the 640KB DOS memory. Depending on the problem size, more memory may be required. This program is compatible with QEMM expanded memory
manager and can be run under the DESQView-386 multitasking environment.
This program can also run in a DOS window under MS-WINDOWS 3.x.
(For MS-DOS 5.0 users: If you want to run the program under DESQView,you must have QEMM-386 version 6.00 or later)At least 5MB of free hard disk space. You may need more free disk
space for large problems.

About System Requirements:
All Pentium based processors and like (as AMD) are compatible and very more powerfull than the legacy 386/486 family and have builtin in the chip the math coprocessor once a time to buy separately as optional (387). So your Celeron 1300 is perfectly in line with the specification you post for the CPU.
DesqWiev and QEMM386 are an old multitasking shell and an old expanded memory manager whose functionality you get as a standard service running Windows 9X or later (QEMM386 services are managed by Microsoft EMM386).
So what do you have to try? First give a more detailed description of your run-time problem at Dos prompt: error message, Dos box under Windows or other operating environment.
May be to optimize your code a new compilation starting from source is required setting some parameter; at the moment you can try to enlarge the size of XMS memory and DPMI memory in the Memory Section of Properties related to your Dos window (supposed you are running the executable in that way). An initial value may be 8192.
Sorry to be not full exaustive, but from your post it is hard to go far away, as the lane is surrounded by fog.

In short:
You have a fast processor and a large amount of RAM.
1. Boot with a MSDOS 7.1 boot diskette. Then no Windows overhead is loaded consuming processor power.
2. Config a lot of extended memory(=XMS) and raw extended memory (=/Int15) ie. 20 Mb each in EMM386
3. Make a large RAMDrive (ie. 20 Mb), copy the program and data files to the RAMDrive and run the program from that drive.
4. Use of cache may help dramatically if used on a low powered 386/486 but should not increase the throughput on your computer with so much RAM. Check your harddisk light is off during execution.Keep us posted
Svend

Certain original dos programs will not run properly or at all, using the embedded dos versions supplied as part of windows. This may well be true when using an old fortran program.
You just have to run using dos (or find a windows equvalent fortran compiler/emulator). Boot up any pc, using dos version 6.2x to check it runs the fortran program ok, then optimise to process as fast as possible. Some of the above tips are fine, however I would suggest trying one at a time. A general rule is that, the larger the memory (also faster cpu, space on hdd) the faster the processing.
It was my understanding at the time QEMM was a memory manager that allowed multi programming and became redundant when windows appeared.
You have to have a memory manager, so the ms HIMEMM is fine. It is unlikely you will require EMM386, as this is for expanded memory. When using a 386, I feel its matching 387 will help.
Good luck - Keep us posted.

Mike Newcomb,
"Certain original dos programs will not run properly or at all, using the embedded dos versions supplied as part of windows ... Boot up any pc, using dos version 6.2x to check it runs the fortran program ok"
This time you are totally wrong.
1: He is perhaps talking about the standalone and complete version of MS-DOS 7.10 from:
http://newdos.yginfo.net/msdos71.htm
2: He has already said "While using it under Ms-Dos 7.1, the solution time can reach many many hours for big sized problems.". So, Mike, please don't give the wrong information!

Wengier - if you disagree with my reply, fine. However it is my opinion that it is correct, based on my own experience, and may well help to overcome Emre's problem.
I have in excess of 30 years in computing including commercial, scientific and control, and yes, there is still a vast amount I do not know.
Computing.net is about helping each other, not nit-picking amongst ourselves. If Emre was close to me, I would love checking out his set up personally as it is some time since I used fortran.Going back to dos, there were a number of versions that had odd quirks (e.g 4 + 5), whereas 3 and 6 were pretty good.
Good luck - Keep us posted.

Mike,
"if you disagree with my reply, fine. However it is my opinion that it is correct, based on my own experience, and may well help to overcome Emre's problem. I have in excess of 30 years in computing including commercial, scientific and control, and yes, there is still a vast amount I do not know."
Nice to meet you here. And I'm sorry if you feel that.
His problem could be caused by any of the following reasons:
1: Incorrect/inappropriate settings in CONFIG.SYS/AUTOEXEC.BAT;
2: Some TSR/drivers/programs which are incompatible with his program are loaded in CONFIG.SYS/AUTOEXEC.BAT or so. Or maybe the versions of QEMM386 and/or DesqView he is using are not incompatible with his program.
3: The extra codes in some IO.SYS. By this I need to say the different between "WinDOS 7.1" (a special version of MS-DOS 7.10 embedded in Win9x) and standard MS-DOS 7.10. Let's take Soft-ICE 2.8 as an example. Soft-ICE 2.8 will NOT work properly on "WinDOS 7.1" due to some extra codes in IO.SYS, but it works great in standard version of MS-DOS 7.10 from "MS-DOS 7.10 Full Version". So it could be a cause if he is actually talking about "WinDOS 7.1".
However, more detailed information (like the error message he saw) is needed.
Keep us posted.

Wengier - it appeared to me that Emre has two problems:-
a) program does not run
b) wants to speed up programI feel the best approach is to tackle them separately. If one makes too many steps at once, should there be a subsequent problem it then is not so easy to locate. I take on board, and accept totally your last observations.
I would like to know how Emre's program performed on a very fast 486DX with maximum memory. Also I would like to benchmark any pc Emre uses and then compare his program's run time on that pc.
Should a later compiler be used, it may require some (probably minor) changes to the source program.
So many old dos programs are not forward compatible, or only partially so. I have seen some fail due to too much memory would you believe.
Using non-ms software (e.g. QEMM) may again cloud the issue.
We used to call some scientific programs number crunchers, as they would process for hours then produce the result(s) - just one big sum really.
I disagree totally when I hear that computers cannot make mistakes, as I personally experienced the opposite. With one huge scientific package, when run in test/validate mode it produced a different result on our computer, to that on the same computer elsewhere (I am talking mainframes). In the end, it turned out there was a fault in one of the mathmatical circuits.
Regards -Mike
Good luck - Keep us posted.

Dear Friends,
Thank you for helpful advices. I applied all the given recommendations. I changed the config.sys file, formed a ramdrive with capacity of 32 MB and started from
a boot disk. But I couldn't succeed to operate the system either in the ms-dos window under windows98 or in a short time under ms-dos7.1. First of all, working with the program in a ramdrive decreased the
operation time to the 10th % of the former for small sized problems (nearly 30 elements). For big sized problems
of 1700 elements it decreased only the 75th % of the former. And for problems having more than 2200 elements it is a complete disappointment because it doesn't work.Finally I decided to make modifications to the program using current Compaq Visual Fortran. By changing some of the
commands of the program concordant with the old lahey F77L-Em32 fortran, I think I may succeed operating it efficiently.Best regards...

Some very old DOS programs do have forward compatibility problem, like BCAD 2.0, which doesn't work correctly if the XMS memory of the system is greater than 64MB. So it's a good idea to use a later or updated version of the compiler.
Good luck!

![]() |
![]() |
![]() |

This post is quite old and has been locked from receiving new replies. Please create a new posting instead.
| Ads by Google |