OT: Good place to ask about XP memory problems

[dennis]

Sorry, almost anyone.

[geoff]

In message , "dennis@home" writes

Go on then, make a list

name them

What you ACTUALLY did, not a fairytale list of what you would have liked to have done

[geoff]

And about the simplest, most basic couple of lines of code one could come up with

Dennis, you are simply one of the most stupid people I have ever come across

[geoff]

Of course it is

In dennis world, you'd print it, cut it out and paste it with a prittstick

[geoff]

In message , "dennis@home" writes

Yeah - your parents

[geoff]

In message , "dennis@home" writes

If that's what your care worker is telling you, listen to them, they're right

[dennis]

Well you could never make or receive any landline calls in the UK. You couldn't use any IN services. You and your business would be pretty well stuffed. Now do your usual thing and call me a liar because you feel inadequate.

There's not much point as you are so stupid that you couldn't understand what is in telephone exchanges.

[Tim Streater]

The following appeared in a Mac newsgroup nearly 20 years ago.

How I Connected My New PC to the Network by mathew

Inspired by those Apple adverts

I unpacked the computer and plugged it in. I had paid extra to get a network card supplied with it, so I didn't have to worry about that.

I powered up the computer. It booted MS-DOS.

I found a copy of the network drivers on floppy disk. Because there are so many varieties of network, the PC manufacturer hadn't been able to supply them all. I typed COPY A:*.* C:\DOS to copy the files to my hard disk, then added two lines to my AUTOEXEC.BAT file to load them.

I rebooted the machine and typed MEM. Unfortunately, although the machine had 16MB of memory fitted, the largest space for running DOS programs was only 500K.

I ran Microsoft's MEMMAKER program. I set the various options correctly and it told me it had altered my CONFIG.SYS and AUTOEXEC.BAT files to increase the available memory. I rebooted.

It was better, but still not enough to run the programs I needed. I had heard that Quarterdeck's QEMM sometimes did a better job, so I edited my AUTOEXEC.BAT and CONFIG.SYS to remove MEMMAKER's changes, then ran QEMM OPTIMIZE.

I typed "O" in the right place to get the hidden options screen, and selected Stealth Mode. QEMM tested the ST:F and ST:M parameters, but neither of them were compatible with the hardware of my new machine. I continued with the standard OPTIMIZE process, which involved rebooting three times.

QEMM had done better than MEMMAKER, but it still wasn't quite enough. I ran the QEMM Manifest program, and noted that the memory above 640K was split into three regions, reducing the number of resident programs that could be squeezed in. I noted that the disk controller extended BIOS ROM was in the middle of an area of RAM.

I read the disk controller instruction manual, and discovered that it was possible to relocate the controller BIOS to a different address by hitting F2 as the machine was powering up. I did this, and relocated it to C8000, which Manifest had shown to be clear.

I rebooted the machine. The disk controller failed to display its start-up message. I checked the troubleshooting section of the manual, which said that this was likely the result of a ROM address clash, and suggested that I remove other cards containing BIOS ROM. I couldn't move the disk controller's address, because the code to do that was only contained in the BIOS ROM which it was failing to load.

I powered down the machine, unplugged the monitor, unscrewed the case and removed the VGA video adaptor. This wasn't easy, because of the three bus connectors; one old 8-bit type connector, a 16- bit extended connector stuck on the end as normal, and another MCA- like 32 bit connector stuck on the end of that for the VESA local bus. The VESA bus is a way of avoiding the PC's internal bus, which still runs at 8MHz even though the processor is 33MHz.

I plugged in an old monochrome adaptor card which I happened to have lying around the office. I plugged the machine into an old monochrome monitor which I also fortuitously had to hand, and powered everything up. Because I had changed the video card, the machine had zapped the setup information in the non-volatile RAM. I went into the BIOS setup program, and told it about the hard disk.

I had to tell it the disk was type 47 (user defined), and supply the number of heads, sectors and cylinders, the DMA address and IRQ number, and three other numbers whose purpose is obscure even to me. Fortunately I've played this game before, so I had written down the three screenfuls of setup information on a piece of paper. I rebooted. Still no setup prompt for the hard disk.

The only other card in the machine was the network card. I powered down, took it out and looked at it, but it was of unfamiliar design. I contacted the manufacturer's technical support line. They put me into their queueing system. After a while they phoned me back, and gave me some suggestions.

I examined the card and found that the ROM BIOS socket for the network card was empty. However, examining the tiny circuit board link marked "RO" indicated that it was connected. So the card thought there was supposed to be a ROM there. Examining the three links marked MEM revealed (with the tech support guy's help in decoding) that the card address was C8000.

Because there was no ROM, the card had not appeared to QEMM Manifest. However, it had succeeded in grabbing the addresses sufficiently well that it was interfering with the disk controller.

I carefully removed the RO link from the network card PCB, and plugged the card back in. I wasn't planning to use the ROM remote boot feature anyway. I powered up. This time, hurrah, I got the disk setup prompt.

Having removed the BIOS address conflict, I powered down once more, removed the monochrome card and replaced the S3 VGA card. I connected the Trinitron monitor, powered up again, and entered the hard disk details into the setup program a second time.

Finally I had the machine displaying all the setup prompts on bootup. I ran the QEMM optimize process again. This time, with only 2 regions in the upper memory block, it managed to load nearly all the network drivers and other memory-resident programs into high memory, leaving me with about 600K for DOS programs.

I checked the memory map with Manifest again. There was still some free space in the upper memory. I went ahead and installed Windows for Workgroups - mostly a plodding process, except that it complained about a possible IRQ contention on LPT2:. In fact, the machine had no LPT2:, it was really a network print device remapped to LPT2:, so I ignored the dire warning.

Once I had Windows for Workgroups installed, I exited it and rebooted. W4W had its own network drivers, so I had to run the QEMM optimise process again, rebooting three times. Once it had finished the result was much the same as before installing W4W.

I examined the CONFIG.SYS and AUTOEXEC.BAT files. I noticed that QEMM had set MSIPX.SYS to load high in CONFIG.SYS; I knew that this was unnecessary, as MSIPX.SYS loads high on its own, and using QEMM to load it high actually wastes memory.

I also noticed that the KEYB UK line (which loads the keyboard driver) was near the end of the AUTOEXEC.BAT, and was therefore failing to load high. I knew from experience that KEY needs lots of memory to load but releases almost all of it after loading. I moved it to the start of AUTOEXEC.BAT, just before SMARTDRV, and rebooted.

Everything worked. All the memory resident programs loaded into high memory, leaving me with over 600K for DOS programs.

I still wasn't quite finished, though. Windows would not work in enhanced mode until I altered its SYSTEM.INI file. I loaded the file, found the [386enh] section, and added a line saying "device=monoumb2.386". I copied the monoumb2.386 file to WINDOWS\SYSTEM.

I then ran Windows for Workgroups, and set up my network settings and printer settings appropriately, following procedures just like those described in the Apple adverts.

It was really easy.

mathew

[ And people ask me why I bought a Mac for my home machine... Still, I should thank Microsoft and Intel for keeping consultants like me in gainful employment. ]

[The Natural Philosopher]

everyday for me at one time..

[John Rumm]

Don't think it actually was true single pass, but it did follow in Turbo Pascal's footsteps of being very small and tight and very quick (probably coded in assembler (TP up to version 5 was), with integrated editor etc). Hence it was much faster to compile than any of the competition. It also managed to keep a fair amount of program source memory resident, and hence could skip some of the multiple file read activities that many C compilers do.

Don't be so hard on Borland. After all, MS liked their products so much they pilfered their chief software designer.

[John Rumm]

Not sure about the step to 4 bit even then...

[John Rumm]

The pipeline (well prefetch queue - might be a bit OTT to call it a pipeline) was 4 bytes on the 8088, and 6 bytes on the 8086 (and I think NEC copied this - although would not swear to it - never actually coded for them). Hence there was a way you could sense which processor you were running on, using a bit of self modifying code to overwrite an (e.g.) and INC with a NOP just ahead of the current execution position. The 8088 wound not execute the INC but the 8086 would because it was already prefetched in the deeper queue.

Could never bring myself round to buying a PC until the 486 appeared!

[John Rumm]

Yup, I remember having a rather confused conversation with someone at a job fair a couple of decades ago. She wanted Assembler programmers according to the billboard, so I gave her a CV to look at. After a few mins study she said she could not find any mention of assembler! I had to explain what 6502, Z80, NSC800, 68K, 80x86 and various others under the title of "Low level programming" meant. She did not seem convinced when I pointed out that by the time you know one or two low level languages, you can pick up others well enough to be productive in a few days. ;-)

[Clive George]

I think a C compiler ought to be written in C :-)

(it's a traditional test for them - compile the compiler, compile again with the new one, see if there are any differences)

[John Rumm]

Yes good point... DX3 was supposed to be a 2.5x multiple version they never actually introduced in the end.

I suppose the justification would be "its just the next number"...

[John Rumm]

Don't think I have ever seen a 288... they may have been mythical.

[John Rumm]

[snip]

I remember pulling the ROM out of an Adaptec AVA 1525 SCSI card being used to talk to a CD ROM drive for just such a purpose ;-)

[The Natural Philosopher]

"If the blind lead the blind, they both shall fall in to the ditch"

[Bob Eager]

Yes, I remember that. And yes, I thought 'pipeline' wasn't the right word, but I was tired!

I wrote an early book on DOS, and used that machine to test a lot of it...

[Bob Martin]

You "concluded" after being "introduced" to it?

I programmed IBM mainframes in assembler almost every day from 1965 to past 2000, during which time I also used every other IBM instruction set plus x86 and TI

9xxx, and nothing came near 360 for effectiveness, ease of use, maintainability etc. I never met a 360 programmer who thought otherwise.