PDP-11 Booting

One of my first summer intern jobs was recreating the boot deck for a PDP 8. The first card in the deck had become damaged and there was no backup. We had an old assembler listing of the boot program but no assembler. I did not know PDP assembly but knew 6502, Z80 and 370. I had done all my Z80 writing machine code directly as I had no assembler there either. So I read the PDP docs and hand assembled the first card. It was a relief when I found that the first instruction on the second card lined up as did the next several and the last instruction on the second card. We were pretty sure we had the right deck. I powered on the PDP, keyed the boot sequence, hit run and the deck loaded. The ready light came on and I loaded the next deck, the copy deck program. It loaded successfully as well and I reloaded the boot deck in to the card reader. Pressing the start button was a moment of suspense but when the card punch started clacking out a duplicate deck I was able to breathe a sigh of relief.

We had talked at lunch that day about making a special boot deck whose sole purpose would be to punch a replacement standard boot deck, but it seemed too ambitious for something rarely needed.

Asking a few years later, the boot deck had lasted until the machine had been retired. The backup deck had never been used

PDP 11/34 had a bootrom and an octal keyboard, which is kind of cheating.

I grew up using the older PDP 11/45 where you had to key in (using binary toggles) a boot program. Or, the way I did it, manually enter write address and byte count to the RK05 disk controller (memory mapped I/O) then writing a "01" Go command to RK05's command register and watch the RK05 disk activity light flicker briefly. Voila, RT11 was loaded. (This worked because on the PDP 11/45 DMA transfers from disk to memory worked also in CPU Halt mode.)

If you want to play the original Adventure without the PDP-11 (though not nearly as authentic), I ported it to the web here: https://grack.com/demos/adventure/

I added a 6Mb hard disk drive to my H11. This required building an interface card, and writing the device driver for it.

I used as a model the device driver for the floppy drive. I was rather amazed at how it worked - it was loading instructions into memory as it was executing the previous instruction just loaded. The instructions were also simultaneously handling two independent values in the upper and lower bytes. A marvel of compactness.

Here is a PDP-11 booting UNix v6: https://www.youtube.com/watch?v=9wFJL9inunQ

The LSI-11 and 11/23 jumped to 173000 at startup, so (re)booting them was just hit halt on the front panel and then enter '173000G' in ODT (the built-in octal debugger). If necessary, it would only take a minute or two to type in a bootstrap somewhere in RAM and execute (my boss at the time could do that faster across the toggle keys on the front panel of his larger PDP-11 than I could type it in).

I coded several bootstraps for PDP-11/23's, managed to get 4 different devices (DX,DY 8" disks, MX 5.25" and MZ 3.5" disks) complete with CLI into 256 words (2x256 byte OTP ROMs). Careful coding time, since one had to send the listing off to someone who could burn the ROMs and post them back.

My favorite PDP-11 / LSI-11 instruction was 014747 octal: mov -(pc), -(pc). It would fill all of memory with copies of itself.

I actually had a use for that, to get a board that had core memory to a known state.

That ASR-33 Teletype comes in handy if your PDP-11 doesn't have a boot ROM: toggle the first stage bootloader into the panel, load your tape into the reader, and let 'er rip.

At a summer job in 1973, we used carved wood sticks to quickly set the toggle switches for the bootloader.

Off-topic: with a bunch of vintage ‘booting a computer’ stuff lately on HN and in my YT recommendations, I keep remembering what I once read about an old trick for ‘faster booting’: on bootup, the computer displayed a static screenshot taken before the shutdown (or rather to-disk suspension, apparently)—and while the user themselves tried to remember what they were doing the day before, the machine actually loaded programs into the memory. Only after that the cursor started blinking. Since after turning on the power the user received fast(er) feedback which kept them occupied, they perceived the bootup as much quicker than it actually was.

Only, I don't remember what computer that was—probably paid more attention to the UX-design side at the time. If anyone knows, I'd be grateful for the reference, so I could finally file this properly among my pile of tech trivia.

I am old enough to remember those days; got a summer job when I was in college at a lab that had lots of PDP-11s in, I think, 1980. Some of them didn't have boot ROMs and you had to toggle in the bootstrap program from the front panel every time (only about a dozen instructions).

I remember booting RT-11 on a DEC PDP/11 clone (LSI-11). You would turn it on (along with the Lear Siegler ADM-3A [or equivalent] terminal), and since it did not come with a hard-disk by default, you would insert the (8") boot floppy, and type 177170L at the (ROM) monitor prompt. The octal address was followed by a "L" which immediately caused the monitor to jump to the boot loader (also in ROM). After about five minutes, RT-11 would be up and running (just as shown on the terminal in the fine article).

I worked at The Computer Museum in Boston (RIP) in the '90s. We had an old computer whose boot program was a 2D array of toggle switches. In other words, it was a small physical-write, electrical-read memory. I'm guessing it was about 16 switches wide by 30 switches high.

Someone needs to write a Y2K patch...

This is so cool. I want to see posts like this for other old school minicomputers and mainframes!

Memories! One of the things I had to do in my first ever job was swap the backup discs in a Vax PDP-11 that ran a big pharmaceutical inventory system. All that time waiting for those huge discs to dismount and re-mount...

If you were lucky enough to have real core then you may not have had to toggle in bootstrap after powering on ...

You can play ADVENT under the BSD-games package.

Also, there's a good ZMachine port (several of them). Look out for advent.z5.