Thank you Henk, no I hadn't come across that system3/basic folder at bitsavers. But it is exactly what I was looking for! One intent was to try to verify that indeed the IBM 5100 is running the same "version" of BASIC as was on the native System/3 itself, and approximately when that version of BASIC was available (which seems to be 1971). Although we confirmed the 5100 is (when in "BASIC mode") running the S/3 emulation (we found the instruction jump table in the ROM), I was never really sure where that BASIC code had come from (the one implemented using S/3 instructions). [ now it might be interesting to do a hex comparison of the actual S/3 native BASIC -- what that S/3 emulator is running, and what we've extracted out of the IBM 5100; note the 5100 arranges things in 6KB segments, not every segment is fully used but code generally does straddle across two segments -- but the size is approximately 54KB, which any early microcomputer would struggle to handle that ] Cory is doing a talk about BASIC at June VCF - mostly from the microcomputer perspective, but I wasn't sure if he'd be interested in touching on any pre-1975 aspects (aside from its Dartmouth origin). The S/3 isn't quite a "boot to BASIC" system -- I was curious on the process of how it BASIC was started on this system, and page 12-14 of that Guide to BASIC PDF in the link you provided describes exactly that sequence (which I see followed in your example1/2.txt outputs-- any chance that S/3 emulator runs on a PC? Imagine a MAME front end with a Model 6 CRT depicted). Anyway, it leads to a question of what was the first boot-to-BASIC system? And tentatively I'm thinking that might have been the Wang 2200 (but I've no idea what the "lineage" of the Wang 2200 BASIC was). Yes, yes, there was the HP9830 (if the BASIC ROM cartridge was inserted) - sales of both systems seemed to have started really in '74. Thanks! Steve On Wed, May 7, 2025 at 12:43 AM Henk Stegeman via cctalk < cctalk(a)classiccmp.org&gt; wrote:

Makes sense, IPL (instruction program load?) - how were the disk packs initialized to indicate it was an IPL media? I assume the content to be loaded had to be on a specific sector or of a file-type indicating it is to be loaded on startup? And it is "boot to BASIC" (in quotes) mainly because the operator has to do the effort of at least inserting the proper disk (but it was a "pack", not just a single disk?) I assume SCP is a serial line of some fashion, to support that idea of remote-loading of an OS or startup software? Kind-of sort of an early form of PXE to "boot from a network"? While we're here, another note: on the SCAMP, they emulated the IBM 1130's version of APL (so I may have misspoken earlier - it wasn't an IBM S/3 they had trouble sourcing, but rather during the SCAMP dev it was an IBM 1130 they had trouble finding to borrow/lease). This SCAMP was the prototype leading to the IBM 5100 (where in the IBM 5100, they switched to using an S/360 based APL). On an off-chance of any IBMers out there, I'm still looking for Kitty Price or Patrick Smith (two known experts of the PALM processor, they wrote a paper referring to it in 1974). In the appendix of that paper, they refer to a "1130 PALM Simulator" (i.e. before the PALM was available to them, someone had developed a simulator of it on the 1130). Then in the next section, they refer to a "1130 Simulator which runs under VM/370". Finding any of that software would be incredible. <SCAMP/IBM_SCAMP_1of3_May1974_AnArchitecturalAndDesignOverviewOfScamp_PatrickSmith_KittyPrice_Part1.pdf at main · voidstar78/SCAMP <https://github.com/voidstar78/SCAMP/blob/main/IBM_SCAMP_1of3_May1974_AnArchitecturalAndDesignOverviewOfScamp_PatrickSmith_KittyPrice_Part1.pdf On Thu, May 8, 2025 at 2:38 PM Henk Stegeman via cctalk < cctalk(a)classiccmp.org&gt; wrote: > A few corrections: > The BASIC on the S/3 model 6 was a "boot to BASIC" system. You could IPL > BASIC from R1 > or business SCP from F1. > This BASIC system had virtual memory implemented. Real storage was 8 or > 16k but in the > BASIC environment approximately 54KB was available during execution. > Only thing IBM did was reuse of the Psuedo machine code definition in > the 5100. > Another nice feature of this BASIC was the online helpfiles. > Regards Henk

"Initial Program Load" I don't know any system other than the 360 and that one only very vaguely. I think it would load one record (a card, if IPL from the card reader; a sector, probably the first one, if disk; don't know about tape) and execute that. And I think that the loaded data is a channel program, so it's handled to the IPL channel to execute. A special case is "Emulator IPL" on the 360 model 44, which loads the special memory (inaccessible from normal execution) that contains the code to emulate SS class instrutions. The emulator is provided as a card deck, in standard assembler object file output format, one object data record per card. And the loader is a self-modifying channel program that reads in the entire emulator deck by an impressive mix of data and command chaining, picking up the data load address and length fields of each object card to modify the CCW that then reads the data bytes into the correct memory location. paul

Interesting - then I wonder why in 1974 they bothered to build a S/3 emulator into the IBM 5100. They had already done the S/360 emulator and gotten APL working, so why not use this OS/360 version of BASIC? (well, OS/360 is the OS -- maybe there were more hooks and baggage in getting that working; i.e. they didn't want to pull over the entire OS/360 environment) Or maybe there were more features they liked about the S/3 version? Did the TSO BASIC have floating point, MAT, PRINT USING, etc? On Wed, May 7, 2025 at 12:43 AM Henk Stegeman via cctalk < cctalk(a)classiccmp.org&gt; wrote:

of what is encompassed by the emulation. It’s at least plausible that the > state of the emulation required for implementing the APL interpreter was missing some non-trivial set of features required by the BASIC interpreter. Certainly possible - we've found the jump tables in both emulators that gives an inventory of what opcodes they supported, and to fit in the capacity and cost (of what amount of software could be fitted into a slotted card at the time) that they were targeting, some compromises had to be made [ so neither were a "general 100% cycle accurate" kind of emulation, rather the focus was on supporting the essentials necessary for those respective applications of APL and BASIC, adapted from line-printer over to kbd/CRT fashion ] I recall in the SCAMP Joe George engineering docs, there was an entry where they had trouble funding or obtaining a "spare" System/3 for their development (that being IBMs own internal developers). I'm on travel and will have to lookup the exact reference later. Perhaps similarly, whatever S/360 they borrowed or used previously was no longer available. So (maybe) just being pragmatic on what was available to them at the time, they went the S/3 route. Another reason might possibly be: iirc, 1975 the trial portion of IBMs long antitrust suite was starting (the decade long one ending right at the release of the IBM PC in 1981). That being the issue of coupling software to their hardware. A thought was that by at least showing the concept that "some other language" could essentially be plugged into the system, it at least gives the impression that a third-party could technically do the same. Of course no one ever did (I'm not sure if they'd have access to enough info then to do so), but there was mention of a FORTRAN card concept for the system. Aspects along these lines are also implied in the dev journal mentioned above (if not for third party, but even in internal debate and justification by Friedl to get the project funded, a la "you want to build WHAT?" kind of discussions to a mainframe oriented business unit). Dennis Robinson might still be available to chime in on this aspect. -Steve On Wed, May 7, 2025 at 5:18 AM r.stricklin via cctalk &lt;cctalk(a)classiccmp.org&gt; wrote: > > > > On May 6, 2025, at 11:32 PM, Steve Lewis via cctalk < > cctalk(a)classiccmp.org&gt; wrote: > > > >>>> OS/360 TSO BASIC, pgm-nr 5734-RC2 round 1971 > > > > Interesting - then I wonder why in 1974 they bothered to build a S/3 > > emulator into the IBM 5100. They had already done the S/360 emulator > and > > gotten APL working, so why not use this OS/360 version of BASIC? > > of what is encompassed by the emulation. It’s at least plausible that the > state of the emulation required for implementing the APL interpreter was > missing some non-trivial set of features required by the BASIC interpreter. > > It’s also worth considering, as you point out, that perhaps the BASIC > inerpreter depended on some OS services. APL\360 had originally been its > own standalone timesharing environment. I’m unsure of whether that > situation was fully at an end by the time of APL\360 on DOS and MVT > (1969ish) but I don’t doubt the standalone code was likely still reasonably > fresh by the time work was underway on the 5100. The S/3 BASIC may have > been in a more reasonable state for adapting to standalone operation on the > 5100. I haven’t kept up with all the scholarship around the history of the > 5100 but it’s also not exactly implausible that there was some phase of the > project where the thing had been two separate products. > > ok > bear. > >

APL on the 5100/5110/5120 was based on the IBM mainframe APL/SV, (follow-on to APL\360), with several added features. The "360 emulator" was not a full emulator of 360/370, but only emulated exactly those instructions needed by the APL interpreter (that was written in 360/370 assembler language). Similarly, the System/3 emulator that was developed for the 5100/5110/5120 BASIC implemented only those instructions that were needed by the S/3 BASIC software. System/3 was a machine with limited amounts of real CORE memory, so developers recognized early on that they needed to implement a kind of a "virtual machine" for BASIC, and that was coded in S/3 assembler language. This layer also provided the "virtual" memory for BASIC use, done via "software paging" on a machine (S/3) that had no hardware for virtual memory, relocation, etc. See also: https://www.glennsmuseum.com/items/s3m6/ Glenn Henry was instrumental in the development of the S/3 Model 6 BASIC, and later led the effort to create the IBM System/38. He was eventually promoted to become an IBM Fellow. The design of the virtual instruction set used for S/3 BASIC was used to create the "scientific instructions" for the S/32, S/34 and S/36 that was used by BASIC and FORTRAN on those systems. Mark S Waterbury