OS/2 for PowerPC Tidbits

In December 1994, IBM shipped the first beta version of OS/2 for the PowerPC to selected developers. This beta included the PowerPC operating system as well as Intel-based cross-development tools that ran on OS/2 2.11 or Warp.

The operating system naturally required a PowerPC system to run on. In late 1994, there was only a single machine that OS/2 for PowerPC supported: IBM Personal Computer Power Series 440, also known as Model 6015 or Sandalfoot. This system was very similar to the RS/6000 Model 7020 (40P). The difference was that the Power Series used PReP firmware, rather than OpenFirmware.

Except for the CPU, the Power Series 440 was not substantially different from Intel-based PCs of the era. It was built around PCI/ISA bus, and unlike the later Power Series 830/850 used SCSI storage. The processor was a 66 MHz PowerPC 601 and the system was capable of running in either little-endian (OS/2, Windows NT) or big-endian mode (AIX, Solaris, Linux).

OS/2 for PowerPC was a strange beast, an OS/2 "personality" sitting on top of the Mach kernel. The OS/2 layer was never quite finished and lacked networking support, even though basic networking (such as TFTP) was provided in the Mach kernel. It is quite possible that this major omission was caused by the fact that the (Intel) OS/2 networking stack—LAN Requester/LAN Server—was built on 16-bit LAN Manager code, with a number of special hooks in the OS/2 kernel, and was thus not easily portable.

The cross-development tools were (perhaps surprisingly) built around MetaWare’s High C compiler. In late 1995, IBM also released (as part of the Developer Connection program) its own VisualAge C++ cross-compiler targeting OS/2 for PowerPC.

By the time OS/2 for PowerPC was officially released (December 1995), the project had been essentially canceled. While the basic OS was functioning, it was incomplete. The OS/2 for PowerPC Overview booklet shipped with the OS cautioned that "information that accompanies this product, such as online books, messages, text on windows and buttons, might have references to networking products (Connect). Please note, there is no networking function available with OS/2 Warp (PowerPC Edition)."

The hardware compatibility list in the final OS/2 for PowerPC version was hardly longer than the one in the December 1994 beta. Only two models were supported, IBM Personal Computer Power Series 830 and 850. These were updated versions of the earlier Power Series 440, with IDE disks rather than SCSI but also a PowerPC 604 processor running at 100 or 133 MHz.

It should be noted that as big a failure as OS/2 for PowerPC was, it did not do much worse than the entire Power Series platform and most operating systems available for it. Windows NT supported PowerPC in versions 3.51 and 4.0, but PowerPC support was canceled before Service Pack 4. Solaris only supported PowerPC in version 2.5.1.

In retrospect, it is apparent that the whole RISC vs. CISC "war" was bogus (not unlike the Type 1 vs. Type 2 hypervisor tussle more recently). What matters to customers is performance, features, and price; not a label. Creating a whole PowerPC ecosystem to compete with Intel would have required the PowerPC machines to be either significantly cheaper or significantly faster than they actually were.

IBM clearly believed its own hype and perhaps that was the reason why OS/2 for PowerPC was not developed to support both PowerPC and Intel platforms. As a consequence, most (though not all) of the OS/2 for PowerPC development effort was wasted and could not be reused in the Intel versions of OS/2. That was a pity, as OS/2 for PowerPC was in many respect much more modern than the Intel version, with no 16-bit legacy and many redesigned internals.

This entry was posted in IBM, OS/2, PowerPC. Bookmark the permalink.

8 Responses to OS/2 for PowerPC Tidbits

  1. Yuhong Bao says:

    And the worst thing is the entire effort would not be needed at all if MS stuck with the original OS/2 2.0 and OS/2 NT plan from 1989.

  2. Richard Wells says:

    MetaWare had compilers that worked on both OS/2 (Intel) and PowerPC that were very reliable before OS/2 PPC started. Not the fastest but good enough for most developers. It was fairly obvious that MetaWare was the right choice for early adopters, especially for those working on device drivers.

    Later versions of VisualAge were took a lot longer to compile than competing products, produced code that ran slowly, and was less than stable. I doubt an initial prerelease version for OS/2 PPC would be better than the 1995/1996 versions I worked with. VisualAge was targeted at the internal office developer where sluggish code matters less than the easy ability to combine objects.

  3. It is such a shame that IBM didn’t release any of the L4 port of OS/2 as clearly it must be free of MS’s ASM job on OS/2 1 & 2 …

    But then again at the same time, native software was such a small thing I dont think there’d be all that much call for a free OS/2 ….

    In the end all you’d wind up is with NT anyways.

    Back in the day I remember all the hype built around this mythical PPC 615 (I think it was the 615) that was going to be plug compatable with some Intel processor that could even boot up and behave like an x86, but let you jump into PowerPC mode. Although having used NT on RS/6000’s it was underwhelming in terms of price/performance and the only thing it really had going for it was IIS that was imune to Win32/x86 worms/trojans/viruses as it couldn’t run them (score one for incompatible CPU type!)..

    But PentiumPRO 200Mhz’s were just too cheap compared to any fastish PPC’s .. And while the Dec Alpha was a generation or two ahead, Intel just kept on closing the gap..

    Sigh but 1996-1998 held so much promise of a non x86 future.

  4. michaln says:

    Are you talking about the PowerPC version of VisualAge that targeted OS/2 or AIX? I’m sure the former was never really finished, just like the OS/2 for PowerPC itself.

    The MetaWare compiler certainly did the job, although it made porting code from Intel OS/2 a bit harder–simply because both VisualAge and High C were quirky compilers and the quirks were different.

  5. michaln says:

    It never ceases to amaze me that Intel thought the Itanium was any different from all those RISC processors when it came to displacing x86.

    And I do wonder if IBM still has the source code to OS/2 for PowerPC, and whether it is owned 100% by IBM or not. They certainly can’t have reused any of the 16-bit code directly, although that by itself doesn’t mean much.

  6. jon banquer says:

    mn : Did Intel *really* think the Itanic would displace x86 ? Or did they figure they’d win either way – along with killing off two viable competitors ? Bye-bye PA-RISC. Bye-bye MIPS.

    Of course, that was aided by corporate stupidity, but that’s a constant in this world 🙂

  7. michaln says:

    Interesting question. My hunch is that some people within Intel must have realized or at least strongly suspected that x86 would win (again). But on the corporate level, everyone was clearly aboard the Itanic… with the resulting Pentium 4 mess, and being caught off-guard by AMD in terms of x86 performance and especially the AMD64 architecture. And then having to resurrect the good old P6 microarchitecture, and hastily cobbling nearly unusable AMD64-compatible CPUs. That to me looks like flailing, not a well thought out strategy. But corporations are complex beasts and I don’t know what exactly was going on within Intel.

  8. Alexander says:

    Just sourced the MetaWare High C/C++ compiler for the PowerPC…

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.