• Re: [OT] Dead DEC Punch card times (was Re: Taking Speedy too seriously)

    From Lawrence =?iso-8859-13?q?D=FFOliveiro?=@ldo@nz.invalid to comp.os.linux.misc,comp.arch,alt.folklore.computers on Tue Nov 3 20:35:31 2026
    From Newsgroup: comp.arch

    On Tue, 3 Nov 2026 07:10:13 -0000 (UTC), Thomas Koenig wrote:

    The specifics of the VAX (especially its addressing modes) made sure
    that it would not scale very well towards higher peformance.

    You have to remember the time (mid-1970s) when it was being designed;
    RISC was just beginning to make its presence felt, and it wouldnrCOt
    have been clear which approach would have been the best for the next
    decade or more -- who could see ahead that far? At that point, the
    world was still full of awkwardly-designed instruction sets that were
    annoying to program for; the PDP-11 was praised as being a major step
    forward in this regard, and the plan for VAX was to make it even more
    so.

    This sort of thing keeps being discussed on comp.arch. The success of
    RISC depended on having lots of CPU registers (and later caching as
    well), and you had to have good compiler technology to manage these
    hardware resources. The economics of doing all this properly took a
    few more years to fall into place.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From anton@anton@mips.complang.tuwien.ac.at (Anton Ertl) to comp.os.linux.misc,comp.arch,alt.folklore.computers on Tue Nov 3 21:11:40 2026
    From Newsgroup: comp.arch

    Lawrence =?iso-8859-13?q?D=FFOliveiro?= <ldo@nz.invalid> writes:
    On Tue, 3 Nov 2026 07:10:13 -0000 (UTC), Thomas Koenig wrote:

    The specifics of the VAX (especially its addressing modes) made sure
    that it would not scale very well towards higher peformance.

    You have to remember the time (mid-1970s) when it was being designed;
    RISC was just beginning to make its presence felt, and it wouldnrCOt
    have been clear which approach would have been the best for the next
    decade or more -- who could see ahead that far?

    John Cocke. A few years later, David Patterson (Berkeley RISC), and
    John Hennessy (Stanford MIPS).

    the PDP-11 was praised as being a major step
    forward in this regard, and the plan for VAX was to make it even more
    so.

    It seems to me that the PDP-11 is easier to run fast than the VAX.

    But performance was not the only or even dominant goal in the VAX
    instruction set, as one can see in the VAX call instruction.

    The success of
    RISC depended on having lots of CPU registers (and later caching as
    well), and you had to have good compiler technology to manage these
    hardware resources.

    No. Pipelining is enough to let a RISC in the same technology
    outperform a VAX 11/780.

    Followups set to comp.arch.

    - anton
    --
    'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
    Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From makendo@makendo@makendo.invalid to comp.arch on Wed Nov 4 14:40:36 2026
    From Newsgroup: comp.arch

    The success of
    RISC depended on having lots of CPU registers (and later caching as
    well), and you had to have good compiler technology to manage these
    hardware resources.

    No. Pipelining is enough to let a RISC in the same technology
    outperform a VAX 11/780.

    Practice from x86 shows that pipelining and caching don't just
    benefit RISC.

    The advantage of RISC for in-order designs is, I guess, you could
    save the area needed by a decoder. Then enters OoO, and even with
    RISC you need a decode stage for OoO designs. Intel tried again by
    introducing a VLIW architecture to workstations and servers, and
    we all know how it ends.

    1970s and 1980s is kinda when every sort of weird instruction set
    architectures fluorished -- you have the textbook CISC called VAX,
    you have early RISC designs using register windows and branch
    delay slots, and you have x87 with its stack-machine-esque model...
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From MitchAlsup@user5857@newsgrouper.org.invalid to comp.arch on Wed Nov 4 18:52:52 2026
    From Newsgroup: comp.arch

    anton@mips.complang.tuwien.ac.at (Anton Ertl) posted:

    /ERROR "unexpected byte sequence starting at index 404: '\xE2'" while decoding/:

    Lawrence =?iso-8859-13?q?D=FFOliveiro?= <ldo@nz.invalid> writes:
    On Tue, 3 Nov 2026 07:10:13 -0000 (UTC), Thomas Koenig wrote:

    The specifics of the VAX (especially its addressing modes) made sure
    that it would not scale very well towards higher peformance.

    You have to remember the time (mid-1970s) when it was being designed;
    RISC was just beginning to make its presence felt, and it wouldn|o-C-Ot >have been clear which approach would have been the best for the next
    decade or more -- who could see ahead that far?

    John Cocke. A few years later, David Patterson (Berkeley RISC), and
    John Hennessy (Stanford MIPS).

    Seymore Cray (and James Thornton) CDC 6600 Concurrency and OoO
    delivered 1963.

    the PDP-11 was praised as being a major step
    forward in this regard, and the plan for VAX was to make it even more
    so.

    It seems to me that the PDP-11 is easier to run fast than the VAX.

    But performance was not the only or even dominant goal in the VAX
    instruction set, as one can see in the VAX call instruction.

    The success of
    RISC depended on having lots of CPU registers (and later caching as
    well), and you had to have good compiler technology to manage these >hardware resources.

    No. Pipelining is enough to let a RISC in the same technology
    outperform a VAX 11/780.

    Followups set to comp.arch.

    - anton
    --- Synchronet 3.22a-Linux NewsLink 1.2