<Allison Parent wrote:
<
<> Competing against the mostly 16bit 8088/6 and the 286 the PDP11 was out
<> front. To match a 16bit cpu against a 32bitter... you must be inhaling
<
<A long running discussion. Allison, I don't understand how you can say
<that the PDP11, with its very simple instruction set, is _more_ CISC
<than (say) the 80286, with which you compare it here. To my mind the
<only really CISC feature of the PDP11 is the MARK instruction. I fear
<we may be talking at cross purposes, and may mean different things by
<RISC and CISC - could you give some specific examples, please?
The CISC features are in the use of general registers (no specific
accumulator) and a huge assortment of addressing modes in many cases
featuring two operand addresses. Most micros are either single address,
primary accumulator and loaded with specific registers. Think of one
micro that would permit PC relative addressing and stack relative. I
happen to know two but they are not common.
Compared to most RISC machines the PDP-11 is very CISC. Also at the
time of the PDP-11 RISC had a specific definition that the -11 clearly
didn't come close to. The definition of the era was all instruction
would execute in 1 to to clocks, register rich and very simple
instructions fast instructions compared to complex many clocks to execute
instructions. The idea of auto index deferred (*x++) is not a RISC concept
as it's far to complex to execute in one or two clocks(minimum of two to
three memory reads and one write).
Opinion: PDP-11 of all 16bit machines of wide spread use was the best
suited for C programming or FORTH due to it's stack archecture and
addressing modes. The only 16 bit machines that can beat it for code
density are a few of the byte instruction oriented machines using far
larger and more complex compilers.
<For those who think a souped up PDP11 could be a pentium killer, bear in
<mind that there was a 32 bit PDP11. I don't mean the VAX, and I don't
<mean the PDP11-68: I mean the Motorola 68000. AFAICT the two
<architectures are very, very similar. Is it a Pentium killer? The
<68070 might have been but it's rather faded away now...
Sorry, the moto is similar in that it borrows some concepts. It lacks
the general resgister archecture, misses the idea of orthoginality and
is a primary accumulator machine. It's at best PDP-11 on bad acid. It
could be a pentium killer as the 680xx was a 32bit machine from day 0.
The VAX is the closest machine to the PDP-11 in general archecture and
instruction set of the lot. I has has compete for years against the X86
machines for a number of years not by surperfast clock numbers but the
ability to manage memory and large numbers of users. The VAX 7000m7xx
series with the NVAX-5(circa 1994) chip was already killing anything
Intel would produce for a few years. People are not junking 7000series
machines over Pentium yet!
A far closer machine is the NS16032, still 32bits but borrows more on
PDP-11 and VAX then moto did.
Even the z8000 was more like PDP-11 than MOTO! It did keep the registers
more or less general, had most of the addressing modes and most all were
applicable to the general registers.
In the intel race there can be lots of competitors as some of the "big
box" systems have far better IO and DISK systems than are married to
most pentium class machines. In the end if you munging gigabyte data
bases raw cpu speed is only a partial solution if your waiting for the
disk!
Allison
Received on Thu Apr 23 1998 - 08:33:40 BST