Re: Alpha AXP is dead
- From: "Peter \"Firefly\" Lund" <firefly@xxxxxxx>
- Date: Fri, 22 Dec 2006 10:30:15 +0100
On Wed, 8 Nov 2006, glen herrmannsfeldt wrote:
I like VAX, too, but it was designed for the past, not the future.
Among others, a 512 byte page is too small, and the instruction length
is too variable, with too many and too complex addressing modes.
I intend to build a VAX clone with TTL chips (74LSxxx) + SRAM chips. It is sort of a long-term goal: I'm not good enough with analog electronics yet so I'm going to practice with smaller things first. I do have the pipeline sketched out, though.
It looks very little like the VAX microarchitectures DEC used -- it's more like the 486/Pentium microarchitectures.
One of my conclusions so far is that the variable instruction length (and the extreme maximum length!) of the VAX is not really such a problem.
(and besides, one that we know how to solve: put instruction start/stop bits in the L1 cache.)
The small page length seems to have been a real problem -- but one that could have been fixed.
As a machine to play with, to write assembly code on, and otherwise use
for a hobby it is fine.
I do like Alpha, though. It does seem well designed, isn't too hard
to figure out how to code in assembly.
Not having byte load/store instructions was a big mistake. It made driver code a lot harder to port. It also meant that individual single-byte variables that happened to be too close together couldn't be accessed atomically -- even though the source code showed that everything was fine.
-Peter
.
- Follow-Ups:
- Re: Alpha AXP is dead
- From: Richard Tomkins
- Re: Alpha AXP is dead
- From: Richard Tomkins
- Re: Alpha AXP is dead
- Prev by Date: Re: Mentec US is gone!
- Next by Date: Re: Alpha AXP is dead
- Previous by thread: Wanted to buy L0001 parts on vax 11/750
- Next by thread: Re: Alpha AXP is dead
- Index(es):
Relevant Pages
|
