13 Nov
2003
13 Nov
'03
11:05 a.m.
On Nov 13, 0:29, Eric Smith wrote:
Then I get three because I understood it, and can think of processors
that do it -- and a few more because I've not only got some (MIPS
R4600, R5K and R10K) but I know how to tell them what the best guess is
(they have forms of branch instruction that can be tagged in assembler
as "likely" or "not likely", to modify the internal branch prediction
mechanism).
The Pentium can do speculative execution as well, and one version even
did a fetch of both branches, but only a pre-fetch, not execution.
--
Pete Peter Turnbull
Network Manager
University of York
Notice to all:
I'm claiming an extra geek point because this entire
message
makes perfect sense to me. --Patrick
OK, then I claim two extra geek points for having written it. 
13 Nov
13 Nov
6:48 p.m.
Peter Turnbull wrote:
Then I get three because I understood it, and can
think of processors
that do it -- and a few more because I've not only got some (MIPS
R4600, R5K and R10K)
None of which were available in November 1993. And having used the
R4600 quite a bit in 1996, I'm fairly sure that it did *not* have
speculative execution.
The Pentium can do speculative execution as well,
The Pentium was superscalar, but did not do speculative execution.
SE was introduced with the Pentium Pro, which was not introduced until
1995.
I suspect that there were some processors with speculative execution
before November 1993, but not microprocessors.
I'm not convinced that you've earned your three extra geek points
today, sorry.
Eric
7:32 p.m.
Gosh, I just had to enter this fray - especially since I have a complete set
of Microprocessor Reports going back to 1990.
In term of the most complex superscaler microprocessors in 1992, the most
significant were the 960MM, 88110 and the SuperSPARC. Now I quote:
"The 88110 has the most extensive partitioned and duplicated resource
organization of any superscaler microprocessor. It offers a branch/
instruction issue unit, a load/store unit, two integer units, a bit-field
shift unit, two graphics units, and three independent floating point units
(add, multiply and divide). The degree-two issue unit can simultaneously
dispatch two instructions to any two execution units with only minimal
exceptions."
"In addition to its impressive symmetry, the 88110 permits speculative
instruction issue and execution after a predicted branch. A register file
history buffer keeps track of registers so that they can be restored to their
pre-branch values if the branch is later discovered to have been
mis-predicted. Speculative execution allows the 88110 to extract the most
performance from its two issue capability"
Microprocessor Report, October 7, 1992
Lyle
On Thursday 13 November 2003 21:48, Eric Smith wrote:
Peter Turnbull wrote:
--
Lyle Bickley
Bickley Consulting West Inc.
Mountain View, CA 94040
"Black holes are where God is dividing by zero"
Then I get three because I understood it, and can
think of processors
that do it -- and a few more because I've not only got some (MIPS
R4600, R5K and R10K)
None of which were available in November 1993. And having used the
R4600 quite a bit in 1996, I'm fairly sure that it did *not* have
speculative execution.
The Pentium can do speculative execution as well,
The Pentium was superscalar, but did not do speculative execution.
SE was introduced with the Pentium Pro, which was not introduced until
1995.
I suspect that there were some processors with speculative execution
before November 1993, but not microprocessors.
I'm not convinced that you've earned your three extra geek points
today, sorry.
Eric 
7:51 p.m.
Lyle Bickley wrote:
"In addition to its impressive symmetry, the 88110
permits speculative
instruction issue and execution after a predicted branch. A register file
history buffer keeps track of registers so that they can be restored to their
pre-branch values if the branch is later discovered to have been
mis-predicted. Speculative execution allows the 88110 to extract the most
performance from its two issue capability"
Speaking of the "oh so cool" 88k- does anyone have one of these? I
always thought a Moto 88k box running some sort of *BSD would be a great
addition to the collection. Anyone?
-Ryan Brooks
ryan(a)hack.net
8 p.m.
On Friday 14 November 2003 01:51, Ryan K. Brooks wrote:
Lyle Bickley wrote:
Yeah, I've got an NCD 19c X-windows terminal which uses an 88k of some variant
as its processor. I *think* that's the only 88k "machine" that I have...
Pat
--
Purdue University ITAP/RCS
Information Technology at Purdue
Research Computing and Storage
http://www.itap.purdue.edu/rcs/
"In addition to its impressive symmetry, the
88110 permits speculative
instruction issue and execution after a predicted branch. A register file
history buffer keeps track of registers so that they can be restored to
their pre-branch values if the branch is later discovered to have been
mis-predicted. Speculative execution allows the 88110 to extract the most
performance from its two issue capability"
Speaking of the "oh so cool" 88k- does anyone have one of these? I
always thought a Moto 88k box running some sort of *BSD would be a great
addition to the collection. Anyone?
7557
days inactive
7558
days old
4 comments
5 participants
participants (5)
-
eric＠brouhaha.com -
lbickley＠bickleywest.com -
pat＠purdueriots.com -
pete＠dunnington.u-net.com -
ryan＠hack.net