[rc5] Pentium vs. Pentium Pro (renamed)

Brian Hechinger wonko at blackhole.arkham.net
Thu Jul 3 11:15:27 EDT 1997


Sanford Olson drunkenly mumbled...
> 
> I'm not an expert either, but I believe that the Pentium and Pentium Pros
> are quite different internally, too.  The Pentium has dual integer
> instruction pipelines (U & V) plus a FPU.  Something like 70% of the time a
> Pentium executes two instructions at once.  It also has some limited branch
> prediction logic.  The Pentium Pro actually has RISC-type execution units
> (I think 3 of them).  The X86 instructions are translated into RISC-type
> sub-instructions which can be executed out of order, with register renaming
> and better branch prediction.  The problem is that the segment registers
> are not aliased and renamed, so that changing a segment register (which
> happens in 16-bit code constantly) causes practically the entire CPU
> context to be dumped and reloaded.  The Pentium II chips handle segment

i knew there was something special about how a 16-bit instruction could cause
the CPU to flush out everything and start over, i couldn't remember why it was
though.  thanks!! :)

> register changes better and don't suffer as much with poor 16-bit

but how much better are they??  obviously they are much better than a Pentium,
but what if i only want to run 32-bit stuff, how does a Pentium-II stack up
against a PPro of the same clock rate??

> performance.  Of course, 512K of sub-10ns L2 cache on a Pentium Pro doesn't
> hurt performance either. :)  So, a .EXE that was optimized for a Pentium

yea, that's such a sweet deal.  i don't know how the PPro does it exactly,
but i think the data path from the CPU to the onboard L2 runs at either interal
clock rate or some fraction of the clock rate, however it works, it's a billion
times faster than the 66Mhz external bus speed of the chip (although some
motherboards can reportedly do 75Mhz and 100Mhz as external clock rates)

> Pro would help performance (many compilers have a P6 optimizing setting).
> I don't know anything about RC5 algorithms to even have a clue about
> whether it would make a noticeable difference.  Generally, a P6-optimized
> .EXE runs quite well on a regular Pentium, so you wouldn't need separate
> .EXE's.

if the PPro has to break down the x86 instructions, how much optimization can
be done with the exception of using an assembler that talks in "native" PPro
instructions instead of x86 instructions, in which case a P6 optimized binary
won't run on a P5 or lower.  i don't even know if something like that is
possable anyway, i'm just taking a guess at what could be done.

-brian

-- 
** Brian Hechinger ** wonko at mail.arkham.net ** http://www.arkham.net **
--
"Yes, evil comes in many forms, whether it be a man-eating cow or
 Joseph Stalin, but you can't let the package hide the pudding!  Evil
 is just plain bad!  You don't cotton to it.  You gotta smack it in the
 nose with the rolled-up newspaper of goodness!  Bad dog!  Bad dog!"
                                --The Tick

----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.



More information about the rc5 mailing list