Richard Gorton (gorton@amt.tay1.dec.com)
Mon, 30 Nov 1998 09:09:36 -0500
Naohiko Shimizu wrote:
>But I myself is not comfortable with 21264, because it does not
>improve the peak floating point performance from 21164. The 21364
>will use the same core with 21264 and then we cannot expect
>outperformed performance with these chips.
If by 'peak floating point performance' you mean "number of floating
point instructions issued/cycle" then this is a correct statement.
However, since it is an out-of-order machine, the actual elapsed time
to complete a job can be significantly less on a 21264 machine than on
a 21164 machine with the same clock rate.
Additionally, the number of cycles to execute some FP operations is
significantly better on the 21264 vs. the 21164 (i.e. divide)
The addition of instructions in the architecture to perform SQRT
(among other items) will result in additional performance gains.
Regards,
Rick
Richard Gorton http://www.digital.com/amt
Digital Equipment Corporation All standard disclaimers apply.
Reply-to: gorton@tallis.enet.dec.com
"A committee is a cul-de-sac down which ideas are lured and then quietly
strangled."
--Sir Barnett Cocks
This archive was generated by hypermail 2.0b3 on Mon Nov 30 1998 - 22:10:34 EST