Re: Another stack approach in the Am29000
- To: ganswijk@xxxxxxxxx
- Subject: Re: Another stack approach in the Am29000
- From: girling@xxxxxxxxx (Doug Girling)
- Date: Wed, 20 Mar 1996 12:19:24 -0800 (PST)
- CC: MISC
- In-reply-to: <199603201159.AA03426@xs1.xs4all.nl> (message from Jaap van Ganswijk on Wed, 20 Mar 1996 12:59:13 +0100)
FWIW, stepping into the Wayback Machine (or at least "Computer", May 1977),
one of the HP minis (the HP 3000) had 4 internal registers which formed
the top 4 stack elements. As new things were pushed into R0, R3 was pushed
into real memory, and vice versa. Adding a bigger "stack cache" would
probably help the algorithmic performance, but would need to be balanced
against the overhead of flushing the stack cache on interrupts and context
switches (i.e., where one changes the stack being used).
Just my half mybble's worht,
Doug
------------------------------v-----------------------------------------
Doug Girling | Tel: (604)291-4428 Fax: (604)291-4951
School of Engineering Science | email: girling@cs.sfu.ca
Simon Fraser University | "Not everything that counts can be counted,
Burnaby, B.C. V5A 1S6 CANADA | Not everything that can be counted counts.
