Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!unido!tub!fauern!tumuc!guug!pcsbst!
From: hgw@rht32.pcs.com (h.-g. willers)
Newsgroups: comp.lang.forth
Subject: Optimization
Message-ID: <1990Feb9.092823.20653@pcsbst.pcs.com>
Date: 9 Feb 90 09:28:23 GMT
Reply-To: hgw@rht32.pcs.com (h.-g. willers)
Organization: PCS Computer Systeme GmbH, D-8000 Muenchen 90
Lines: 25

Please note, that in most modern RISC-processors (such as
Intels i860 or MIPSco's R2000 / R3000) there is no
classical call instruction.

Subroutine linkage is done with a machine register which
holds the return address. So, for routines calling no
further routine, there is no overhead for storing/retrieving
information to/from the machine stack.

Another issue, that has to be taken care of, is the
'branch delay slot' that follows a JAL (jump-and-link
instruction). The branch delay slot is an instruction that
is always executed after the JAL.

So, we have to fill the branch delay slot with something
useful (besides a NOP). That could be the first instruction
of the called routine  (calling effectively routine+4 on a
32-bit machine), if it is not also a branch or jump instruction.

H.-G.

--
H.-G. Willers       PCS-Mail: hgw       internal phone ( -271 )
DOMAIN:  hgw@rht32.pcs.de   (EUR) or  hgw@rht32.pcs.com    (US)
BANG:    ..unido!pcsbst!hgw (EUR) or  ..pyramid!pcsbst!hgw (US)