Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!husc6!hao!ames!amdahl!amdcad!tim
From: tim@amdcad.UUCP
Newsgroups: comp.arch
Subject: Re: register windows
Message-ID: <18903@amdcad.AMD.COM>
Date: Wed, 28-Oct-87 19:50:04 EST
Article-I.D.: amdcad.18903
Posted: Wed Oct 28 19:50:04 1987
Date-Received: Sat, 31-Oct-87 18:08:03 EST
References: <201@PT.CS.CMU.EDU> <28200058@ccvaxa>
Reply-To: tim@amdcad.UUCP (Tim Olson)
Organization: Advanced Micro Devices
Lines: 28

In article <28200058@ccvaxa> aglew@ccvaxa.UUCP writes:

| How do you count load/stores with non-zero offsets for the AMD29000,
| seeing as it has no such addressing mode? Was this for a design
| alternative, that did have indexing? Or do you count by looking at the
| number of loads that have an add immediately preceding to the register
| used in the address of the load/store? Variations of this last method
| would seem to me to be very susceptible to masking by code rearrangement
| in the optimizer.

Yes, we counted the "offset" loads and stores by looking for an add
(dynamic, at runtime) preceding the load or store which computes into
the load/store address register.

You are correct that this might be suceptible to code rearrangement --
however, our internal compiler doesn't schedule loads, and the only
other code motion is due to loop-invarience [not a problem here] and
filling branch delay slots [the add, load/store will still be sequential
at runtime].

| A while back I posted static counts of address constants on VAX 4.2 BSD.
| As I remember it, about 40% were for locals (stack offsets), but 20% were
| for pointer+field_offset. How do you optimize those away?

Perhaps there is a big difference in static vs dynamic statistics?

	-- tim