Xref: utzoo comp.arch:5472 comp.sys.ibm.pc:17165
Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!mailrus!uflorida!gatech!hubcap!rjschw
From: rjschw@hubcap.UUCP (Robert Schwartz)
Newsgroups: comp.arch,comp.sys.ibm.pc
Subject: 80x86 Measurement
Message-ID: <2167@hubcap.UUCP>
Date: 13 Jul 88 17:48:18 GMT
Organization: Clemson University, Clemson, SC
Lines: 27


As part of ongoing research, I want to make some measurements on the
80x86 architecture.  However, we have some questions some of you out
there on the network may be able to answer.  

What measurements would be useful?  Study of some previous research on 
the VAX-11/780 shows various categories.  For example, Clark and Levy
(IEEE 1982) measure instruction frequency for a benchmark program, a
compiler, and a timesharing workload. Wiecek (ACM 1982) looks at the
execution of six different compilers.

My plans currently are to look at the following:

	* Opcode frequency:  Static, dynamic, for various programs, compilers
		etc.

	* Addressing modes:  Same conditions as the opcode measurements.

Are other measurements desirable (e.g. instruction pair or triples,
opcode successors or predecessors)?  

Please e-mail responses, I can summarize
if there is sufficient interest.  Thanks in advance.
-- 
Robert Schwartz		 	UUCP: ...gatech!hubcap!rjschw
Clemson University		Internet: rjschw@hubcap.clemson.edu
	"What do ya want for nothin', R-r-r-r-r-ubber biscuit?"
					--Elwood Blues