Path: utzoo!utgpu!water!watmath!clyde!rutgers!husc6!hao!gatech!purdue!i.cc.purdue.edu!j.cc.purdue.edu!pur-ee!uiucdcs!uxc.cso.uiuc.edu!ccvaxa!aglew
From: aglew@ccvaxa.UUCP
Newsgroups: comp.arch
Subject: Re: conditional branches
Message-ID: <28200098@ccvaxa>
Date: 14 Feb 88 21:54:00 GMT
References: <191@telesoft.UUCP>
Lines: 33
Nf-ID: #R:telesoft.UUCP:191:ccvaxa:28200098:000:1329
Nf-From: ccvaxa.UUCP!aglew    Feb 14 15:54:00 1988


..> Comparisons, branches, and booleans

The test x = 0, and therefore x = y, can almost certainly be done with
minimum delay in a pipeline, as can x < 0, looking just at the sign bit.
A full arithmetic compare might be a problem because of carry.

When storing the result of a compare into a register, it might be better
to store all the CC bits, rather than just a boolean corresponding to
what you wanted to test. Eg.

	compare result := x ? y
	if result was < branch foo
	if result was > branch bar
	if result was unordered branch foobar
	...

And have a loose encoding so that the "if result was ..." instruction is cheap.

This is similar to what some vector processors do with vector compares: create
a vector of CCs, and be able to do multiple tests with them without rescanning
the vector. 


Andy "Krazy" Glew. Gould CSD-Urbana.    1101 E. University, Urbana, IL 61801   
    aglew@gould.com     	- preferred, if you have nameserver
    aglew@gswd-vms.gould.com    - if you don't
    aglew@gswd-vms.arpa 	- if you use DoD hosttable
    aglew%mycroft@gswd-vms.arpa - domains are supposed to make things easier?
   
My opinions are my own, and are not the opinions of my employer, or any
other organisation. I indicate my company only so that the reader may
account for any possible bias I may have towards our products.