Re: Strahler number and register allocation

Tomasz Kowaltowski <tk@ic.unicamp.br>
Wed, 14 Jul 2010 10:54:28 -0300

          From comp.compilers

Related articles
Strahler number and register allocation krzikalla@gmx.de (Olaf Krzikalla) (2010-07-13)
Re: Strahler number and register allocation tk@ic.unicamp.br (Tomasz Kowaltowski) (2010-07-14)
Re: Strahler number and register allocation gneuner2@comcast.net (George Neuner) (2010-07-14)
Re: Strahler number and register allocation gneuner2@comcast.net (George Neuner) (2010-07-14)
Re: Strahler numbers tk@ic.unicamp.br (Tomasz Kowaltowski) (2010-07-15)
Re: Strahler number and register allocation blog@rivadpm.com (Alex McDonald) (2010-07-15)
Re: Strahler numbers gneuner2@comcast.net (George Neuner) (2010-07-16)
Re: Strahler numbers cr88192@hotmail.com (BGB / cr88192) (2010-07-21)
[1 later articles]
| List of all articles for this month |

From: Tomasz Kowaltowski <tk@ic.unicamp.br>
Newsgroups: comp.compilers
Date: Wed, 14 Jul 2010 10:54:28 -0300
Organization: Compilers Central
References: 10-07-014
Keywords: registers, optimize
Posted-Date: 14 Jul 2010 12:37:04 EDT

> is it always correct, that the Strahler number of an expression tree
> denotes the minimal number of registers needed? The link in wikipedia
> referring to the original source is apparently broken. However IMHO the
> Strahler number can only be applied if the tree contains binary
> expressions only (which may not be the case anymore with e.g. fused
> multiply-add operations).


Strahler numbers work as long as you have an expression tree. It is a
very simple task to adapt the algorithm for operators with any any
arity: unary, binary, ternary and so on.


-- Tomasz Kowaltowski
[How does it compare to Sethi-Ullman numbering? -John]



Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.