Re: Graph colouring and local register allocation

George Neuner <gneuner2@comcast.net>
Sat, 01 Dec 2007 04:41:17 -0500

          From comp.compilers

Related articles
Graph colouring and local register allocation shafitvm@gmail.com (Mohamed Shafi) (2007-10-31)
Re: Graph colouring and local register allocation preston.briggs@gmail.com (preston.briggs@gmail.com) (2007-10-31)
Re: Graph colouring and local register allocation jle@ural.owlnet.rice.edu (2007-11-03)
Re: Graph colouring and local register allocation parthaspanda22@gmail.com (2007-11-04)
Re: Graph colouring and local register allocation SidTouati@inria.fr (Sid Touati) (2007-11-05)
Re: Graph colouring and local register allocation miles.bader@necel.com (Miles Bader) (2007-11-22)
Re: Graph colouring and local register allocation Sid.Touati@uvsq.fr (Sid Touati) (2007-11-30)
Re: Graph colouring and local register allocation gneuner2@comcast.net (George Neuner) (2007-12-01)
Re: Graph colouring and local register allocation sjdutoit@gmail.com (sdt) (2007-12-01)
Re: Graph colouring and local register allocation preston.briggs@gmail.com (preston.briggs@gmail.com) (2007-12-01)
Re: Graph colouring and local register allocation gneuner2@/comcast.net (George Neuner) (2007-12-01)
Re: Graph colouring and local register allocation pertti.kellomaki@tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=) (2007-12-03)
Re: Graph colouring and local register allocation SidTouati@inria.fr (Sid Touati) (2007-12-04)
Re: Graph colouring and local register allocation torbenm@app-1.diku.dk (2007-12-05)
[4 later articles]
| List of all articles for this month |
From: George Neuner <gneuner2@comcast.net>
Newsgroups: comp.compilers
Date: Sat, 01 Dec 2007 04:41:17 -0500
Organization: Compilers Central
References: 07-10-103 07-11-019 07-11-063 07-11-091
Keywords: registers
Posted-Date: 01 Dec 2007 09:46:32 EST

On Fri, 30 Nov 2007 16:17:40 +0100, Sid Touati <Sid.Touati@uvsq.fr>
wrote:


>Miles Bader a icrit :
>
>> So what do "modern" compilers use?
>
>The term "modern" has no sense. It's marketing. Maybe you mean "actual".
>
>Indeed, you should make a distinction what actual compilers do and
>what they should do. Most of the compilers use old register
>allocation techniques, designed for sequential processors. When ILP
>came, few of them used advanced techniques based on strong theoretical
>research results. What happened is that many ad hoc practical
>heuristics have been quickly designed to tackle the problem of
>register allocation with ILP. Some improvement has been measured, but
>on some simple non representative benchmarks. On real life codes, the
>old register allocation techniques are not as efficient as people
>think...


You didn't answer Miles's question.


Frankly, I don't see what impact ILP has on register allocation. The
fact that operations may be performed in parallel has no bearing
whatsoever on the number of live values in a block of code. Either
the processor has enough registers to handle them all concurrently or
it doesn't, and if there aren't enough registers to handle all the
live values, you have to prioritize. Existing instruction scheduling
and register allocation methods are perfectly adequate to do that.


If you know something different then how about providing some cites to
help the rest of us?


George


Post a followup to this message

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