Re: Graph colouring and local register allocation

On Sat, 1 Dec 2007 10:12:01 -0800 (PST), "preston.briggs@gmail.com"
<preston.briggs@gmail.com> wrote:


>> Frankly, I don't see what impact ILP has on register allocation [...]
>
>If we schedule to minimize register pressure, then allocate, then try
>to reschedule to maxmize ILP, we'll have a hard time. The initial
>schedule and allocation will leave us very little "room to work". On
>the other hand, if we initially schedule for maximum ILP, then
>allocation will require more registers, perhaps causing spill code.


My point was that register allocation for ILP is not, as Sid Touati
seemed to be claiming, a fundamentally different problem from
allocating for a sequential processor. There is some additional
effort to iterate scheduling and allocation to find a good balance,
but existing methods are up to that task already (whether existing
compilers do as well as they could using them is a different
discussion).


New methods that reduce the effort to find the sweet spot are always
welcome, but IMO claiming the old methods can't do the job is just
wrong.




>Google up "Register Allocation with Instruction Scheduling: A New
>Approach" by Schlomit Pinter. That and the papers that cite it will
>get you off to a good start.


I've seen that. Thanks.


George

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