This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFA: patch for usage of more accurate live information for theglobal and the reload.

On May 25, 2004, at 4:11 PM, Vladimir Makarov wrote:

Steven Bosscher wrote:

As for data flow equation solver functions from df.c, I found that gcc
bootstrap will be ~ 1% slower with them (besides a lot of changes to the
patch). So the modified patch still contains the original data solver.

Please no, not another ad hoc dataflow solver. We already have too many of those.

IMHO, df data slover is ad hoc solver too with incovinient interface (I beleive my interface could be more general).

So redesign the interface :). Heck, tell me what interface you want, and *i'll* reimplement the interface if you are too busy.
As we both know, the main important part of the df.c solver is the algorithm it uses to solve the dataflow problem, not the equations it allows you to use in the problem.
The algorithm it uses will work for any equations, nobody ever needed more out of the interface, so nobody ever made it do more.

That's not just a cop out however, since when i originally wrote it, I believe I posted some mail to the mailing list asking if i should bother making it even more general (so that it could handle more complex dataflow problems), and nobody seemed to think that was necessary.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]