|
From: | Lars Segerlund |
Subject: | Re: [Gomp-discuss] Plan ... coments wanted ! |
Date: | Thu, 30 Jan 2003 10:59:43 +0100 |
User-agent: | Mozilla/5.0 (X11; U; Linux i586; en-US; rv:1.2.1) Gecko/20021226 Debian/1.2.1-9 |
Now I have looked at some existing compilers, and also at the Intel compiler, and the intel guys are doing some things right.
They handle the OpenMP stuff to libcalls after some optimizations ( how do you spell this ? ), thus in the GIMPLE->RTL phase. However some work has to be done before this !
Also they don't generate subroutines which are executed in parallell, they generate a 'reentrant code section' with shared locals available to all threads, and a copy in/out scheme for private ( real local variables , for thae thread ).
I have a suspicion that we might be able to target GENERIC and generate annotated GIMPLE, how does this sound ? ( This is one part which I need to read up on, sorry my understanding of some issues are lacking ).
In other words, build the structure and impose some restriction before the optimization passes and handle the generating of openMP code after this is done, does this sound reasonable ?
Also we might want to keep the option of generating copy in/out subroutines open as this would make future adoption of PHF or parrallell statements running over MPI or PVM possible. It also would make testing much easier.
/ regards, Lars Segerlund. Steven Bosscher wrote:
Op wo 29-01-2003, om 15:50 schreef Diego Novillo:You really want to work on GIMPLE. That's the language over which GCC will do tree optimizations.---- 8< ----As far as the optimizations go, almost everything will be analyzed and optimized at the GIMPLE level.So where should the OpenMP stuff be translated to libcalls and who knows what else? During GENERIC->GIMPLE, or during GIMPLE->RTL? Greetz Steven _______________________________________________ Gomp-discuss mailing list address@hidden http://mail.nongnu.org/mailman/listinfo/gomp-discuss
[Prev in Thread] | Current Thread | [Next in Thread] |