This is the mail archive of the gcc-patches@gcc.gnu.org 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]

Re: [patches] Re: preofile.c versus setjmp


On Tue, Aug 07, 2001 at 07:30:49PM +0200, Jan Hubicka wrote:
> Assume that the abnormal edge targets instruction after CALL.  Then
> emitting any code to the fallthru edge between call and next basic block
> would result in incorrect code (as it will get executed on the abnormal
> path too).

Yep.  We cannot insert instructions on either edge.

> Should be the fallthru edge also abnormal?

I wouldn't have thought so, but I can see how it might solve some
problems.

> Assume that the abnormal edge targets the basic block starting CALL.

This would be wildly incorrect.

> OK, would be sane to write patch to make both edges (fallthru and
> entry-to-the-next-block abnormal? What other flags they should get?)

Yes, but this is starting to sound like something for a branch.


r~


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