Re: [PATCH][Modulo-sched] Change SMS behavior upon failure

[Revital1 Eres <ERES at il dot ibm dot com>]

> There's one thing to check (and some comments below to add to better
explain
> what's going on): now that we do not unschedule nodes, the inner i loop
either
> completes succefully after having scheduled all nodes successfully
(perhaps
> having bumped the ii inserting empty rows), or raises the
> try_again_with_larger_ii (better called flush_and_start_over) flag. So
the
> outer loop can now iterate only as long as try_again_with_larger_ii is
true,
> without the additional condition if all tobe_scheduled insns have been
> scheduled. Please check if this can be cleaned-up.

It seems to be OK.  The attached patch (which includes your comments)
was tested on ppc (r127558) with no regressions and make-check with and
without the patch without new regressions for the C testsuite on spu.

OK for mainline once x86_64 and IA-64 tests will be finished successfully?

Thanks,
Revital

(See attached file: sms_1_9.txt)

Attachment: sms_1_9.txt
Description: Text document