This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH GCC]Fix pr60363 by adding backtraced value of phi arg along jump threading path
- From: Jeff Law <law at redhat dot com>
- To: "bin.cheng" <bin dot cheng at arm dot com>
- Cc: gcc-patches List <gcc-patches at gcc dot gnu dot org>
- Date: Sun, 04 May 2014 07:05:04 -0600
- Subject: Re: [PATCH GCC]Fix pr60363 by adding backtraced value of phi arg along jump threading path
- Authentication-results: sourceware.org; auth=none
- References: <006901cf4292$a78407e0$f68c17a0$ at arm dot com> <534F66DF dot 6060701 at redhat dot com> <CAHFci2_sLQP2asq3EaW3xMf+JEYE=8f4-x+=sgLahTSU9_rzAw at mail dot gmail dot com> <53597994 dot 30908 at redhat dot com> <003701cf6763$8d4bc4d0$a7e34e70$ at arm dot com>
On 05/04/14 00:39, bin.cheng wrote:>
I updated and rebased the patch against latest trunk. It passes bootstrap
and regression test on x86/x86_64. Also pr60363 is fixed on
logical_op_short_circuit targets. Is it OK?
Since ssa-dom-thread-4.c is fixed now, I also reverted the XFAIL test for
Yes. This is fine. Please install.
Most of the time, const propagation opportunities on the original path
are limited to PHIs that become degenerate as a result of isolation of
the threaded path. We use a specialized propagator which seeds strictly
from PHIs to catch those.
An irrelevant point, the const propagation opportunities exist not only for
threaded jump path, but also for the original path after threading jump. So
maybe it's worth to do the general transformation as Jeff suggested before.
If you've got testcases where there's const propagation opportunities on
the original path that do not originate from PHIs that become degenerate
as a result of threading, certainly file a BZ so we can all take a look.