From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 48) id 0C9AC3858C52; Wed, 11 Oct 2023 12:58:33 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 0C9AC3858C52 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1697029113; bh=vKcR2SlzYnc643BE25lSwZjZcrVCs63Nw0MW0nLxc9k=; h=From:To:Subject:Date:In-Reply-To:References:From; b=mngb+4kM24WMWb6gf7T+LjvLDI9gl51B2rLd8bdffZMwXN2v6o6HQX6aBv4f+/YbH BCMzRVnb9Aq5vAQpXBj3RdNzkcuGXAUA35nIsJTpQYRuORavKv+YWnguwfLk14W661 iLhTZLXvWpx3BVdcRwWEVIheTv2rKPY/5O3bAuJs= From: "rguenth at gcc dot gnu.org" To: gcc-bugs@gcc.gnu.org Subject: [Bug tree-optimization/111770] predicated loads inactive lane values not modelled Date: Wed, 11 Oct 2023 12:58:32 +0000 X-Bugzilla-Reason: CC X-Bugzilla-Type: changed X-Bugzilla-Watch-Reason: None X-Bugzilla-Product: gcc X-Bugzilla-Component: tree-optimization X-Bugzilla-Version: 14.0 X-Bugzilla-Keywords: missed-optimization X-Bugzilla-Severity: normal X-Bugzilla-Who: rguenth at gcc dot gnu.org X-Bugzilla-Status: NEW X-Bugzilla-Resolution: X-Bugzilla-Priority: P3 X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org X-Bugzilla-Target-Milestone: --- X-Bugzilla-Flags: X-Bugzilla-Changed-Fields: cf_reconfirmed_on everconfirmed bug_status Message-ID: In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/ Auto-Submitted: auto-generated MIME-Version: 1.0 List-Id: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D111770 Richard Biener changed: What |Removed |Added ---------------------------------------------------------------------------- Last reconfirmed| |2023-10-11 Ever confirmed|0 |1 Status|UNCONFIRMED |NEW --- Comment #1 from Richard Biener --- "Generally" we need an "else" value for .MASK_LOAD and there model "don't care". Elsewhere we observed that most uarchs do zero masked elements (or at least offer so) and thus .MASK_LOAD without else value might be interpreted as doing that (until I came along in the other related PR sugggesting an omitt= ed 'else' value means 'don't care' - but IIRC the RISC-V folks ended up implementing that with default-defs). Btw, we should possibly vectorize this with a COND_DOT_PROD, since adding zeros isn't correct for HONOR_SIGNED_ZEROS/HONOR_SIGN_DEPENDENT_ROUNDING.=