From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 48) id 28CA4382DB21; Thu, 1 Sep 2022 09:03:55 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 28CA4382DB21 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1662023035; bh=NzmZHUqE+O/ueOcdOgIw0rT8Ddlb8QJV3dh35QfWuJY=; h=From:To:Subject:Date:In-Reply-To:References:From; b=kt76V0DH1mUgN+4qUwDMKUjaEQC+bYWZUnbW+U2JvWnLgwBTBNrISpWNq9CTEsWWr vBOpJGm13tWGgV0dn5K00ssxVG2FkAgWJaZ3X2byx8mUhWh/TXNSLW9SDTKgFfDDra YjJiBKhJkPqUX34ofvyTZ1YgtNVmf/gKs7xwYBUI= From: "dodji at redhat dot com" To: libabigail@sourceware.org Subject: [Bug default/29464] abidw performance regression on vmlinux Date: Thu, 01 Sep 2022 09:03:53 +0000 X-Bugzilla-Reason: CC X-Bugzilla-Type: changed X-Bugzilla-Watch-Reason: None X-Bugzilla-Product: libabigail X-Bugzilla-Component: default X-Bugzilla-Version: unspecified X-Bugzilla-Keywords: X-Bugzilla-Severity: normal X-Bugzilla-Who: dodji at redhat dot com X-Bugzilla-Status: ASSIGNED X-Bugzilla-Resolution: X-Bugzilla-Priority: P2 X-Bugzilla-Assigned-To: dodji at redhat dot com X-Bugzilla-Target-Milestone: --- X-Bugzilla-Flags: X-Bugzilla-Changed-Fields: Message-ID: In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://sourceware.org/bugzilla/ Auto-Submitted: auto-generated MIME-Version: 1.0 List-Id: https://sourceware.org/bugzilla/show_bug.cgi?id=3D29464 --- Comment #2 from dodji at redhat dot com --- (In reply to gprocida from comment #1) > The performance regression is reproducible with this test case: >=20 > https://github.com/myxoid/libabigail/raw/ > 51fd7ce8d9b896bf4736bb638266bda5c761ee8a/vmlinux.lzma Thank you Giuliano, this was very useful. I could reproduce the issue, ind= eed. The issue comes from the fixing of the earlier canonical propagation during= the computing of canonical type DIEs, when canonicalizing type DIEs. Before 7ecef6361799326b99129a479b43b138f0b237ae, we were propagating canoni= cal DIEs "too much". Things were fast. But wrong. That is why I tried to fix= the issue. But then, it appears that with that fix, we are not propagating canonical DIEs "enough", so things are super slow. I have a tentative approach to this in the branch users/dodji/libabigail-perf-regr, accessible at https://sourceware.org/git/?p=3Dlibabigail.git;a=3Dshortlog;h=3Drefs/heads/= users/dodji/libabigail-perf-regr. It should bring things back to an acceptable performance, albeit slower to = what it was previously (probably in the 35s range for you). I still need to comment and split the content into proper patches but you c= an already test it to see if things work for you with that branch. In the mean time, I am moving forward with the commenting & splitting. --=20 You are receiving this mail because: You are on the CC list for the bug.=