From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <sourceware-bugzilla@sourceware.org>
Received: by sourceware.org (Postfix, from userid 48)
 id 3D3B5386EC52; Fri, 12 Mar 2021 12:35:38 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3D3B5386EC52
From: "vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>
To: dwz@sourceware.org
Subject: [Bug default/27557] Use threading at sub-file level
Date: Fri, 12 Mar 2021 12:35:38 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: dwz
X-Bugzilla-Component: default
X-Bugzilla-Version: unspecified
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: vries at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P2
X-Bugzilla-Assigned-To: nobody at sourceware dot org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-27557-11298-bXg2SiI4KP@http.sourceware.org/bugzilla/>
In-Reply-To: <bug-27557-11298@http.sourceware.org/bugzilla/>
References: <bug-27557-11298@http.sourceware.org/bugzilla/>
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
X-BeenThere: dwz@sourceware.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Dwz mailing list <dwz.sourceware.org>
List-Unsubscribe: <https://sourceware.org/mailman/options/dwz>,
 <mailto:dwz-request@sourceware.org?subject=unsubscribe>
List-Archive: <https://sourceware.org/pipermail/dwz/>
List-Help: <mailto:dwz-request@sourceware.org?subject=help>
List-Subscribe: <https://sourceware.org/mailman/listinfo/dwz>,
 <mailto:dwz-request@sourceware.org?subject=subscribe>
X-List-Received-Date: Fri, 12 Mar 2021 12:35:38 -0000

https://sourceware.org/bugzilla/show_bug.cgi?id=3D27557

--- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
An interesting question is whether find_dups can be parallelized.

The common resource that is used, is the dup_htab. For an actual insertion,=
 the
htab as a whole needs to be locked.

For comparing, comparing DIEs with different checksums can be done in paral=
lel.=20
So, we could fold the ref_checksum to a byte, and use that as in index into=
 an
array of 256 locks.  The problem however is the optimistic marking of DIEs =
as
duplicates when following references.  Marking those optimistically would a=
lso
require acquiring the lock for the corresponding checksum, which opens the =
door
to deadlock problems. We could mitigate for that by detecting cycles upfron=
t,
and avoiding parallel processing of those, but that would again reduce
effectivity.

--=20
You are receiving this mail because:
You are on the CC list for the bug.=