From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
 id A05C5385842A; Tue, 31 Aug 2021 09:39:24 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org A05C5385842A
From: "gcc@christoph-conrads.name" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libstdc++/89979] subtract_with_carry_engine incorrect carry flag
Date: Tue, 31 Aug 2021 09:39:24 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libstdc++
X-Bugzilla-Version: 8.2.0
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: gcc@christoph-conrads.name
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: 
Message-ID: <bug-89979-4-9wdWP8gCDH@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-89979-4@http.gcc.gnu.org/bugzilla/>
References: <bug-89979-4@http.gcc.gnu.org/bugzilla/>
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
X-BeenThere: gcc-bugs@gcc.gnu.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Gcc-bugs mailing list <gcc-bugs.gcc.gnu.org>
List-Unsubscribe: <https://gcc.gnu.org/mailman/options/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=unsubscribe>
List-Archive: <https://gcc.gnu.org/pipermail/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-request@gcc.gnu.org?subject=help>
List-Subscribe: <https://gcc.gnu.org/mailman/listinfo/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=subscribe>
X-List-Received-Date: Tue, 31 Aug 2021 09:39:24 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D89979

--- Comment #4 from Christoph Conrads <gcc@christoph-conrads.name> ---
> LLVM's libc++ does not go into the 0 loop but still does not do a good jo=
b:

The subtract-with-carry PRNG is a simple PRNG, it has a very long period wh=
ose
length can be proved with elementary number theory but the randomness
properties of an SWC on its own are bad which is why you have to discard a =
lot
of values. For reference, std::ranlux48 uses about 11 out of 389 values and=
 you
can detect (self-advertisement: it discards much less variates if you you u=
se
my parameters: b=3D32, s=3D2, r=3D13, see
https://github.com/boostorg/random/issues/57).

Here are the random values are drawing 100 times:

114294 141465 4294632814 252684 106646 4294497449 475286

Here are the random values after drawing 1000 times:

3844519667 4029384828 773557245 3005408692 1891078342 2314767096 2602358454

After drawing 2000 times:

671901017 3735908022 2276383719 2942324048 2197640583 3544339665 2823890385=