220 33602 article Path: news.gmane.org!.POSTED!not-for-mail From: Phil Bouchard Newsgroups: gmane.comp.lang.c++.isocpp.proposals Subject: Re: Deterministic C++ memory manager Date: Fri, 4 Aug 2017 21:15:26 -0700 (PDT) Lines: 251 Approved: news@gmane.org Message-ID: References: <68af3311-d2d7-42ff-8cfd-f8bcdc20bf0c@isocpp.org> Reply-To: std-proposals@isocpp.org NNTP-Posting-Host: blaine.gmane.org Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="----=_Part_1410_995564218.1501906526670" X-Trace: blaine.gmane.org 1501906532 15628 195.159.176.226 (5 Aug 2017 04:15:32 GMT) X-Complaints-To: usenet@blaine.gmane.org NNTP-Posting-Date: Sat, 5 Aug 2017 04:15:32 +0000 (UTC) To: ISO C++ Standard - Future Proposals Original-X-From: std-proposals+bncBCAYTFXFZMBBBX4MSXGAKGQE2ZTWFMY@isocpp.org Sat Aug 05 06:15:24 2017 Return-path: Envelope-to: gclcip-std-proposals@m.gmane.org Original-Received: from mail-it0-f70.google.com ([209.85.214.70]) by blaine.gmane.org with esmtp (Exim 4.84_2) (envelope-from ) id 1ddqUl-0003Qc-3x for gclcip-std-proposals@m.gmane.org; Sat, 05 Aug 2017 06:15:23 +0200 Original-Received: by mail-it0-f70.google.com with SMTP id r9sf29690389ita.7 for ; Fri, 04 Aug 2017 21:15:29 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=isocpp-org.20150623.gappssmtp.com; s=20150623; h=date:from:to:message-id:in-reply-to:references:subject:mime-version :x-original-sender:reply-to:precedence:mailing-list:list-id :list-post:list-help:list-archive:list-subscribe:list-unsubscribe; bh=yUG7SvNgBvdPyhX1mueNi3qxRBZ3ZSCGAP/PJ4lVfJY=; b=AFzgPWOH4Bj1Hb/fV5qXOc8LWrAlth69axXuSRHMh3ocJ/TvQyH1LIYVINxio/xo0n 8F7TxgYOJYnS5VlP8RGlAjjAE0DJy+fsPt51WOumKPa2XnFeIB63KFVBmztS6FOxrAmW GQi5I2n0uX4nxpt+uvHAivWka9gV5VWgS8y1T68YRr2/PRZzMxGYPWcfeaIhSPWw2ARL O2HeSMh+S2puwn0UWdw5iAINN9nH5UntLjwJosvsUYn2mcKsM3f5S2lNQU6Hji8WzewB /acsCbfjgqfqyTko8+m1O4xZ9R3CL9LhqBIuauzmETaWlGIx0pPQecmeDc0djDtfITvQ 2RIA== DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=date:from:to:message-id:in-reply-to:references:subject:mime-version :x-original-sender:reply-to:precedence:mailing-list:list-id :list-post:list-help:list-archive:list-subscribe:list-unsubscribe; bh=yUG7SvNgBvdPyhX1mueNi3qxRBZ3ZSCGAP/PJ4lVfJY=; b=vePjCrV67cdy/3WedbzlrsPtUqgHGyTuHBkw5TbJGnL3BWEG9DIfoAxGDWOQOJdUgQ ZRdfAJx8D8yHSxGzXlqH66l7WFslk1Y09UptlapnGKiWs6YqAMxXPelq7dse+IwSlw8F 7XRCDOYDlyfdKQGWWJlzZxOHZEZviWPR80MzHwQ8lOk0uuTROlxaNGi0j1jpMgoAlIs4 hbhOBq22/dkDNbc3eIvH5+IArXVyxzCxAu/afVoMRlmnvDOEdRLeXFGzGlb8ObgkJO6O RV4XreevYWDWuNPzACSNLcY1uq/5T81HymR5DWmRUN3SwsSU9D2lvnEY3lCy7X8VX0xa ve7w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:date:from:to:message-id:in-reply-to:references :subject:mime-version:x-original-sender:reply-to:precedence :mailing-list:list-id:x-spam-checked-in-group:list-post:list-help :list-archive:list-subscribe:list-unsubscribe; bh=yUG7SvNgBvdPyhX1mueNi3qxRBZ3ZSCGAP/PJ4lVfJY=; b=ZoD9+/oW9Nsxypj7u9drUvJcRpBtB982Xh4yenErUTlDowmD7bpeEcZdNM+5+xmmDR ncs5IWaC9rPtLti2HI+xovdvo7t8Ny9veAECSFAUL0NaILZS1IAJQIyfyp5fpKHkcBo1 wF8r23F7hVIkIwtFFBcSQWpdhfgLZOaePis6LWv5a40g76tmaDyOfa+A4ObLWtXSQdKW iQA+dGQZ4uCN9VFcDKd8udjYX/edMy05fuR0h3yXKljXrjd3H+w8Xm5QG9hJ4mMvSCbv SWyLPZNy7fLDP0A/tEB9GuhTBOBFMS1DZwn0P+kzf8Bi/yLleWHhoLH4pyO9jINSAO+V eACg== X-Gm-Message-State: AIVw1117rWJaC2MRMZ6CwdXrSWhWky0qddBh8XDnMv1kWT0v4PVoTk3a 8X7m6e+R+KxyTh2R X-Received: by 10.107.205.131 with SMTP id d125mr1464484iog.45.1501906528517; Fri, 04 Aug 2017 21:15:28 -0700 (PDT) X-BeenThere: std-proposals@isocpp.org Original-Received: by 10.107.24.69 with SMTP id 66ls4439727ioy.18.gmail; Fri, 04 Aug 2017 21:15:27 -0700 (PDT) X-Received: by 10.31.169.17 with SMTP id s17mr22953vke.11.1501906527083; Fri, 04 Aug 2017 21:15:27 -0700 (PDT) In-Reply-To: X-Original-Sender: philippeb8@gmail.com Precedence: list Mailing-list: list std-proposals@isocpp.org; contact std-proposals+owners@isocpp.org List-ID: X-Google-Group-Id: 399137483710 List-Post: , List-Help: , List-Archive: List-Subscribe: , List-Unsubscribe: , Xref: news.gmane.org gmane.comp.lang.c++.isocpp.proposals:33602 Archived-At: ------=_Part_1410_995564218.1501906526670 Content-Type: multipart/alternative; boundary="----=_Part_1411_1379060365.1501906526671" ------=_Part_1411_1379060365.1501906526671 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Friday, August 4, 2017 at 6:40:45 PM UTC-4, Arthur O'Dwyer wrote: > > On Friday, August 4, 2017 at 1:07:31 PM UTC-7, Phil Bouchard wrote: >> >> Hi ISO C++,=20 >> >> I have developed a deterministic memory manager in C++ and I was=20 >> wondering if ISO C++ would be interested to hear more about it. >> >> For the moment I have included my solution in a brand new language I cal= l=20 >> "BB++" because I need to add functionality that C++ doesn't yet have suc= h=20 >> as: >> - to add instances of an object implicitly as each scope; >> - to add implicit references to these objects in top-level classes >> - to overload 'operator .' >> - to use 'auto' for member variables instead of 'decltype' >> - to add metadata of the classes in order to propagate the proxy=20 >> associated to the pointers within those classes >> >> The homepage of my project is: >> https://github.com/philippeb8/root_ptr/tree/bb++/bbpp2cpp >> >> And I have a 17 minutes presentation available here: >> https://youtu.be/GrNDYWyasxg >> >> Please let me know if this is the type of research you are interested in= ..=20 >> The work is a step away from being production ready. It has to be noted= =20 >> that Root.Ptr is already faster than Shared.Ptr in multithreaded mode. >> > > It would help if you could edit down that YouTube video to cut all the=20 > dead air; it's 17 minutes of which the relevant 5 minutes is scattered=20 > around separated by painful silent sections. Even at 200% playback speed,= =20 > it's a slog. > Certainly for your present purposes, *nothing* prior to 6m22s is=20 > relevant. (Anyone who cares already knows how mark-and-sweep and=20 > refcounting work, and doesn't need a tutorial on the subject.) > Thanks I'll make the corrections this weekend.=20 =20 > You never show code or diagrams for root_ptr in the video, but from the= =20 > usage example and caveats, I infer that it is (almost?) exactly what Herb= =20 > Sutter called a "deferred_heap" at CppCon 2016. > > https://internals.rust-lang.org/t/herb-sutter-deferred-heaps-and-pointers= /4183 > Of course everyone has their own implementation of this idea; and there= =20 > are plenty of hybrids, such as "local arena allocators", or mixing in a b= it=20 > of refcounting to get Objective-C's "autorelease pools", or mixing in a b= it=20 > of RCU to get "RCU domains", or whatever. > Firstly I would like to point out that I came up with the idea long before= =20 Microsoft did (2011 actually). You can easily find references to my=20 previous attempts in the development Boost mailing list. Back in the days= =20 my library was called: "block_ptr" and "shifted_ptr".=20 Secondly deferred_ptr is not production ready and doesn't support=20 multithreading mode whereas root_ptr was correctly unit tested and it does= =20 support multithreading mode but I just need to optimize one mutex (that is= =20 currently a static member) to make it even faster.=20 =20 > Starting around 12m45s, the examples seem to be concerned entirely with= =20 > showing examples of successful escape analysis =E2=80=94 that is, you cre= ate a=20 > "node_proxy" (a.k.a. deferred_heap) and allocate some objects in that hea= p,=20 > and then you return a pointer to one of those objects from the current=20 > function. Normally this would cause dangling-pointer/use-after-free error= s,=20 > but in this case your BB++ language is doing some sort of "escape analysi= s"=20 > to promote the returned object into the surrounding scope. Some explanati= on=20 > of this promotion mechanism would be useful, since it's the one thing=20 > that's not obvious how to do in today's C++. > I will clarify what is currently happening at that moment in my next video.= =20 Thanks for your patience. Regards, -Phil=20 --=20 You received this message because you are subscribed to the Google Groups "= ISO C++ Standard - Future Proposals" group. To unsubscribe from this group and stop receiving emails from it, send an e= mail to std-proposals+unsubscribe@isocpp.org. To post to this group, send email to std-proposals@isocpp.org. To view this discussion on the web visit https://groups.google.com/a/isocpp= ..org/d/msgid/std-proposals/c019348f-8a1f-46b3-a4a5-b5fdc3a23226%40isocpp.or= g. ------=_Part_1411_1379060365.1501906526671 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable


On Friday, August 4, 2017 at 6:40:45 PM UTC-4, Art= hur O'Dwyer wrote:
On Friday, August 4, 2017 at 1:07:31 PM UTC-7, Phil Bouchard wrote= :
Hi ISO C++,

I have developed a deterministic memory manager in C++ and I was=20 wondering if ISO C++ would be interested to hear more about it.

For the moment I have included my solution in a brand new language I call=20 "BB++" because I need to add functionality that C++ doesn't y= et have=20 such as:
- to add instances of an object implicitly as each scope;
- = to add implicit references to these objects in top-level classes
- to ov= erload 'operator .'
- to use 'auto' for member variables= instead of 'decltype'
- to add metadata of the classes in order= to propagate the proxy associated to the pointers within those classes

=

The homepage of = my project is:
https://github.com/= philippeb8/root_ptr/tree/bb++/bbpp2cpp

And I have a 17 minutes presentation ava= ilable here:
https://youtu.be/GrNDYWyasxg

Please let me know if this is the type of research you are interested in. The work is a step away from being production ready. It has to be noted th= at Root.Ptr is already faster than Shared.Ptr in multithreaded mode.


It would help if you could edit down th= at YouTube video to cut all the dead air; it's 17 minutes of which the = relevant 5 minutes is scattered around separated by painful silent sections= .. Even at 200% playback speed, it's a slog.
Certainly for you= r present purposes, nothing prior to 6m22s is relevant. (Anyone who = cares already knows how mark-and-sweep and refcounting work, and doesn'= t need a tutorial on the subject.)

<= div>Thanks I'll make the corrections this weekend.=C2=A0
= =C2=A0
<= div>You never show code or diagrams for root_ptr in the video, but from the= usage example and caveats, I infer that it is (almost?) exactly what Herb = Sutter called a "deferred_heap" at CppCon 2016.
Of course everyone has their own implement= ation of this idea; and there are plenty of hybrids, such as "local ar= ena allocators", or mixing in a bit of refcounting to get Objective-C&= #39;s "autorelease pools", or mixing in a bit of RCU to get "= ;RCU domains", or whatever.

Firstly I would like to point out that I came up with the idea long=20 before Microsoft did (2011 actually). You can easily find references to=20 my previous attempts in the development Boost mailing list. Back in the=20 days my library was called: "block_ptr" and "shifted_ptr&quo= t;.

Secondly deferred_ptr is not production ready and doesn't support= =20 multithreading mode whereas root_ptr was correctly unit tested and it=20 does support multithreading mode but I just need to optimize one mutex=20 (that is currently a static member) to make it even faster.=C2=A0
=

=C2=A0
Starting around 12m45s, the examples seem to be con= cerned entirely with showing examples of successful escape analysis =E2=80= =94 that is, you create a "node_proxy" (a.k.a. deferred_heap) and= allocate some objects in that heap, and then you return a pointer to one o= f those objects from the current function. Normally this would cause dangli= ng-pointer/use-after-free errors, but in this case your BB++ language = is doing some sort of "escape analysis" to promote the returned o= bject into the surrounding scope. Some explanation of this promotion mechan= ism would be useful, since it's the one thing that's not obvious ho= w to do in today's C++.

I will clarify what is currently happening at that moment in my= next video. Thanks for your patience.


<= div>Regards,
-Phil=C2=A0

--
You received this message because you are subscribed to the Google Groups &= quot;ISO C++ Standard - Future Proposals" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to std-proposa= ls+unsubscribe@isocpp.org.
To post to this group, send email to std-proposals@isocpp.org.
To view this discussion on the web visit https://groups.google.= com/a/isocpp.org/d/msgid/std-proposals/c019348f-8a1f-46b3-a4a5-b5fdc3a23226= %40isocpp.org.
------=_Part_1411_1379060365.1501906526671-- ------=_Part_1410_995564218.1501906526670-- .