From lloyd@acm.jhu.edu Mon Apr 08 11:58:31 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16ueLU-0008Hh-00 for ; Mon, 08 Apr 2002 11:58:28 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id C8F0B13E98 for ; Mon, 8 Apr 2002 14:58:21 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id A584BD843; Mon, 8 Apr 2002 14:58:21 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 8FFFB782E for ; Mon, 8 Apr 2002 14:58:21 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: OpenCL Devel List Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/CIRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] BigInt is being merged. Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.5 Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Mon Apr 8 11:59:27 2002 X-Original-Date: Mon, 8 Apr 2002 14:58:21 -0400 (EDT) I have merged BigInt alpha2 into the main library, meaning the next release of OpenCL (0.8.0) will feature some public key algorithms. I'm working on RSA encryption and signatures. Anyway, the point of this is to ask what people would like to see implemented soon in terms of PK algorithms and support. Particularly, does anyone think support for PKCS #1 v1.5 is needed, or should I just go with the IEEE 1363 functions? DSA? ElGamal encryption? ElGamal signatures? Diffie-Hellman? I need to balance these two factors: 1) Get 0.8.0 out as soon as I possibly can (which will still take a while, given how much work that still needs to be done on BigInt). 2) Ensure it has enough features that people can use it For example, I could implement Montgomery reduction in 0.8.0, have it be released a month later than it would otherwise, or release it sooner, but it will be pretty slow, then do another release later that implements fast modular arithmetic. BTW, the mp_core interfaces have pretty well solidified, so if someone is interested in working on assembly versions of it, let me know. I'll do SPARC and maybe x86 myself, but I'm not even touching it until I finish the important stuff. So, please provide me with guidance, ye users. -Jack From Hany.Greiss@inbusiness.com Thu Apr 11 13:52:27 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16vlYQ-0001cL-00 for ; Thu, 11 Apr 2002 13:52:26 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id <21YDB8C7>; Thu, 11 Apr 2002 16:51:54 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032E3B@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] Windows 2000 Port Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Apr 11 13:56:06 2002 X-Original-Date: Thu, 11 Apr 2002 16:51:53 -0400 Hello, I finally had a few cycles and started porting OpenCL to Windows 2000. I am using Visual C++ V6.0 SP5. This compiler has a number of rather annoying "features". Some of these are well known (listed already on the OpenCL home page). They include the inability to specify initialized static const members. These were fixed easily enough with enums. Throughout the code, I have wrapped my changes as follows: #ifdef OPENCL_WIN new modified code #else existing code #endif I have only 7 errors to go!!! I have hit a roadblock however. Inexplicably, SecureVector seems to kill VC++. I have no idea why. I tried commenting out most of the code and it still fails. Any thoughts, idea, etc. would be appreciated. Hany From Hany.Greiss@inbusiness.com Thu Apr 11 13:55:39 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16vlbW-0002GP-00 for ; Thu, 11 Apr 2002 13:55:38 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id <21YDB8D1>; Thu, 11 Apr 2002 16:55:07 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032E3C@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] Feedback regarding public key algorithms Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Apr 11 14:00:28 2002 X-Original-Date: Thu, 11 Apr 2002 16:55:07 -0400 I think support for DSA, ElGamal etc first would be my preference. I also think that getting the release out and then adding the fine bits later makes sense. Just my two cents worth. Hany From mads@opencs.com.br Fri Apr 12 05:44:46 2002 Received: from [200.182.84.33] (helo=sparta.opencs.com.br) by usw-sf-list1.sourceforge.net with smtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16w0Q1-0002sS-00 for ; Fri, 12 Apr 2002 05:44:45 -0700 Received: (qmail 11763 invoked from network); 12 Apr 2002 12:26:38 -0000 Received: from spatacus-dmz.opencs.com.br (HELO olimpia.opencs.com.br) (200.182.84.62) by sparta.opencs.com.br with SMTP; 12 Apr 2002 12:26:38 -0000 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: RES: [OpenCL-devel] Feedback regarding public key algorithms X-MimeOLE: Produced By Microsoft Exchange V6.0.4417.0 Message-ID: <499FF4AFD7F16F4C99618F59BC5DA5561BCD37@olimpia.opencs.com.br> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [OpenCL-devel] Feedback regarding public key algorithms Thread-Index: AcHhnEUhKSSxzkkfQcWG3LGfmsMOFAAg5d6w From: "Mads Rasmussen" To: Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri Apr 12 05:45:05 2002 X-Original-Date: Fri, 12 Apr 2002 09:45:52 -0300 I might be able to provide test values, if interested Mads Rasmussen Open Communication Security -----Mensagem original----- De: Hany Greiss [mailto:Hany.Greiss@InBusiness.com]=20 Enviada em: quinta-feira, 11 de abril de 2002 17:55 Para: 'opencl-devel@lists.sourceforge.net' Assunto: [OpenCL-devel] Feedback regarding public key algorithms =09 I think support for DSA, ElGamal etc first would be my preference. I also think that getting the release out and then adding the fine bits later makes sense. Just my two cents worth. Hany --=20 Opencl-devel mailing list https://lists.sourceforge.net/lists/listinfo/opencl-devel From lloyd@acm.jhu.edu Fri Apr 12 11:27:07 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16w5lJ-0001og-00 for ; Fri, 12 Apr 2002 11:27:06 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 4851F13EAE; Fri, 12 Apr 2002 14:27:00 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 887A4D849; Fri, 12 Apr 2002 14:26:59 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 6BF6F782E; Fri, 12 Apr 2002 14:26:59 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: Mads Rasmussen Cc: Subject: Re: RES: [OpenCL-devel] Feedback regarding public key algorithms In-Reply-To: <499FF4AFD7F16F4C99618F59BC5DA5561BCD37@olimpia.opencs.com.br> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/CIRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri Apr 12 11:32:35 2002 X-Original-Date: Fri, 12 Apr 2002 14:26:59 -0400 (EDT) That would be amazingly useful. Currently the only one I have is a single RSA OAEP test vector that I found in some RSA DSI publication. I imagine the DSA spec might have one test vector, but I can't even imagine where I could find ElGamal test vectors. Right now I'm working on optimizing the memory allocation system. Last night I profiled the code and found about 90% of the time was being spent in there, which is way too much. It was much worse, at one point about 30 seconds were being spent inside malloc. That was painful. Other than that things are looking pretty good. After I finish (or possibly give up) working on the memory system, all I have planned before the next release is the IEEE 1363 signature padding methods EMSA1 and EMSA2, and DSA and ElGamal (doing both of those should ensure I haven't made any particularly fatal mistakes in my interface designs). Everything will be pretty slow until I do Montgomery reductions. I have implemented Barrett reductions, but the problem is that it can only reduce things that are at most twice the bit length of the modulus minus a small constant. But multiplying two n bit numbers and trying to take the modulus of it with another n bit number often overflows. So it falls back onto the more general divide algorithm way too often. Well, that's basically where things are at right now. -Jack On Fri, 12 Apr 2002, Mads Rasmussen wrote: > > I might be able to provide test values, if interested > > Mads Rasmussen > Open Communication Security > > -----Mensagem original----- > De: Hany Greiss [mailto:Hany.Greiss@InBusiness.com] > Enviada em: quinta-feira, 11 de abril de 2002 17:55 > Para: 'opencl-devel@lists.sourceforge.net' > Assunto: [OpenCL-devel] Feedback regarding public key algorithms > > > > I think support for DSA, ElGamal etc first would be my preference. > I also think that getting the release out and then adding the fine bits > later makes sense. Just my two cents worth. > > Hany > > From mads@opencs.com.br Fri Apr 12 11:43:37 2002 Received: from [200.182.84.33] (helo=sparta.opencs.com.br) by usw-sf-list1.sourceforge.net with smtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16w61I-0005ZO-00 for ; Fri, 12 Apr 2002 11:43:36 -0700 Received: (qmail 12868 invoked from network); 12 Apr 2002 18:25:32 -0000 Received: from spatacus-dmz.opencs.com.br (HELO olimpia.opencs.com.br) (200.182.84.62) by sparta.opencs.com.br with SMTP; 12 Apr 2002 18:25:32 -0000 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: RES: RES: [OpenCL-devel] Feedback regarding public key algorithms X-MimeOLE: Produced By Microsoft Exchange V6.0.4417.0 Message-ID: <499FF4AFD7F16F4C99618F59BC5DA55609771B@olimpia.opencs.com.br> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: RES: [OpenCL-devel] Feedback regarding public key algorithms Thread-Index: AcHiUAFtVwS3jPD2SM2WvC/hPR3q5QAAWDtQ From: "Mads Rasmussen" To: "Jack Lloyd" Cc: Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri Apr 12 11:44:13 2002 X-Original-Date: Fri, 12 Apr 2002 15:45:14 -0300 Test vectors for OAEP and PSS ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1d2-vec.zip I cannot find any vectors for PKCS 1.5, I'm currently fighting with RSA security to make them see the point in making them available. El Gamal might be difficult I am corresponding with the guys at nist to receive test vectors for AES and DSA, you are aware that 512 bit DSA is invalid? There was a correction (Fips 182-2 change notice) that specifies to use only 1024 bit prime p. What about Rabin? You can use OAEP for that one as well, but I cannot find any vectors for that one either. Have you got any vectors for RC4? Regards, Mads -----Mensagem original----- De: Jack Lloyd [mailto:lloyd@acm.jhu.edu]=20 Enviada em: sexta-feira, 12 de abril de 2002 15:27 Para: Mads Rasmussen Cc: opencl-devel@lists.sourceforge.net Assunto: Re: RES: [OpenCL-devel] Feedback regarding public key algorithms That would be amazingly useful. Currently the only one I have is a single RSA OAEP test vector that I found in some RSA DSI publication. I imagine the DSA spec might have one test vector, but I can't even imagine where I could find ElGamal test vectors. Right now I'm working on optimizing the memory allocation system. Last night I profiled the code and found about 90% of the time was being spent in there, which is way too much. It was much worse, at one point about 30 seconds were being spent inside malloc. That was painful. Other than that things are looking pretty good. After I finish (or possibly give up) working on the memory system, all I have planned before the next release is the IEEE 1363 signature padding methods EMSA1 and EMSA2, and DSA and ElGamal (doing both of those should ensure I haven't made any particularly fatal mistakes in my interface designs). Everything will be pretty slow until I do Montgomery reductions. I have implemented Barrett reductions, but the problem is that it can only reduce things that are at most twice the bit length of the modulus minus a small constant. But multiplying two n bit numbers and trying to take the modulus of it with another n bit number often overflows. So it falls back onto the more general divide algorithm way too often. Well, that's basically where things are at right now. -Jack On Fri, 12 Apr 2002, Mads Rasmussen wrote: > > I might be able to provide test values, if interested > > Mads Rasmussen > Open Communication Security > > -----Mensagem original----- > De: Hany Greiss [mailto:Hany.Greiss@InBusiness.com] > Enviada em: quinta-feira, 11 de abril de 2002 17:55 > Para: 'opencl-devel@lists.sourceforge.net' > Assunto: [OpenCL-devel] Feedback regarding public key algorithms > > > > I think support for DSA, ElGamal etc first would be my preference. > I also think that getting the release out and then adding the fine bits > later makes sense. Just my two cents worth. > > Hany > > From lloyd@acm.jhu.edu Sun Apr 14 11:03:38 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16woLh-0000wX-00 for ; Sun, 14 Apr 2002 11:03:37 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 12EEF13EA4 for ; Sun, 14 Apr 2002 14:03:33 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id C3925D849; Sun, 14 Apr 2002 14:03:32 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id B05EF78B0 for ; Sun, 14 Apr 2002 14:03:32 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: OpenCL Devel List Subject: Re: RES: RES: [OpenCL-devel] Feedback regarding public key algorithms In-Reply-To: <499FF4AFD7F16F4C99618F59BC5DA55609771B@olimpia.opencs.com.br> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/CIRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Sun Apr 14 11:04:04 2002 X-Original-Date: Sun, 14 Apr 2002 14:03:32 -0400 (EDT) On Fri, 12 Apr 2002, Mads Rasmussen wrote: > Test vectors for OAEP and PSS > > ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1d2-vec.zip Ah, goody. > I cannot find any vectors for PKCS 1.5, I'm currently fighting with RSA > security to make them see the point in making them available. If you can get ahold of some, please let me know. I'm going to be adding the PKCS stuff in the release after the next one, and having a test vector or three might be nice. > El Gamal might be difficult Yeah. I've decided to hold of on ElGamal for 0.8.0 and just do RSA and DSA. ElGamal signatures have been standardized by the IEEE, but AFAIK nobody has done any real work for standadizing ElGamal encryption, except OpenPGP. It's a mess that I don't want to have to deal with right now. > I am corresponding with the guys at nist to receive test vectors for AES > and DSA, you are aware that 512 bit DSA is invalid? There was a > correction (Fips 182-2 change notice) that specifies to use only 1024 > bit prime p. Oh, I didn't know about that. Thanks. > What about Rabin? You can use OAEP for that one as well, but I cannot > find any vectors for that one either. I'm not worrying about Rabin for now, though that will become an issue in the future, I suppose. > Have you got any vectors for RC4? Yes, plenty. Thanks, Jack From lloyd@acm.jhu.edu Sun Apr 14 11:32:13 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16wonL-00058P-00 for ; Sun, 14 Apr 2002 11:32:12 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 4D71113E9A for ; Sun, 14 Apr 2002 14:32:08 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 23458D849; Sun, 14 Apr 2002 14:32:08 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 0E42978B0 for ; Sun, 14 Apr 2002 14:32:08 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: OpenCL Devel List Subject: Re: [OpenCL-devel] Windows 2000 Port In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032E3B@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/CIRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Sun Apr 14 11:33:04 2002 X-Original-Date: Sun, 14 Apr 2002 14:32:08 -0400 (EDT) On Thu, 11 Apr 2002, Hany Greiss wrote: > Hello, > > I finally had a few cycles and started porting OpenCL to Windows 2000. I > am using Visual C++ V6.0 SP5. This compiler has a number of rather annoying > "features". Some of these are well known (listed already on the OpenCL home > page). They include the inability to specify initialized static const > members. These were fixed easily enough with enums. > > Throughout the code, I have wrapped my changes as follows: > > #ifdef OPENCL_WIN > new modified code > #else > existing code > #endif > > I have only 7 errors to go!!! Great! > I have hit a roadblock however. Inexplicably, SecureVector seems to > kill VC++. I have no idea why. I tried commenting out most of the code and > it still fails. Any thoughts, idea, etc. would be appreciated. I don't know about this one. The only thing I can think of is to put most of the source in a .cpp file and compile it with the rest of the libary, and instantiate versions for byte, u16bit, u32bit, and u64bit inside that file. If that doesn't work, I'm pretty much out of ideas. Maybe try putting all the function definitions inline, or all of them seperate from the declaration? Maybe the mix of some inline, some not is confusing the compiler? You may run into problems using the new SecureBuffer, as it uses the same allocation mechanism as SecureVector. If so, swap it out for the old one that was used in 0.7.8 that uses an array (and if the new SecureBuffer works, I'm curious as to why it does when SecureVector won't). -Jack From lloyd@acm.jhu.edu Sun Apr 14 13:21:27 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 16wqV4-0000Dg-00 for ; Sun, 14 Apr 2002 13:21:26 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 33D6013E9A for ; Sun, 14 Apr 2002 16:21:22 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id E131CD849; Sun, 14 Apr 2002 16:21:21 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id BDB6378B0 for ; Sun, 14 Apr 2002 16:21:21 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: OpenCL Devel List Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/CIRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] Status update Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Sun Apr 14 13:22:02 2002 X-Original-Date: Sun, 14 Apr 2002 16:21:21 -0400 (EDT) Well, it turned out that, while the memory allocation mechanisms are still pretty slow, my real problem was that I misread the spec for Barrett reduction. I thought it could only handle reductions of integers of size 2n - 1 words, where n is the number of words in the modulus. But it can actually handle integers that are up to 2n words. So it was falling back on the more general divide algorithm far more often than it needed to (chewing up a *lot* of memory). That has been fixed and things are much faster. I also redid inverse_mod to use a better algorithm, which doesn't use division. Between these two, RSA is now reasonbly fast, though hardly a speed demon. I have about half a dozen other ideas for good optimizations, though I will probably hold off on most of them until 0.8.1. I'm working on EMSA2 (a padding method for RSA signatures specfied by IEEE 1363) and DSA. The only things I have after that are EMSA1 (really easy), some infrastructure stuff (some of it seems nasty), and documentation for everything. The release WILL be out by the end of the month. Even if I have to ship it with just RSA/EME1 + RSA/Raw encryption and RSA/Raw signatures (though I doubt that will become necessary). OpenCL has been out for over a year and I think it's finally time this stuff was around. Oh, BTW: the special allocators (MemoryLocking_Allocator and MemoryMapping_Allocator) were very broken due to a stupid and horrible bug I made in ManagedAllocator. Basically it's possible that it will return memory that is not initialized to zero, given certain conditions. This breaks many things. It's an easy one line fix that's now in. Also new: I've added a global RNG. It's really pretty convenient, saving you from passing RandomNumberGenerator references every which way. The interface should be thread safe as well. If anyone wants, I can put up a pre-release version in a couple of days so people can see where things are headed. -Jack From lloyd@acm.jhu.edu Tue Apr 30 07:53:11 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 172Z08-00024U-00 for ; Tue, 30 Apr 2002 07:53:08 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 7544813E9B; Tue, 30 Apr 2002 10:52:59 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 3966DD823; Tue, 30 Apr 2002 10:52:59 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 12FA778B2; Tue, 30 Apr 2002 10:52:59 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: Hany Greiss Cc: OpenCL Devel List In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032E72@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] Re: Port of Library to Windows Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Apr 30 07:54:51 2002 X-Original-Date: Tue, 30 Apr 2002 10:52:59 -0400 (EDT) Wow! That is fantastic (big happy grin). Please keep me updated on the progress. -Jack On Tue, 30 Apr 2002, Hany Greiss wrote: > Hi Jack, > > I think I've managed to complete the port of the library to Windows. It > was simply impossible with VC++ v6.0 with any of the available service > packs. Using V7.0 (the .NET version) the library compiled with very few > issues (mainly things like std::ferror, which I simply removed, as per my > previous e-mail to you). > > I will continue testing and let you know if I encounter any problems. I > will also put together a simple makefile that we can use with NMAKE. > > Hany From ParsonD@diebold.com Fri May 10 11:35:15 2002 Received: from [208.228.181.120] (helo=msexch13.diebold.com) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 176FEZ-0001tH-00 for ; Fri, 10 May 2002 11:35:15 -0700 Received: by msexch13.diebold.com with Internet Mail Service (5.5.2653.19) id ; Fri, 10 May 2002 14:35:13 -0400 Message-ID: <2FD9E707FC76D511B68100B0D0F0B7790EAC7F@msexch11.diebold.com> From: "Parsons, Donald" To: "'opencl-devel@lists.sourceforge.net'" Cc: "McCoy, Donald" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] comments on OpenCL 0.8.1 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri May 10 11:36:02 2002 X-Original-Date: Fri, 10 May 2002 14:35:12 -0400 I just discovered OpenCL and tried it on Linux, and it works quite well. Only comment on Linux is that you have to do a make install before make check, otherwise you get: ./check: error while loading shared libraries: libopencl-0.8.1.so: cannot open shared object file: No such file or directory make: *** [check] Error 127 On Windows the configure.pl fails on line 964 as I see now that the mail list more or less warns about. I get: WARNING: Using i386 as default type for family IA-32 Can't link ..\md5.h to include\opencl\md5.h (No such file or directory) at D:\src\OpenCL\OpenCL-0.8.1\configure.pl line 964, line 105. (I don't see where the MODFILE line 105 comes from?) Is there any chance someone (Hany or Jack) could post the upcoming nmake file? I would like to see how OpenCL runs on Windows too. (I was going to compile with gcc 2.95.3-7 [mingw experimental], but I also have VS 7.0, ie, .net) Since recent cygwin is unknown, I will try it too. Sorry fails too: Many warnings like: g++ -Iinclude/ -O3 -fstrict-aliasing -fomit-frame-pointer -march=i686 -ansi -W -Wall -c src/base.cpp -o build/lib/base.obj src/base.cpp: In method `void OpenCL::RandomNumberGenerator:: add_entropy(OpenCL::EntropySource &, bool = true)': src/base.cpp:174: warning: choosing `OpenCL::SecureVector:: operator byte *()' over `OpenCL::SecureVector:: operator const byte *() const' src/base.cpp:174: warning: for conversion from `OpenCL:: SecureVector' to `const unsigned char *' src/base.cpp:174: warning: because conversion sequence for the argument is better and error (64-bit unsupported in cygwin?) g++ -Iinclude/ -O3 -fstrict-aliasing -fomit-frame-pointer -march=i686 -ansi -W -Wall -c src/des_tab.cpp -o build/lib/des_tab.obj src/des_tab.cpp:371: integer constant out of range src/des_tab.cpp:371: integer constant out of range ....omitting several 100 lines.... src/des_tab.cpp:632: integer constant out of range src/des_tab.cpp:632: integer constant out of range make: *** [build/lib/des_tab.obj] Error 1 Since in the end I would like to run just RSA decrypt on an embedded ARM7, does anyone have embedded Crypto Library recommendations. I looked at openSSL and crypto++ but it looks to me that OpenCL would be the easiest to get working. Thanks for any answers. Don From lloyd@acm.jhu.edu Fri May 10 13:27:55 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 176Gza-0002bm-00 for ; Fri, 10 May 2002 13:27:54 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 6A4E913EA1; Fri, 10 May 2002 16:27:50 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id DED03D831; Fri, 10 May 2002 16:27:49 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id C748F78B2; Fri, 10 May 2002 16:27:49 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: "Parsons, Donald" Cc: "'opencl-devel@lists.sourceforge.net'" , "McCoy, Donald" Subject: Re: [OpenCL-devel] comments on OpenCL 0.8.1 In-Reply-To: <2FD9E707FC76D511B68100B0D0F0B7790EAC7F@msexch11.diebold.com> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri May 10 14:00:04 2002 X-Original-Date: Fri, 10 May 2002 16:27:49 -0400 (EDT) On Fri, 10 May 2002, Parsons, Donald wrote: > > I just discovered OpenCL and tried it on Linux, and it works quite well. > Only comment on Linux is that you have to do a make install before > make check, otherwise you get: > > ./check: error while loading shared libraries: libopencl-0.8.1.so: cannot > open shared object file: No such file or directory > make: *** [check] Error 127 Alternately, you can "export LD_LIBRARY_PATH=.", so the dynamic linker will look in the current directory for shared libraries it can't find in /lib, /usr/lib, ... I should probably mention that in the documentation somewhere. > On Windows the configure.pl fails on line 964 as I see now that the mail > list more or less warns about. I get: > > WARNING: Using i386 as default type for family IA-32 > Can't link ..\md5.h to include\opencl\md5.h (No such file or directory) > at > D:\src\OpenCL\OpenCL-0.8.1\configure.pl line 964, line 105. > > (I don't see where the MODFILE line 105 comes from?) MODFILE is a file handle (IIRC) that is used to read information about the add-on modules. Why Perl is reporting it... I don't know. :) Hrm... that's an odd error, too. I'll have to install windows again (my Win2K install went south a few months ago) and play with this. > Is there any chance someone (Hany or Jack) could post the upcoming > nmake file? I don't have it. Hany? > I would like to see how OpenCL runs on Windows too. (I was going to > compile with gcc 2.95.3-7 [mingw experimental], but I also have VS 7.0, > ie, .net) Well, the report is that it works on VC7, modulo a few bugs. I've changed some things to work around one problem Hany mentioned to me, but I haven't seen all of the changes that might be necessary yet. > Since recent cygwin is unknown, I will try it too. Sorry fails too: > Many warnings like: > > g++ -Iinclude/ -O3 -fstrict-aliasing -fomit-frame-pointer -march=i686 -ansi > -W > -Wall -c src/base.cpp -o build/lib/base.obj > src/base.cpp: In method `void OpenCL::RandomNumberGenerator:: > add_entropy(OpenCL::EntropySource &, bool = true)': > src/base.cpp:174: warning: choosing `OpenCL::SecureVector:: > operator byte *()' over `OpenCL::SecureVector:: > operator const byte *() const' > src/base.cpp:174: warning: for conversion from `OpenCL:: > SecureVector' to `const unsigned char *' > src/base.cpp:174: warning: because conversion sequence for the argument is > better Ah, I know what the problem is. GCC 2.95.x is the only compiler that gives those warnings. They're really pretty silly things; all they tell you is that the compiler chose one conversion operator over another for you - which is what it's supposed to do! Sigh. > and error (64-bit unsupported in cygwin?) > g++ -Iinclude/ -O3 -fstrict-aliasing -fomit-frame-pointer -march=i686 -ansi > -W -Wall -c src/des_tab.cpp -o build/lib/des_tab.obj > src/des_tab.cpp:371: integer constant out of range > src/des_tab.cpp:371: integer constant out of range > ....omitting several 100 lines.... > src/des_tab.cpp:632: integer constant out of range > src/des_tab.cpp:632: integer constant out of range > make: *** [build/lib/des_tab.obj] Error 1 What you need to do is edit the Makefile and add -fpermissive to the LIB_OPTS variable. 2.95.x doesn't like having 64-bit constants without a LL suffix. This is bad, because some other compilers use different things. The next release will let you pass the option --gcc295x to configure.pl, and it will do this for you. People have also run into this on FreeBSD and other systems that still use 2.95.x. There will still be a ton of warnings when you compile the library, but it compiles and runs just fine. > Since in the end I would like to run just RSA decrypt on an embedded > ARM7, does anyone have embedded Crypto Library recommendations. I'll abstain from any commentary here, since I'm obviously very biased. But I'll mention that I have compiled OpenCL on ARM (with GCC) and it works. That was a while ago, however. In fact, I think after I take care of a few things I'll try it out again. -Jack From lloyd@acm.jhu.edu Fri May 10 15:27:21 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 176IrA-000399-00 for ; Fri, 10 May 2002 15:27:20 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id BA89013EA1; Fri, 10 May 2002 18:27:16 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 99E05D831; Fri, 10 May 2002 18:27:16 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 8300A78B2; Fri, 10 May 2002 18:27:16 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: "Parsons, Donald" Cc: "'opencl-devel@lists.sourceforge.net'" , "McCoy, Donald" Subject: Re: [OpenCL-devel] comments on OpenCL 0.8.1 In-Reply-To: Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri May 10 15:33:02 2002 X-Original-Date: Fri, 10 May 2002 18:27:16 -0400 (EDT) On Fri, 10 May 2002, Jack Lloyd wrote: > > Since in the end I would like to run just RSA decrypt on an embedded > > ARM7, does anyone have embedded Crypto Library recommendations. > > I'll abstain from any commentary here, since I'm obviously very biased. But > I'll mention that I have compiled OpenCL on ARM (with GCC) and it works. > That was a while ago, however. In fact, I think after I take care of a few > things I'll try it out again. OK, I just ran some tests on ARM. I compiled 0.8.1 using an Alpha -> ARM crosscompiler (gcc 2.95.2) and ran it on an iPaq (running Linux). /proc/cpuinfo: Processor : Intel StrongARM-1110 rev 6 (v4l) BogoMIPS : 194.15 Hardware : Compaq iPAQ Revision : 0200 Everything ran through just fine. It's not particularly fast, which doesn't suprise me much. It takes 37 seconds for the validation set to complete, compared with about 2.5 on my 1.4 Ghz Athlon at home. But then again, it is just a cute little ARM. :) And there have been a lot of optimizer improvements between 2.95.2 and 3.0.4. I'm in the process of compiling it natively on another, slightly faster ARM box, but it hasn't finished yet. I'll post the results when it's done. -Jack From ParsonD@diebold.com Tue May 14 06:46:21 2002 Received: from [208.228.181.120] (helo=msexch13.diebold.com) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 177cdB-0002nJ-00 for ; Tue, 14 May 2002 06:46:21 -0700 Received: by msexch13.diebold.com with Internet Mail Service (5.5.2653.19) id ; Tue, 14 May 2002 09:46:11 -0400 Message-ID: <2FD9E707FC76D511B68100B0D0F0B7790EAC82@msexch11.diebold.com> From: "Parsons, Donald" To: "'opencl-devel@lists.sourceforge.net'" Cc: "McCoy, Donald" Subject: RE: [OpenCL-devel] comments on OpenCL 0.8.1 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue May 14 06:47:06 2002 X-Original-Date: Tue, 14 May 2002 09:46:10 -0400 Don Parsons wrote: > Since recent cygwin is unknown, I will try it too. Sorry fails too: > Many warnings like: (cygwin gcc-2.95.3-5) Well I compiled a new gcc-3.1 rc2 (i.e., May 10, 2002) under cygwin and tried it on OpenCL-0.8.1 And it all worked perfectly. Validation passed, 1.90sec on a dual 1GHz P3 under Cygwin-1.3.10 on Win2k sp2 [ Only "g++ -Wall" output was an unused parameter: g++ -Iinclude/ -O3 -fstrict-aliasing -fomit-frame-pointer -march=i686 -ansi -W -Wall -c src/init.cpp -o build/lib/init.obj src/init.cpp: In member function `OpenCL::LibraryInitializer::LibraryInitializer(const std::string&)': src/init.cpp:17: warning: unused parameter `const std::string&args' ] Thanks, Don PS: I built the gcc-3.1rc2 on Linux and OpenCL passed there too. gcc-3.1 seems very good. The Intel 6 C++ compiler is usually slower and fails on last test (#11) on pari-gp-2.2.2 (and gcc-3.1 gets right). From dbakhash@yahoo.com Sun Jun 09 18:07:15 2002 Received: from web12801.mail.yahoo.com ([216.136.174.36]) by usw-sf-list1.sourceforge.net with smtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17HDeN-0001Ad-00 for ; Sun, 09 Jun 2002 18:07:15 -0700 Message-ID: <20020610010714.1957.qmail@web12801.mail.yahoo.com> Received: from [24.46.16.56] by web12801.mail.yahoo.com via HTTP; Sun, 09 Jun 2002 18:07:14 PDT From: Dave Bakhash To: opencl-devel@lists.sourceforge.net MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Subject: [OpenCL-devel] bad name! Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Sun Jun 9 18:08:02 2002 X-Original-Date: Sun, 9 Jun 2002 18:07:14 -0700 (PDT) Hey, there is an ANSI-standardized language called Common Lisp, which has been abbreviated "CL" for ages. I think that the name you guys have chosen for your software is _not_ a good name, and will cause confusion. Just my $0.02 dave __________________________________________________ Do You Yahoo!? Yahoo! - Official partner of 2002 FIFA World Cup http://fifaworldcup.yahoo.com From lloyd@acm.jhu.edu Mon Jun 10 11:49:28 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17HUEJ-00040F-00 for ; Mon, 10 Jun 2002 11:49:27 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id 4EAF713E97; Mon, 10 Jun 2002 14:49:23 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 4E08937BEA; Mon, 10 Jun 2002 14:49:23 -0400 (EDT) From: Jack Lloyd To: Hany Greiss Cc: OpenCL Devel List In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032EC7@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] Re: Win Port 0.8.3 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Mon Jun 10 11:50:05 2002 X-Original-Date: Mon, 10 Jun 2002 14:49:23 -0400 (EDT) On Mon, 10 Jun 2002, Hany Greiss wrote: (cc-ing the devel list, since this seems on topic for there) > Hi Jack, > > Sorry this took so long. I was off a couple of weeks. In any case as > promised, I've attached a Windows Makefile that works with the latest > Microsoft compiler. Ah! This is _great_. > I had to make a few simple changes. > > a) pipe_io.cpp std:: was removed in a few places. Not much I can do about this one. Does VC come with a (usable) patch tool? If so I'll start generating a patch for this as part of the release script. Or possibly I'll just stop using the std:: in there altogether, which seems ugly, but also very useful, because many systems seem to have problems with that. I'll have to think about this one a bit. > b) init.cpp problems with externs OK... I think that's a real bug; I would presume the solution is to just put extern in front of each of: void load_algorithms_list(); void initalize_memory_subsystem(); void destroy_lookup_tables(); Is this right? If so I'll fix this in 0.8.4. > c) pk_bench timer code What goes wrong here? I don't see anything that could be problematic, at least a first glance. > d) pk.cpp - the function decl should be u32bit otherwise we can't > find the functions during the link > > e) validate.cpp - ditto Those two are plainly just bugs. I'll make sure they get fixed in the next release. I'm suprised it's worked so far... [...] > As of now, the changes are manual. Every time a new release arrives the > Makefile will require changes. I suppose the best way to go would be to > incorporate these changes into configure.pl but this may be ok for the short > term. The Makefile syntax is different enough that I'm sure I'll have to re-write large portions of the configure generation stuff, but it seems worthwhile enough to do. Especially because I have a bad tendency to rename/add/delete files quite a bit. I'll also have to remember to generate the makefile for VC with each release, since we can't really count on people having Perl installed on Windows machines. > I haven't really QA'd the port much, so if problems are encountered let me > know. I've edited the Makefile with this comment on the top: # This makefile, and the OpenCL Win32 port, were contributed by # Hany Greiss . If you encounter problems # with the Win32 port, please contact both Hany and Jack (lloyd@acm.jhu.edu) > One final point. You will need the latest compiler, what Microsoft call > version 7.0 or .NET to build it yourself. I also attached the built library > to save people some time. OK. I've temporarily put the files up as: http://opencl.randombit.net/files/Makefile.Win32 http://opencl.randombit.net/files/libopencl.lib I'll find a more permanent home for them (and set up links from the main pages) in a day or two. Thanks very much for your work, Jack From Hany.Greiss@inbusiness.com Mon Jun 10 11:59:51 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17HUON-0004m0-00 for ; Mon, 10 Jun 2002 11:59:51 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Mon, 10 Jun 2002 15:00:08 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032ECB@OTTEXC00> From: Hany Greiss To: OpenCL Devel List MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] RE: Win Port 0.8.3 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Mon Jun 10 12:00:06 2002 X-Original-Date: Mon, 10 Jun 2002 15:00:08 -0400 Hello, With regards to init.cpp, yes all that was needed were extern statements. And as for pk_bench.cpp, its more of that std:: stuff again. I did have to add an include for the time.h file as well. Hany From lloyd@acm.jhu.edu Mon Jun 10 12:11:38 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17HUZl-000619-00 for ; Mon, 10 Jun 2002 12:11:37 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id 8BFDC13E97; Mon, 10 Jun 2002 15:11:33 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 8B6DF37BEA for ; Mon, 10 Jun 2002 15:11:33 -0400 (EDT) From: Jack Lloyd To: OpenCL Devel List Subject: Re: [OpenCL-devel] RE: Win Port 0.8.3 In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032ECB@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Mon Jun 10 12:12:19 2002 X-Original-Date: Mon, 10 Jun 2002 15:11:33 -0400 (EDT) On Mon, 10 Jun 2002, Hany Greiss wrote: > Hello, > > With regards to init.cpp, yes all that was needed were extern > statements. OK, I'll fix that. > And as for pk_bench.cpp, its more of that std:: stuff again. Odd. What specifically causes problems here? The only things I'm using from std:: in pk_bench.cpp are things that are used without problems elsewhere, basically just: cout, ifstream, string, getline, clock, clock_t, exception. That looks like it. > I did have to add an include for the time.h file as well. Oops. Either me == dumb, or me == lazy. Wait: if the std:: problems were related to clock/clock_t, try including instead of ; that should fix them. -Jack From Hany.Greiss@inbusiness.com Mon Jun 10 12:27:03 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17HUoh-0007eA-00 for ; Mon, 10 Jun 2002 12:27:03 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Mon, 10 Jun 2002 15:27:25 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032ECC@OTTEXC00> From: Hany Greiss To: 'OpenCL Devel List' MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] RE: Win Port 0.8.3 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Mon Jun 10 12:28:03 2002 X-Original-Date: Mon, 10 Jun 2002 15:27:24 -0400 Hi again, Thanks for the tip. Adding an include to pk_bench fixed the std:: problem. So all that's needed is an #include . Hany From Hany.Greiss@inbusiness.com Tue Jun 11 06:56:43 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17Hm8X-0007So-00 for ; Tue, 11 Jun 2002 06:56:41 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Tue, 11 Jun 2002 09:57:01 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032ED1@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] HMAC/EMAC question Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jun 11 06:57:08 2002 X-Original-Date: Tue, 11 Jun 2002 09:56:53 -0400 Hello, Prior to version 8.0.3, the HMAC/EMAC classes were template classes instantiated on the underlying message digest or block cipher. How can we now extract the hash length or the block size of the digest or cipher. For example, if I have OpenCL::HMAC x(new OpenCL::SHA1); is there an easy way to extract the underlying hash length? If there was a method that returned a const pointer to the hash or block cipher, we could then use it to determine the value of the HASH_BLOCK_SIZE or MAXIMUM_KEYLENGTH or whatever. Any thoughts on this? Hany From lloyd@acm.jhu.edu Wed Jun 12 07:16:06 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17I8uq-0003Ui-00 for ; Wed, 12 Jun 2002 07:16:04 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 63E6C13E97 for ; Wed, 12 Jun 2002 10:16:00 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 27CC1D831; Wed, 12 Jun 2002 10:16:00 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id 0F75378B2 for ; Wed, 12 Jun 2002 10:16:00 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: "'opencl-devel@lists.sourceforge.net'" Subject: Re: [OpenCL-devel] HMAC/EMAC question In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032ED1@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jun 12 07:17:06 2002 X-Original-Date: Wed, 12 Jun 2002 10:15:59 -0400 (EDT) On Tue, 11 Jun 2002, Hany Greiss wrote: > > Hello, > > Prior to version 8.0.3 [0.8.3 -ed], the HMAC/EMAC classes were > template classes instantiated on the underlying message digest or block > cipher. How can we now extract the hash length or the block size of the > digest or cipher. For example, if I have > > OpenCL::HMAC x(new OpenCL::SHA1); > > is there an easy way to extract the underlying hash length? Well, in the case of HMAC, the size of the underlying hash output will always be the same length as the output of HMAC, so you could do this: u32bit output_size_of_x = x.OUTPUT_LENGTH; There is a somewhat sillier way do to this, which is to simply mac a "non-message": // create x SecureVector mac = x.final(); // now call mac.size() At this point there is no reasonable way to find out the block size of the hash. Do you really need this for some reason? I didn't bother to do anything for the block size simply because AFAIK the only thing that ever needs to know that is HMAC itself. > If there was a method that returned a const pointer to the hash or block > cipher, we could then use it to determine the value of the HASH_BLOCK_SIZE > or MAXIMUM_KEYLENGTH or whatever. In priciple, I'm not opposed to this. But the only time you would actually need to do this is for the hash's block size: For EMAC and HMAC, here are the equivalences that I know of. The underlying type's characteristics are on the left, the higher level type's cooresponding value (if any) on the right. EMAC: Block size (BLOCK_SIZE) OUTPUT_LENGTH ie: block_size_of_cipher_used_in_emac = emac_obj.OUTPUT_LENGTH Minimum Keylength / Maximum Keylength / Keylength Mod (MINIMUM_KEYLENGTH, MAXIMUM_KEYLENGTH, KEYLENGTH_MULTIPLE) Currently these are wrong in EMAC. It will state that it will accept anything between 1 and 32 bytes inclusive for a key, but it's lying; it will throw an exception if the block cipher doesn't like the key. I'll make sure to fix this so the keylength values reflect what the block cipher can actually handle. HMAC: Hash Size (OUTPUT_LENGTH) OUTPUT_LENGTH Block Size (HASH_BLOCK_SIZE) No sane way. What you can do is call name(), parse the output to get just the hash name, then call get_hash(hash_name) and look at what you get back. But that's really pretty nasty. For keylengths HMAC can accept arbitrary values, which I, just as arbitrarily, have bounded between 8 and 256 bits (inclusive). Let me know if knowing the hash's block size with HMAC really is problem for you. Figuring out what keys you can safely pass to EMAC are also an issue right now. Sorry it took so long to respond, I took yesterday completely off. -Jack From Hany.Greiss@inbusiness.com Thu Jun 13 05:01:05 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17ITHl-0007jl-00 for ; Thu, 13 Jun 2002 05:01:05 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Thu, 13 Jun 2002 08:01:24 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032EE2@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" Subject: Re: [OpenCL-devel] HMAC/EMAC question follow-up & RandPool questi on MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Jun 13 05:02:02 2002 X-Original-Date: Thu, 13 Jun 2002 08:01:22 -0400 Hello, It turns out I can do whatever I needed using the OUTPUT_LENGTH as you suggested. Thanks. The following program seg faults on Linux RedHat v7.2 gcc 2.96. #include #include int main() { try { OpenCL::Randpool rndPool; } catch(OpenCL::Exception& e) { std::cerr << "Caught exception from OpenCL" << std::endl; } return 0; } Any ideas? It seems simple enough. Thanks again, Hany From Hany.Greiss@inbusiness.com Thu Jun 13 05:21:15 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17ITbH-00012T-00 for ; Thu, 13 Jun 2002 05:21:15 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Thu, 13 Jun 2002 08:21:34 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032EE4@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] Version 0.8.3 fails checks Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Jun 13 05:22:03 2002 X-Original-Date: Thu, 13 Jun 2002 08:21:33 -0400 Hello, I should have gone through the checks before, but I hadn't. In any case, here is the output: [hany@hanylap OpenCL-0.8.3]$ ./check --validate Beginning validation tests... Aborted [hany@hanylap OpenCL-0.8.3]$ ./check --benchmark Blowfish: 7.63 Mbytes/sec Aborted This could explain the RandPool problem I was looking into earlier. I'm running Linux RedHat v7.2, it's a 2.4 kernel with gcc v2.96 Thanks, Hany From Hany.Greiss@inbusiness.com Thu Jun 13 05:31:19 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17ITl1-0001sS-00 for ; Thu, 13 Jun 2002 05:31:19 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Thu, 13 Jun 2002 08:31:38 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032EE5@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] More data on the failed checks Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Jun 13 05:32:02 2002 X-Original-Date: Thu, 13 Jun 2002 08:31:37 -0400 Hi, I should have added that I did a make check on version 0.8.2 of OpenCL on the same machine and all tests passed. Hope that helps, Hany From lloyd@acm.jhu.edu Thu Jun 13 08:16:47 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17IWL8-0002uZ-00 for ; Thu, 13 Jun 2002 08:16:46 -0700 Received: from sol.galaxy.acm.jhu.edu (sol.galaxy.acm.jhu.edu [128.220.223.70]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 490C813E99 for ; Thu, 13 Jun 2002 11:16:42 -0400 (EDT) Received: by sol.galaxy.acm.jhu.edu (Postfix, from userid 528) id 15ADAD831; Thu, 13 Jun 2002 11:16:41 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by sol.galaxy.acm.jhu.edu (Postfix) with ESMTP id EDB4778B2 for ; Thu, 13 Jun 2002 11:16:41 -0400 (EDT) From: Jack Lloyd X-X-Sender: To: "'opencl-devel@lists.sourceforge.net'" Subject: Re: [OpenCL-devel] HMAC/EMAC question follow-up & RandPool questi on In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032EE2@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Jun 13 08:17:06 2002 X-Original-Date: Thu, 13 Jun 2002 11:16:41 -0400 (EDT) Well, one problem is that you're not creating a LibraryInitializer object, which is needed for setting up the memory allocators and what-not. However, in that case, it should be throwing an exception telling you that it can't find a memory allocator. Which is what it's doing on my machine. I'm betting, 10 to 1, that it's a compiler bug, because I think I ran into this when building the RPMs an a RedHat 7.3 system. Did you compile OpenCL from source? If so, edit the makefile, remove -fomit-frame-pointer, and recompile. That should do it. However, I cannot be completely sure, because at the moment all the machines I have access to are old Mandrake Linux boxes, which have a different version of GCC (on these machines, your program runs fine). Well, it runs fine and dies from an exception, until I added the line: OpenCL::LibraryInitializer opencl_init; as the first line of the try. But it didn't segfault. Anyway, let let me know if recompiling without -fomit-frame-pointer fixes it. -Jack On Thu, 13 Jun 2002, Hany Greiss wrote: > > Hello, > > It turns out I can do whatever I needed using the OUTPUT_LENGTH as you > suggested. Thanks. > > The following program seg faults on Linux RedHat v7.2 gcc 2.96. > > #include > #include > > int main() { > try { > OpenCL::Randpool rndPool; > } catch(OpenCL::Exception& e) { > std::cerr << "Caught exception from OpenCL" << std::endl; > } > return 0; > } > > > Any ideas? It seems simple enough. > > > Thanks again, > Hany > > > _______________________________________________________________ > > Don't miss the 2002 Sprint PCS Application Developer's Conference > August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink > > From Hany.Greiss@inbusiness.com Thu Jun 13 08:44:29 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17IWlw-0006B5-00 for ; Thu, 13 Jun 2002 08:44:29 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Thu, 13 Jun 2002 11:44:47 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032EE6@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" Subject: RE: [OpenCL-devel] HMAC/EMAC question follow-up & RandPool questi on MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Thu Jun 13 08:45:04 2002 X-Original-Date: Thu, 13 Jun 2002 11:44:46 -0400 Hello, That did the trick. I recompiled the sources with -fomit-frame-pointer removed and sure enough all the tests passed. It's definetly a compiler bug somewhere. I also added the call to the LibraryInitializer (oops!) and my app now works with version 0.8.3. Thanks! Hany -----Original Message----- From: Jack Lloyd [mailto:lloyd@acm.jhu.edu] Sent: Thursday, June 13, 2002 11:17 AM To: 'opencl-devel@lists.sourceforge.net' Subject: Re: [OpenCL-devel] HMAC/EMAC question follow-up & RandPool questi on Well, one problem is that you're not creating a LibraryInitializer object, which is needed for setting up the memory allocators and what-not. However, in that case, it should be throwing an exception telling you that it can't find a memory allocator. Which is what it's doing on my machine. I'm betting, 10 to 1, that it's a compiler bug, because I think I ran into this when building the RPMs an a RedHat 7.3 system. Did you compile OpenCL from source? If so, edit the makefile, remove -fomit-frame-pointer, and recompile. That should do it. However, I cannot be completely sure, because at the moment all the machines I have access to are old Mandrake Linux boxes, which have a different version of GCC (on these machines, your program runs fine). Well, it runs fine and dies from an exception, until I added the line: OpenCL::LibraryInitializer opencl_init; as the first line of the try. But it didn't segfault. Anyway, let let me know if recompiling without -fomit-frame-pointer fixes it. -Jack On Thu, 13 Jun 2002, Hany Greiss wrote: > > Hello, > > It turns out I can do whatever I needed using the OUTPUT_LENGTH as you > suggested. Thanks. > > The following program seg faults on Linux RedHat v7.2 gcc 2.96. > > #include > #include > > int main() { > try { > OpenCL::Randpool rndPool; > } catch(OpenCL::Exception& e) { > std::cerr << "Caught exception from OpenCL" << std::endl; > } > return 0; > } > > > Any ideas? It seems simple enough. > > > Thanks again, > Hany > > > _______________________________________________________________ > > Don't miss the 2002 Sprint PCS Application Developer's Conference > August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink > > _______________________________________________________________ Don't miss the 2002 Sprint PCS Application Developer's Conference August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink -- Opencl-devel mailing list https://lists.sourceforge.net/lists/listinfo/opencl-devel From Hany.Greiss@inbusiness.com Tue Jun 18 08:03:01 2002 Received: from [204.101.222.242] (helo=ottexc00.zoo) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17KKVZ-0007i3-00 for ; Tue, 18 Jun 2002 08:03:01 -0700 Received: by OTTEXC00 with Internet Mail Service (5.5.2653.19) id ; Tue, 18 Jun 2002 11:03:11 -0400 Message-ID: <21D0245AB12FD611A6C5000103E7A7F8032EFF@OTTEXC00> From: Hany Greiss To: "'opencl-devel@lists.sourceforge.net'" MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: text/plain; charset="iso-8859-1" Subject: [OpenCL-devel] RSA Examples Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jun 18 08:03:05 2002 X-Original-Date: Tue, 18 Jun 2002 11:03:10 -0400 Hello, In the change log for v0.8.3, there's mention of some RSA examples. Are they part of the distribution? Thanks, Hany From lloyd@acm.jhu.edu Tue Jun 18 08:09:01 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17KKbM-0000Ay-00 for ; Tue, 18 Jun 2002 08:09:00 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id 4A33113E97; Tue, 18 Jun 2002 11:08:56 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id 4103137B88 for ; Tue, 18 Jun 2002 11:08:56 -0400 (EDT) From: Jack Lloyd To: "'opencl-devel@lists.sourceforge.net'" Subject: Re: [OpenCL-devel] RSA Examples In-Reply-To: <21D0245AB12FD611A6C5000103E7A7F8032EFF@OTTEXC00> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jun 18 08:10:08 2002 X-Original-Date: Tue, 18 Jun 2002 11:08:56 -0400 (EDT) Yes, they are. doc/examples/rsa*.cpp -Jack On Tue, 18 Jun 2002, Hany Greiss wrote: > > Hello, > > > In the change log for v0.8.3, there's mention of some RSA examples. > Are they part of the distribution? > > Thanks, > Hany > > ---------------------------------------------------------------------------- > Bringing you mounds of caffeinated joy > >>> http://thinkgeek.com/sf <<< > > From lloyd@acm.jhu.edu Wed Jun 26 12:11:02 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17NIBx-0001qD-00 for ; Wed, 26 Jun 2002 12:11:01 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id D222413E99; Wed, 26 Jun 2002 15:10:56 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id D160237ADF for ; Wed, 26 Jun 2002 15:10:56 -0400 (EDT) From: Jack Lloyd To: OpenCL Devel List Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] PK algorithms, round 2 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jun 26 12:12:01 2002 X-Original-Date: Wed, 26 Jun 2002 15:10:56 -0400 (EDT) Hello all, I added Nyberg-Rueppel signatures (as spec'ed in IEEE 1363) to OpenCL a couple of days ago. It only took me about 4-5 hours to do the whole thing, which seems to be implying that the PK framework is fairly stable (I didn't have to touch anything outside of nr.* to do it) So, I've been thinking about adding ElGamal encryption and/or signatures. The problem is that there is no standard I know of for how to do either of these. NR sigs are very similiar to ElGamal sigs, and so the additional bonus of them being well standardized made it seem worthwhile. And as there are powerful attacks on plain ElGamal encryption, I'm hesitant to simply toss it out there without the supporting stuff to make it secure. If anyone has any info on this, I could use some pointers to relevant standards (or well-known heuristics), test vectors, etc. So, I now once again ask for some feedback. Send answers either to the list or to me directly, as you prefer. Do you need or want: 1) PK encryption algorithms beyond RSA? - I feel a little nervous that this is the only PK encryption algorithm. But it doesn't seem like anyone really uses anything else, except maybe ElGamal in a few random applications like PGP. 2) PK signature algorithms beyond RSA, DSA, Rabin-Williams, and Nyberg-Rueppel? 3) Key agreement schemes such as Diffie-Hellman? - I was planning on putting this off until 1.1.x (ie sometime this fall). Is this the right choice? 4) Elliptic curve algorithms (like ECDSA)? As far as updates since 0.8.3, I've speeded up PK key generation quite a bit. Now generating a 1024 bit RSA key takes "only" about 4 seconds on a 1.4 Ghz Athlon. Key gen for DL-based algorithms (DSA and NR) still royally sucks (20 seconds - 2 minutes). Thankfully, the primes for such algorithms are public and can be shared among keys, so you can use something like the IPSEC primes for a system. And I'm pretty sure I know why they are so slow (though I'm not sure if I can figure out how to fix it). Other changes... hrm... a bunch of fairly minor bug fixes, mostly. I'm probably going to be adding EMSA4 (aka PSS) before the next release, but haven't gotten around to it just yet. Scheduled release for 0.8.4 is July 7. -Jack From foxcm@purdue.edu Tue Jul 16 12:14:07 2002 Received: from scribe.cc.purdue.edu ([128.210.11.6]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17UXlt-0001zC-00 for ; Tue, 16 Jul 2002 12:14:05 -0700 Received: from scribe.cc.purdue.edu (localhost [127.0.0.1]) by scribe.cc.purdue.edu (8.12.3/8.12.3/scribe) with ESMTP id g6GJDve9010539 for ; Tue, 16 Jul 2002 14:13:57 -0500 (EST) Received: (from nobody@localhost) by scribe.cc.purdue.edu (8.12.3/8.12.3/Submit) id g6GJDvZ3010537 for opencl-devel@lists.sourceforge.net; Tue, 16 Jul 2002 14:13:57 -0500 (EST) X-Authentication-Warning: scribe.cc.purdue.edu: nobody set sender to foxcm@purdue.edu using -f To: opencl-devel@lists.sourceforge.net Message-ID: <1026846837.3d3470753f72f@webmail.purdue.edu> From: foxcm@purdue.edu MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: IMP/PHP IMAP webmail program 2.2.8 Subject: [OpenCL-devel] Undefined reference problem when compiling Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jul 16 12:15:04 2002 X-Original-Date: Tue, 16 Jul 2002 14:13:57 -0500 (EST) Hi all, I just got the tarball of openCL yesterday and it looks pretty neat, but I'm having trouble getting anything to compile. The library itself compiled and installed OK, but any code I try to link to it (including the example files) returns this error: /usr/local/lib/libopencl.so: undefined reference to `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' collect2: ld returned 1 exit status I get this with the default make, and also with this g++ line: g++ -Wall -ansi -L/usr/local/lib -I/usr/local/include -lopencl main.cpp I'm running a Gentoo system with gcc 2.95.3 (I used the --gcc295x option for configuring the library). /usr/local/lib is in my ld.so.conf and I ran ldconfig after installing the libraries. All that I have in my simple main.cpp is the instantiation of the OpenCL::LibraryInitializer, but that is enough to break it. What am I missing that is causing this to fail? Thanks for your help. -- Chuck Fox From lloyd@acm.jhu.edu Tue Jul 16 12:34:28 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17UY5a-00051W-00 for ; Tue, 16 Jul 2002 12:34:26 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id D00D413E98; Tue, 16 Jul 2002 15:34:22 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id BFDA237B88 for ; Tue, 16 Jul 2002 15:34:22 -0400 (EDT) From: Jack Lloyd To: Subject: Re: [OpenCL-devel] Undefined reference problem when compiling In-Reply-To: <1026846837.3d3470753f72f@webmail.purdue.edu> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jul 16 12:35:02 2002 X-Original-Date: Tue, 16 Jul 2002 15:34:22 -0400 (EDT) On Tue, 16 Jul 2002 foxcm@purdue.edu wrote: > Hi all, > I just got the tarball of openCL yesterday and it looks pretty neat, but > I'm having trouble getting anything to compile. The library itself compiled > and installed OK, but any code I try to link to it (including the > example files) returns this error: > > /usr/local/lib/libopencl.so: undefined reference to > `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' > collect2: ld returned 1 exit status > > I get this with the default make, and also with this g++ line: > g++ -Wall -ansi -L/usr/local/lib -I/usr/local/include -lopencl main.cpp That looks OK. You may want to try putting -lopencl after main.cpp. I don't think this is the problem but there is a small chance. Some compilers are picky about that and I can't remember if gcc 2.95.3 is like that or not. (Did you run 'make check'? If so, did that link correctly?) > I'm running a Gentoo system with gcc 2.95.3 (I used the --gcc295x option for > configuring the library). /usr/local/lib is in my ld.so.conf and I ran > ldconfig after installing the libraries. That sounds OK. I haven't been able to test on 2.95.x recently though, because the only machine around that has it (a Solaris/SPARC machine) has a buggy assembler that causes me no end of trouble. > All that I have in my simple main.cpp is the instantiation of the > OpenCL::LibraryInitializer, but that is enough to break it. What am I missing > that is causing this to fail? Thanks for your help. This smells a lot like an ABI problem (GCC's C++ ABI changes pretty often). But if you installed it from source, that shouldn't occur. Just to be on the safe side, I'll ask: You are compiling your main.cpp with the same g++ you used to compile the library, right? I have one other idea about the problem, but if any of these easy cases are to blame I'll just be glad that it wasn't. But if none of what I mentioned above is relevant or works, email back, because at that point I'll be 90% sure what the problem is (and how to fix it). -Jack From foxcm@purdue.edu Wed Jul 17 10:16:10 2002 Received: from scribe.cc.purdue.edu ([128.210.11.6]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17UsPK-0004ZX-00 for ; Wed, 17 Jul 2002 10:16:10 -0700 Received: from scribe.cc.purdue.edu (localhost [127.0.0.1]) by scribe.cc.purdue.edu (8.12.3/8.12.3/scribe) with ESMTP id g6HHG5e9006026; Wed, 17 Jul 2002 12:16:05 -0500 (EST) Received: (from nobody@localhost) by scribe.cc.purdue.edu (8.12.3/8.12.3/Submit) id g6HHG4me006019; Wed, 17 Jul 2002 12:16:04 -0500 (EST) X-Authentication-Warning: scribe.cc.purdue.edu: nobody set sender to foxcm@purdue.edu using -f To: opencl-devel@lists.sourceforge.net Message-ID: <1026926164.3d35a6546fa2c@webmail.purdue.edu> From: foxcm@purdue.edu Cc: foxcm@purdue.edu MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: IMP/PHP IMAP webmail program 2.2.8 Subject: [OpenCL-devel] Problem with linking to openCL Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jul 17 10:17:02 2002 X-Original-Date: Wed, 17 Jul 2002 12:16:04 -0500 (EST) Hi all, I just installed openCL yesterday, and the library itself compiled fine, but everytime I try to compile code linked to the library (including the sample files) I get this error: make main g++ -O2 -ansi -W -Wall -I/usr/local/include -L/usr/local/lib main.cpp -lopencl -o main /usr/local/lib/libopencl.so: undefined reference to `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' collect2: ld returned 1 exit status make: *** [main] Error 1 The simple main file I have only instantiates an object of type OpenCL::LibraryInitializer, and I get the exact same error on all of the sample source files too. The openCL libraries all appear to be OK, and the /usr/local/lib path is in my ld.so.conf, and I ran ldconfig. I'm running Gentoo with gcc-2.95.3, but I used the --gcc295x flag for compiling the libraries. What am I missing here that will make this work? Thanks for your help to make this go, it looks like a cool library. -- Chuck Fox From lloyd@acm.jhu.edu Wed Jul 17 11:22:38 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17UtRd-0008V5-00 for ; Wed, 17 Jul 2002 11:22:37 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id D640813EA1; Wed, 17 Jul 2002 14:22:31 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id D569F37BEA; Wed, 17 Jul 2002 14:22:31 -0400 (EDT) From: Jack Lloyd To: OpenCL Devel List Cc: Subject: Re: [OpenCL-devel] Problem with linking to openCL In-Reply-To: <1026926164.3d35a6546fa2c@webmail.purdue.edu> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: MULTIPART/MIXED; BOUNDARY="-595574305-1591347817-1026929684=:2904" Content-ID: Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jul 17 11:23:07 2002 X-Original-Date: Wed, 17 Jul 2002 14:22:31 -0400 (EDT) This message is in MIME format. The first part should be readable text, while the remaining parts are likely unreadable without MIME-aware tools. Send mail to mime@docserver.cac.washington.edu for more info. ---595574305-1591347817-1026929684=:2904 Content-Type: TEXT/PLAIN; CHARSET=US-ASCII Content-ID: OK. I think I know what the deal is, then. I have attached a new version of pipe.cpp. Copy it into src/ and make/install the library. It basically moves the definition of the NullFilter class out of the function and into an anonymous namespace at the top of the file. I don't think this should be necessary, I strongly suspect this is really a linker bug. But if it can indeed be worked around this way it seems worthwhile to include in the next release, since it's minimally invasive to fix. So let me know if this works for you (or if it doesn't, of course). Regards, Jack On Wed, 17 Jul 2002 foxcm@purdue.edu wrote: > Hi all, > I just installed openCL yesterday, and the library itself compiled fine, > but everytime I try to compile code linked to the library (including the > sample files) I get this error: > > make main > g++ -O2 -ansi -W -Wall -I/usr/local/include -L/usr/local/lib main.cpp -lopencl > -o main > /usr/local/lib/libopencl.so: undefined reference to > `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' > collect2: ld returned 1 exit status > make: *** [main] Error 1 > > The simple main file I have only instantiates an object of type > OpenCL::LibraryInitializer, and I get the exact same error on all of the > sample source files too. The openCL libraries all appear to be OK, and the > /usr/local/lib path is in my ld.so.conf, and I ran ldconfig. > > I'm running Gentoo with gcc-2.95.3, but I used the --gcc295x flag for compiling > the libraries. What am I missing here that will make this work? Thanks for > your help to make this go, it looks like a cool library. > -- Chuck Fox > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > ---595574305-1591347817-1026929684=:2904 Content-Type: TEXT/PLAIN; CHARSET=US-ASCII; NAME="pipe.cpp" Content-Transfer-Encoding: BASE64 Content-ID: Content-Description: Content-Disposition: ATTACHMENT; FILENAME="pipe.cpp" LyoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioNCiogUGlwZSBTb3VyY2UgRmlsZSAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAqDQoqIChDKSAxOTk5LTIwMDIgVGhlIE9wZW5DTCBQcm9q ZWN0ICAgICAgICAgICAgICAgKg0KKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKi8NCg0KI2luY2x1ZGUgPG9wZW5j bC9waXBlLmg+DQojaW5jbHVkZSA8b3BlbmNsL3NlY3F1ZXVlLmg+DQoNCm5h bWVzcGFjZSBPcGVuQ0wgew0KDQpuYW1lc3BhY2Ugew0KDQpjbGFzcyBOdWxs RmlsdGVyIDogcHVibGljIEZpbHRlcg0KICAgew0KICAgcHVibGljOg0KICAg ICAgdm9pZCB3cml0ZShjb25zdCBieXRlIGlucHV0W10sIHUzMmJpdCBsZW5n dGgpDQogICAgICAgICB7IHNlbmQoaW5wdXQsIGxlbmd0aCk7IH0NCiAgIH07 DQoNCn0NCg0KLyoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioNCiogUGlwZSBDb25zdHJ1Y3RvciAgICAgICAgICAg ICAgICAgICAgICAgICAgICAgICAqDQoqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqLw0KUGlwZTo6UGlwZShGaWx0 ZXIqIGYxLCBGaWx0ZXIqIGYyLCBGaWx0ZXIqIGYzLCBGaWx0ZXIqIGY0KQ0K ICAgew0KICAgaW5pdCgpOw0KICAgYXBwZW5kKGYxKTsNCiAgIGFwcGVuZChm Mik7DQogICBhcHBlbmQoZjMpOw0KICAgYXBwZW5kKGY0KTsNCiAgIH0NCg0K LyoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioNCiogUGlwZSBDb25zdHJ1Y3RvciAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAqDQoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqLw0KUGlwZTo6UGlwZShGaWx0ZXIqIGZpbHRl cl9hcnJheVtdLCB1MzJiaXQgY291bnQpDQogICB7DQogICBpbml0KCk7DQog ICBmb3IodTMyYml0IGogPSAwOyBqICE9IGNvdW50OyBqKyspDQogICAgICBh cHBlbmQoZmlsdGVyX2FycmF5W2pdKTsNCiAgIH0NCg0KLyoqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioNCiogSW5p dGlhbGl6ZSB0aGUgUGlwZSAgICAgICAgICAgICAgICAgICAgICAgICAgICAq DQoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqLw0Kdm9pZCBQaXBlOjppbml0KCkNCiAgIHsNCiAgIHBpcGUgPSAw Ow0KICAgZGVmYXVsdF9yZWFkID0gMDsNCiAgIGxvY2tlZCA9IGZhbHNlOw0K ICAgfQ0KDQovKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKg0KKiBEZXN0cm95IHRoZSBQaXBlICAgICAgICAgICAg ICAgICAgICAgICAgICAgICAgICoNCioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKiovDQp2b2lkIFBpcGU6OmRlc3Ry dWN0KEZpbHRlciogdG9fa2lsbCkNCiAgIHsNCiAgIGlmKCF0b19raWxsIHx8 IGR5bmFtaWNfY2FzdDxTZWN1cmVRdWV1ZSo+KHRvX2tpbGwpKQ0KICAgICAg cmV0dXJuOw0KICAgZm9yKHUzMmJpdCBqID0gMDsgaiAhPSB0b19raWxsLT50 b3RhbF9wb3J0cygpOyBqKyspDQogICAgICBkZXN0cnVjdCh0b19raWxsLT5u ZXh0W2pdKTsNCiAgIGRlbGV0ZSB0b19raWxsOw0KICAgdG9fa2lsbCA9IDA7 DQogICB9DQoNCi8qKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqDQoqIFNldCB0aGUgZGVmYXVsdCByZWFkIG1lc3Nh Z2UgICAgICAgICAgICAgICAgICAgKg0KKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKi8NCnZvaWQgUGlwZTo6c2V0 X2RlZmF1bHRfbXNnKHUzMmJpdCBtc2cpDQogICB7DQogICBpZihtc2cgPj0g bWVzc2FnZXMuc2l6ZSgpKQ0KICAgICAgdGhyb3cgSW52YWxpZF9Bcmd1bWVu dCgiUGlwZTo6c2V0X2RlZmF1bHRfbXNnOiBtc2cgbnVtYmVyIGlzIHRvbyBo aWdoIik7DQogICBkZWZhdWx0X3JlYWQgPSBtc2c7DQogICB9DQoNCi8qKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq DQoqIFN0YXJ0IGEgbmV3IG1lc3NhZ2UgICAgICAgICAgICAgICAgICAgICAg ICAgICAgKg0KKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKi8NCnZvaWQgUGlwZTo6c3RhcnRfbXNnKCkNCiAgIHsN CiAgIGlmKGxvY2tlZCkgcmV0dXJuOw0KICAgaWYocGlwZSA9PSAwKQ0KICAg ICAgcGlwZSA9IG5ldyBOdWxsRmlsdGVyOw0KICAgZmluZF9lbmRwb2ludHMo cGlwZSk7DQogICBwaXBlLT5uZXdfbXNnKCk7DQogICBsb2NrZWQgPSB0cnVl Ow0KICAgfQ0KDQovKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKg0KKiBFbmQgdGhlIGN1cnJlbnQgbWVzc2FnZSAg ICAgICAgICAgICAgICAgICAgICAgICoNCioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKiovDQp2b2lkIFBpcGU6OmVu ZF9tc2coKQ0KICAgew0KICAgaWYoIWxvY2tlZCkgcmV0dXJuOw0KICAgcGlw ZS0+ZmluaXNoX21zZygpOw0KICAgbG9ja2VkID0gZmFsc2U7DQogICB9DQoN Ci8qKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqDQoqIEZpbmQgdGhlIGVuZHBvaW50cyBvZiB0aGUgUGlwZSAgICAg ICAgICAgICAgICAgKg0KKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKi8NCnZvaWQgUGlwZTo6ZmluZF9lbmRwb2lu dHMoRmlsdGVyKiBmKQ0KICAgew0KICAgZm9yKHUzMmJpdCBqID0gMDsgaiAh PSBmLT50b3RhbF9wb3J0cygpOyBqKyspDQogICAgICBpZihmLT5uZXh0W2pd ICYmICFkeW5hbWljX2Nhc3Q8U2VjdXJlUXVldWUqPihmLT5uZXh0W2pdKSkN CiAgICAgICAgIGZpbmRfZW5kcG9pbnRzKGYtPm5leHRbal0pOw0KICAgICAg ZWxzZQ0KICAgICAgICAgew0KICAgICAgICAgU2VjdXJlUXVldWUqIHEgPSBu ZXcgU2VjdXJlUXVldWU7DQogICAgICAgICBmLT5uZXh0W2pdID0gcTsNCiAg ICAgICAgIG1lc3NhZ2VzLnB1c2hfYmFjayhxKTsNCiAgICAgICAgIH0NCiAg IH0NCg0KLyoqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioNCiogQXBwZW5kIGEgRmlsdGVyIHRvIHRoZSBQaXBlICAg ICAgICAgICAgICAgICAgICAqDQoqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqLw0Kdm9pZCBQaXBlOjphcHBlbmQo RmlsdGVyKiBmaWx0ZXIpDQogICB7DQogICBpZihsb2NrZWQpDQogICAgICB0 aHJvdyBFeGNlcHRpb24oIkNhbm5vdCBhdHRhY2ggdG8gYSBQaXBlIHdoaWxl IGl0IGlzIGxvY2tlZCIpOw0KICAgaWYoIWZpbHRlcikNCiAgICAgIHJldHVy bjsNCiAgIGlmKGR5bmFtaWNfY2FzdDxTZWN1cmVRdWV1ZSo+KGZpbHRlcikp DQogICAgICB0aHJvdyBJbnZhbGlkX0FyZ3VtZW50KCJQaXBlOjphcHBlbmQ6 IFNlY3VyZVF1ZXVlIGNhbm5vdCBiZSB1c2VkIik7DQoNCiAgIGlmKCFwaXBl KSBwaXBlID0gZmlsdGVyOw0KICAgZWxzZSAgICAgIHBpcGUtPmF0dGFjaChm aWx0ZXIpOw0KICAgfQ0KDQovKioqKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKg0KKiBQcmVwZW5kIGEgRmlsdGVyIHRv IHRoZSBQaXBlICAgICAgICAgICAgICAgICAgICoNCioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKiovDQp2b2lkIFBp cGU6OnByZXBlbmQoRmlsdGVyKiBmaWx0ZXIpDQogICB7DQogICBpZihsb2Nr ZWQpDQogICAgICB0aHJvdyBFeGNlcHRpb24oIkNhbm5vdCBwcmVwZW5kIHRv IGEgUGlwZSB3aGlsZSBpdCBpcyBsb2NrZWQiKTsNCiAgIGlmKCFmaWx0ZXIp DQogICAgICByZXR1cm47DQogICBpZihkeW5hbWljX2Nhc3Q8U2VjdXJlUXVl dWUqPihmaWx0ZXIpKQ0KICAgICAgdGhyb3cgSW52YWxpZF9Bcmd1bWVudCgi UGlwZTo6cHJlcGVuZDogU2VjdXJlUXVldWUgY2Fubm90IGJlIHVzZWQiKTsN Cg0KICAgaWYocGlwZSkgZmlsdGVyLT5hdHRhY2gocGlwZSk7DQogICBwaXBl ID0gZmlsdGVyOw0KICAgfQ0KDQovKioqKioqKioqKioqKioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKg0KKiBQb3AgYSBGaWx0ZXIgb2Zm IHRoZSBQaXBlICAgICAgICAgICAgICAgICAgICAgICoNCioqKioqKioqKioq KioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKiovDQp2b2lk IFBpcGU6OnBvcCgpDQogICB7DQogICBpZihsb2NrZWQpDQogICAgICB0aHJv dyBFeGNlcHRpb24oIkNhbm5vdCBwb3Agb2ZmIGEgUGlwZSB3aGlsZSBpdCBp cyBsb2NrZWQiKTsNCiAgIGlmKCFwaXBlKSByZXR1cm47DQogICBpZihwaXBl LT50b3RhbF9wb3J0cygpID4gMSkNCiAgICAgIHRocm93IEV4Y2VwdGlvbigi Q2Fubm90IHBvcCBvZmYgYSBGaWx0ZXIgd2l0aCBtdWx0aXBsZSBwb3J0cyIp Ow0KICAgRmlsdGVyKiBmID0gcGlwZTsNCiAgIHUzMmJpdCBvd25zID0gZi0+ b3ducygpOw0KICAgcGlwZSA9IHBpcGUtPm5leHRbMF07DQogICBkZWxldGUg ZjsNCg0KICAgd2hpbGUob3ducy0tKQ0KICAgICAgew0KICAgICAgZiA9IHBp cGU7DQogICAgICBwaXBlID0gcGlwZS0+bmV4dFswXTsNCiAgICAgIGRlbGV0 ZSBmOw0KICAgICAgfQ0KICAgfQ0KDQp9DQo= ---595574305-1591347817-1026929684=:2904-- From foxcm@purdue.edu Wed Jul 17 13:36:19 2002 Received: from scribe.cc.purdue.edu ([128.210.11.6]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17UvX0-00070r-00 for ; Wed, 17 Jul 2002 13:36:19 -0700 Received: from scribe.cc.purdue.edu (localhost [127.0.0.1]) by scribe.cc.purdue.edu (8.12.3/8.12.3/scribe) with ESMTP id g6HKaAe9022903 for ; Wed, 17 Jul 2002 15:36:10 -0500 (EST) Received: (from nobody@localhost) by scribe.cc.purdue.edu (8.12.3/8.12.3/Submit) id g6HKaA3x022900 for opencl-devel@lists.sourceforge.net; Wed, 17 Jul 2002 15:36:10 -0500 (EST) X-Authentication-Warning: scribe.cc.purdue.edu: nobody set sender to foxcm@purdue.edu using -f To: opencl-devel@lists.sourceforge.net Subject: Re: [OpenCL-devel] Problem with linking to openCL Message-ID: <1026938170.3d35d53ab0b53@webmail.purdue.edu> From: foxcm@purdue.edu References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: IMP/PHP IMAP webmail program 2.2.8 Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jul 17 13:37:02 2002 X-Original-Date: Wed, 17 Jul 2002 15:36:10 -0500 (EST) Thanks Jack, that new version of pipe.cpp fixed the problem and everything is compiling fine. I don't know if my linker is fubarred (I think it's OK since Gentoo compiles all packages from source and they mostly work), but the new file is definitely better. -- Chuck Quoting Jack Lloyd : > OK. I think I know what the deal is, then. > > I have attached a new version of pipe.cpp. Copy it into src/ and > make/install the library. It basically moves the definition of the > NullFilter class out of the function and into an anonymous namespace at > the > top of the file. > > I don't think this should be necessary, I strongly suspect this is > really a > linker bug. But if it can indeed be worked around this way it seems > worthwhile to include in the next release, since it's minimally invasive > to > fix. So let me know if this works for you (or if it doesn't, of > course). > > Regards, > Jack > > On Wed, 17 Jul 2002 foxcm@purdue.edu wrote: > > > Hi all, > > I just installed openCL yesterday, and the library itself compiled > fine, > > but everytime I try to compile code linked to the library (including > the > > sample files) I get this error: > > > > make main > > g++ -O2 -ansi -W -Wall -I/usr/local/include -L/usr/local/lib main.cpp > -lopencl > > -o main > > /usr/local/lib/libopencl.so: undefined reference to > > > `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' > > collect2: ld returned 1 exit status > > make: *** [main] Error 1 > > > > The simple main file I have only instantiates an object of type > > OpenCL::LibraryInitializer, and I get the exact same error on all of > the > > sample source files too. The openCL libraries all appear to be OK, > and the > > /usr/local/lib path is in my ld.so.conf, and I ran ldconfig. > > > > I'm running Gentoo with gcc-2.95.3, but I used the --gcc295x flag for > compiling > > the libraries. What am I missing here that will make this work? > Thanks for > > your help to make this go, it looks like a cool library. > > -- Chuck Fox > > > > > > > > ------------------------------------------------------- > > This sf.net email is sponsored by:ThinkGeek > > Welcome to geek heaven. > > http://thinkgeek.com/sf > > > From lloyd@acm.jhu.edu Wed Jul 17 15:54:45 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17Uxgy-0007me-00 for ; Wed, 17 Jul 2002 15:54:44 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id F2CA413E97; Wed, 17 Jul 2002 18:54:39 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id F240837C14 for ; Wed, 17 Jul 2002 18:54:39 -0400 (EDT) From: Jack Lloyd To: Subject: Re: [OpenCL-devel] Problem with linking to openCL In-Reply-To: <1026938170.3d35d53ab0b53@webmail.purdue.edu> Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Wed Jul 17 15:55:04 2002 X-Original-Date: Wed, 17 Jul 2002 18:54:39 -0400 (EDT) The particular bug in question is rather obscure, and I would guess only a very small number of software packages would be affected. Well, whatever. Long as it works, I suppose. :) Good luck with your project(s), and let me know if you run into any problems. -Jack On Wed, 17 Jul 2002 foxcm@purdue.edu wrote: > Thanks Jack, that new version of pipe.cpp fixed the > problem and everything is compiling fine. I don't know > if my linker is fubarred (I think it's OK since Gentoo > compiles all packages from source and they mostly work), > but the new file is definitely better. > -- Chuck > > Quoting Jack Lloyd : > > > OK. I think I know what the deal is, then. > > > > I have attached a new version of pipe.cpp. Copy it into src/ and > > make/install the library. It basically moves the definition of the > > NullFilter class out of the function and into an anonymous namespace at > > the > > top of the file. > > > > I don't think this should be necessary, I strongly suspect this is > > really a > > linker bug. But if it can indeed be worked around this way it seems > > worthwhile to include in the next release, since it's minimally invasive > > to > > fix. So let me know if this works for you (or if it doesn't, of > > course). > > > > Regards, > > Jack > > > > On Wed, 17 Jul 2002 foxcm@purdue.edu wrote: > > > > > Hi all, > > > I just installed openCL yesterday, and the library itself compiled > > fine, > > > but everytime I try to compile code linked to the library (including > > the > > > sample files) I get this error: > > > > > > make main > > > g++ -O2 -ansi -W -Wall -I/usr/local/include -L/usr/local/lib main.cpp > > -lopencl > > > -o main > > > /usr/local/lib/libopencl.so: undefined reference to > > > > > `write__Q46OpenCL4Pipe27start_msg__Q26OpenCL4Pipe.0_10NullFilterPCUcUi.1116' > > > collect2: ld returned 1 exit status > > > make: *** [main] Error 1 > > > > > > The simple main file I have only instantiates an object of type > > > OpenCL::LibraryInitializer, and I get the exact same error on all of > > the > > > sample source files too. The openCL libraries all appear to be OK, > > and the > > > /usr/local/lib path is in my ld.so.conf, and I ran ldconfig. > > > > > > I'm running Gentoo with gcc-2.95.3, but I used the --gcc295x flag for > > compiling > > > the libraries. What am I missing here that will make this work? > > Thanks for > > > your help to make this go, it looks like a cool library. > > > -- Chuck Fox > > > > > > > > > > > > ------------------------------------------------------- > > > This sf.net email is sponsored by:ThinkGeek > > > Welcome to geek heaven. > > > http://thinkgeek.com/sf > > > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > From lloyd@acm.jhu.edu Tue Jul 23 06:46:51 2002 Received: from centaur.acm.jhu.edu ([128.220.223.65]) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17X002-0003Vu-00 for ; Tue, 23 Jul 2002 06:46:50 -0700 Received: by centaur.acm.jhu.edu (Postfix, from userid 528) id E3A6A13E9A; Tue, 23 Jul 2002 09:46:44 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by centaur.acm.jhu.edu (Postfix) with ESMTP id E2DAE37BEA for ; Tue, 23 Jul 2002 09:46:44 -0400 (EDT) From: Jack Lloyd To: OpenCL Devel List Message-ID: X-GPG-Key-ID: 4DCDF398 X-GPG-Key-Fingerprint: 2DD2 95F9 C7E3 A15E AF29 80E1 D6A9 A5B9 4DCD F398 X-Red-Robot-Approved: CRUSH ALL HU-MANS! Organization: JHU ACM/CS/SRL MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: [OpenCL-devel] Minor brokenness in 0.8.5 + update Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Tue Jul 23 06:47:04 2002 X-Original-Date: Tue, 23 Jul 2002 09:46:44 -0400 (EDT) I've realized that the pk_types.h header in 0.8.5 is busted. It's fairly easy to patch up the parts that are wrong, or alternately you can switch to the interface in look_pk.h, or continue to use 0.8.4. There's some things that are not currently deprecated but will be in the next release. I mentioned them in the announce notice but I'm fairly sure that some people over here are not on that list so I figured I would mention them over here as well. If you read that (or just plain don't care), you can stop reading here. The current method of creating most of the filters goes like this: SomeFilter(new SomeAlgorithmObject, [other params like keys, etc]) As of 0.8.5 (some of these changes were in 0.8.4 but they were complete in 0.8.5), all of these things can also be done by: SomeFilter("SomeAlgorithm", [other params]) As you can see, this is quite a bit nicer, particular if the type is something strange like "EMAC(Lion(RIPEMD-160,SEAL(4)))" but I've found that even in simple cases like "HMAC(SHA-1)" it's significantly more readable, and has the bonus that you can switch algorithms arbitrarily at runtime. The documentation is seriously out of date in this regard (not to mention in other ways). I've fixed it and added a bunch of new stuff for the next release. I even found references in the section about cipher modes to using template parameters, which changed ages ago. Oops. The reason that I didn't deprecate the old versions in 0.8.5 was because the code handling the name->object mappings (which as you've probably seen in the changelogs, I refer to as the 'lookup' subsystem) didn't handle things like 'HMAC(SHA-1)'; it could only do simple types like 'Blowfish', 'MD5', etc. But, the next release supports those just great, even really complicated ones like the EMAC example above, so the old versions will be deprecated in 0.8.6 and removed in 0.8.7. This affects primarly uses of: StreamCipher_Filter, Hash_Filter, MAC_Filter, and all of the cipher modes. Actually, Hash_Filter's constructor taking a HashFunction* is already deprecated in 0.8.5, because there (were) no hash functions in OpenCL that had any kind of parameterization, and that constructor will be removed in the next release. Basically, I would recommend you convert the ones you can now, and leave the ones that don't work in 0.8.5 (I imagine 95% of these cases will be HMAC) until 0.8.6 is released. Gah. This makes four times I've broken all of the filter code. I think this is finally it, though. :) If you're reading all the way down here, I'm surprised. Regards, Jack From lcheng1@gmu.edu Fri Aug 02 03:43:22 2002 Received: from portalknot.gmu.edu ([129.174.0.8] helo=portal.gmu.edu) by usw-sf-list1.sourceforge.net with esmtp (Exim 3.31-VA-mm2 #1 (Debian)) id 17aZtx-0002bo-00 for ; Fri, 02 Aug 2002 03:43:21 -0700 Received: from gmu.edu (mail02.gmu.edu [129.174.0.112]) by portal.gmu.edu (8.8.8/8.8.8) with ESMTP id GAA16139 for ; Fri, 2 Aug 2002 06:43:19 -0400 (EDT) From: To: opencl-devel@lists.sourceforge.net Message-ID: <14e2a1d14e5823.14e582314e2a1d@gmu.edu> X-Mailer: Netscape Webmail MIME-Version: 1.0 Content-Language: en X-Accept-Language: en Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: 7bit Subject: [OpenCL-devel] Fork behavior Sender: opencl-devel-admin@lists.sourceforge.net Errors-To: opencl-devel-admin@lists.sourceforge.net X-BeenThere: opencl-devel@lists.sourceforge.net X-Mailman-Version: 2.0.9-sf.net Precedence: bulk Reply-To: opencl-devel@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: A list discussing OpenCL development List-Unsubscribe: , List-Archive: Date: Fri Aug 2 07:12:03 2002 X-Original-Date: Fri, 02 Aug 2002 06:43:19 -0400 Hi all, I am a newcomer to OpenCL