Comments for Homepage of Mate Soos http://www.msoos.org A blog about SAT solving and cryptography Fri, 03 Feb 2012 13:44:27 +0000 hourly 1 http://wordpress.org/?v=3.3.1 Comment on FPGA programming and SAT by Car http://www.msoos.org/2010/09/fpga-programming-and-sat/#comment-96 Car Fri, 03 Feb 2012 13:44:27 +0000 http://www.msoos.org/?p=401#comment-96 I am wondering the same myself! I mean you can map NDA to FPGAs.. But the SAT problem requires state, right? I need to look at the FPGA/SAT literature to see what has been done. It's a very interesting problem. I am wondering the same myself! I mean you can map NDA to FPGAs.. But the SAT problem requires state, right? I need to look at the FPGA/SAT literature to see what has been done. It’s a very interesting problem.

]]> Comment on anf2cnf hell in Sage by msoos http://www.msoos.org/2010/09/anf2cnf-hell-in-sage/#comment-95 msoos Tue, 31 Jan 2012 09:06:52 +0000 http://www.msoos.org/?p=416#comment-95 Hi, Yes, I think <a href="http://www.msoos.org/2010/09/anf2cnf-script-released/" rel="nofollow"><strong>this</strong></a> one gets around these problems. Cheers, Mate Hi,

Yes, I think this one gets around these problems.

Cheers,

Mate

]]> Comment on Thoughts on SAT@home by SAT@home project is going ahead - Start - International Desktop Grid Federation http://www.msoos.org/2011/12/thoughts-on-sat-at-home/#comment-88 SAT@home project is going ahead - Start - International Desktop Grid Federation Sun, 22 Jan 2012 09:25:26 +0000 http://www.msoos.org/?p=1066#comment-88 [...] SAT@home BOINC project (http://sat.isa.ru/pdsat/) is aimed at solving hard problems that can be effectively reduced to the Boolean satisfiability problem. This project was implemented by two IDGF members: Institute for System Dynamics and Control Theory of the Siberian Branch of the Russian Academy of Sciences and Institute for Systems Analysis of Russian Academy of Sciences. Now this project is rapidly growing. New results are regularly obtained and published on the project web page. The project already attracted a worldwide attention. For instance recently it was discussed in the SAT Blog by Mate Soos. The record is available from here: http://www.msoos.org/thoughts-on-sat-at-home. [...] [...] SAT@home BOINC project (http://sat.isa.ru/pdsat/) is aimed at solving hard problems that can be effectively reduced to the Boolean satisfiability problem. This project was implemented by two IDGF members: Institute for System Dynamics and Control Theory of the Siberian Branch of the Russian Academy of Sciences and Institute for Systems Analysis of Russian Academy of Sciences. Now this project is rapidly growing. New results are regularly obtained and published on the project web page. The project already attracted a worldwide attention. For instance recently it was discussed in the SAT Blog by Mate Soos. The record is available from here: http://www.msoos.org/thoughts-on-sat-at-home. [...]

]]> Comment on anf2cnf hell in Sage by cgoyet http://www.msoos.org/2010/09/anf2cnf-hell-in-sage/#comment-53 cgoyet Thu, 19 Jan 2012 20:31:00 +0000 http://www.msoos.org/?p=416#comment-53 Hi mate, in fact your second example ([x1*x2 + 1, x1*x2 + x1]) needs only 3 clauses and 2 variables: p cnf 3 3 c x1*x2 + 1 1 2 c x1*x2 + x1 = x1*(x2+1) -1 2 Is there a way to detect and to simplify these kind of unnecessary clauses ? Hi mate,
in fact your second example ([x1*x2 + 1, x1*x2 + x1]) needs only 3 clauses and 2 variables:
p cnf 3 3
c x1*x2 + 1
1
2
c x1*x2 + x1 = x1*(x2+1)
-1 2

Is there a way to detect and to simplify these kind of unnecessary clauses ?

]]> Comment on AMD’s OpenCL heaven and hell by Vincent Hindriksen http://www.msoos.org/2012/01/amds-opencl-heaven-and-hell/#comment-94 Vincent Hindriksen Tue, 10 Jan 2012 11:32:23 +0000 http://www.msoos.org/?p=1079#comment-94 Thank you for this review of AMD OpenCL on Linux. A honest review works better than a big smile. Just be sure that if they solve the problems, to mention that on this page. As I develop OpenCL on Linux, I recognise the problems with stability of AMD's drivers on Linux. At AMD they know about these problems and have taken it more seriously lately, as their drivers were blacklisted for i.e. WebGL. At Phoronix there has been some updates on this. Trust me, I bugged them A LOT with Linux driver issues. Have you tried the latest 11.12-drivers? They are much, much more stable. You missed that you cannot use Radeons in a headless configuration, making it impossible to use it as a compute-device only. Also that it is quite cumbersome to program host-code that transfers data really fast, and the super-fast "zero-copy" is not implemented under Linux. Their AMD APP Fusion competition is Windows-only, giving us Linux-devs a disadvantage from the start. Part 1 of the "The OpenCL Programming Guide" is simply a waste of paper, but the second part with the examples is really useful. "OpenCL in Action" is a much better read, and what I see around it is currently the most popular . "Heterogeneous Computing with OpenCL" is very AMD-orientated, but together with "CUDA by example" and the various optimisation-guides it forms a good base for GPGPU with OpenCL. More resources are on my website, if you are looking for more. Copy&paste is hopefully solved when OpenCL is going to support templates. Each new OpenCL-compiler is better, so the optimisation-problems certainly will get less in time. As you discovered a lot of responsibility is given to the developer to take care of memory-type choices, buffer-sizes and optimisations - it can sometimes be very frustrating, can't it? :) Integration in other IDEs than VisualStudio is my big wish too, as highlighting and project-wizards are only a part of what is needed. Thank you for this review of AMD OpenCL on Linux. A honest review works better than a big smile. Just be sure that if they solve the problems, to mention that on this page.

As I develop OpenCL on Linux, I recognise the problems with stability of AMD’s drivers on Linux. At AMD they know about these problems and have taken it more seriously lately, as their drivers were blacklisted for i.e. WebGL. At Phoronix there has been some updates on this. Trust me, I bugged them A LOT with Linux driver issues. Have you tried the latest 11.12-drivers? They are much, much more stable.

You missed that you cannot use Radeons in a headless configuration, making it impossible to use it as a compute-device only. Also that it is quite cumbersome to program host-code that transfers data really fast, and the super-fast “zero-copy” is not implemented under Linux. Their AMD APP Fusion competition is Windows-only, giving us Linux-devs a disadvantage from the start.

Part 1 of the “The OpenCL Programming Guide” is simply a waste of paper, but the second part with the examples is really useful. “OpenCL in Action” is a much better read, and what I see around it is currently the most popular . “Heterogeneous Computing with OpenCL” is very AMD-orientated, but together with “CUDA by example” and the various optimisation-guides it forms a good base for GPGPU with OpenCL. More resources are on my website, if you are looking for more.

Copy&paste is hopefully solved when OpenCL is going to support templates. Each new OpenCL-compiler is better, so the optimisation-problems certainly will get less in time. As you discovered a lot of responsibility is given to the developer to take care of memory-type choices, buffer-sizes and optimisations – it can sometimes be very frustrating, can’t it? :) Integration in other IDEs than VisualStudio is my big wish too, as highlighting and project-wizards are only a part of what is needed.

]]>