- gsll-devel - mailman3.common-lisp.net

[GSLL-devel] Installing gsll with allegro
by ilencaco＠yahoo.com 20 Sep '11

20 Sep '11

hi Liam, I'm new with lisp and I would like to install the gsll in my machine with allegro cl- . However noting seems to work I would appreciate your help. Best regards, Ismael

2 1

[GSLL-devel] Antik, a Common Lisp system that provides a foundation for scientific and engineering computation
by Liam Healy 18 Aug '11

18 Aug '11

Hello GSLLers, Over the past eight months or so, I have been working on a new Lisp system to provide common definitions useful for a variety of scientific and engineering computing. It is called "Antik" and it is now ready to be used by others. Though I say "new", in reality, much of it is old, as it is composed of two parts: one are the grid and foreign-array modules that were already required by GSLL (and formerly supplied by the GSD repository), and the other consist of several modules, most of which I have written over the past decade or more and used myself but never made public. These include elementary generic mathematical functions, physical dimension quantities (i.e., computation with units), and simplified interfaces to some GSLL functions. There are other, less well developed modules, such as one to help organize computation, that I hope will get more complete in the future. There is detailed documentation for the mature parts of Antik. The homepage for Antik is http://www.common-lisp.net/project/antik/. Because Antik has all the grid definitions from GSD, GSLL now uses Antik instead of GSD, and GSD is now obsolete. The grid definitions are for the most part the same as before, with a few changes. First, grid:aref is preferred to grid:gref for grid element access, though the latter is still supported. The reader macro #m by default expands to generate CL arrays, but this can easily be changed to expanding to foreign arrays by changing grid:*default-grid-type*, which GSLL does on loading. Since GSLL needs the grid and foreign-array modules, and most GSLL users are interested in broader scientific computing tasks than just GSL, I thought I would let you know about Antik via this mailing list. If you load GSLL using git, you will need to load Antik from the git repository (see http://repo.or.cz/w/antik.git) If you load GSLL using quicklisp, you will need to wait a bit until the next update, which should automatically bring in the right systems. An independent change to GSLL is that there is no longer a separate system gsll-tests that needs to be loaded. Instead, the tests are defined by an asdf-system-connections system, so that if gsll and lisp-unit are both loaded (in either order), the tests will be defined. When convenient, please switch to the new GSLL and Antik. Your comments are welcome; there is a mailing list for Antik at http://common-lisp.net/cgi-bin/mailman/listinfo/antik-devel. Liam

1 0

[GSLL-devel] GSD fails to load in CCL
by Leo 29 Jul '11

29 Jul '11

Hello, I am running "Clozure Common Lisp Version 1.7-dev-r14818M-trunk (DarwinX8664)" and GSD is from quicklisp. It fails to load with this error: Bound is not *, a SINGLE-FLOAT or a list of a SINGLE-FLOAT: GRID::VALUE [Condition of type SIMPLE-ERROR] Restarts: 0: [RETRY-COMPILE-FILE] Retry compiling #P"/Users/leo/quicklisp/dists/quicklisp/software/gsd-20101207-git/foreign-array/methods.lisp" 1: [SKIP-COMPILE-FILE] Skip compiling #P"/Users/leo/quicklisp/dists/quicklisp/software/gsd-20101207-git/foreign-array/methods.lisp" 2: [TRY-RECOMPILING] Try recompiling methods 3: [RETRY] Retry compiling #<CL-SOURCE-FILE "foreign-array" "methods">. 4: [ACCEPT] Continue, treating compiling #<CL-SOURCE-FILE "foreign-array" "methods"> as having been successful. 5: [ABORT] Give up on "gsll" --more-- Backtrace: 0: (CCL::VALUES-SPECIFIER-TYPE (SINGLE-FLOAT GRID::VALUE) #<CCL::LEXICAL-ENVIRONMENT #x30200187A93D>) Locals: CCL::SPEC = (SINGLE-FLOAT GRID::VALUE) CCL::ENV = #<CCL::LEXICAL-ENVIRONMENT #x30200187A93D> CONDITION = #<SIMPLE-ERROR #x30200187A89D> CCL::TYPE-CACHE-CTYPES = #(0 0 0 0 0 0 ...) CCL::HITS = 1839929 CCL::TYPE-CACHE-SPECS = #(0 0 0 0 0 0 ...) CCL::PROBES = 1865864 CCL::LOCKED = NIL LOCK = #<RECURSIVE-LOCK [ptr @ #x5204C0] #x3020000A04BD> Leo

2 1

[GSLL-devel] hi!
by Angelo Rossi 29 Jul '11

29 Jul '11

Hello, I'm new to GSLL and I like to know if there are general function to perform operations between matrices like add, sum and so on... Thanks Angelo Rossi

2 1

Re: [GSLL-devel] matrix-product: transpose keyword resulting in memory fault
by Liam Healy 29 Jul '11

29 Jul '11

On Fri, Jul 29, 2011 at 4:31 PM, Liam Healy <lhealy(a)common-lisp.net> wrote: > Please post a complete (all definitions provided) example of this error (as > small as possible please). Thanks. > > Liam > > > On Sun, Jul 17, 2011 at 12:07 PM, Ravi Desai <rd7190(a)gmail.com> wrote: > >> Hello all, >> I needed to do a matrix-vector multiplication where the matrix would >> have to be transposed (I checked the dimensions). I tried to use the >> transpose functionality in the matrix-product function itself. >> >> (matrix-product v (f1 pattern-index) v0 1.0 1.0 :trans :notrans) >> where 'v' is a vector of weights and (f1 pattern-index) returns a vector, >> and v0 was an empty vector I'd made just prior to this line in the let >> definition. >> >> It compiled without any errors, but would say corruption error, and memory >> fault at 1 (pc=...., sp=...), and that the integrity of this image is >> possibly compromised, continuing with fingers crossed. >> >> when I removed the line defining v0 and placed the following instead, >> >> (matrix-product (matrix-transpose v) (f1 pattern-index)) >> >> The thing compiled perfectly and ran without any errors. >> >> To be quite honest I prefer using it this way since its more descriptive >> of whats going on. >> >> Thanks >> >> ______________________________**_________________ >> GSLL-devel mailing list >> GSLL-devel(a)common-lisp.net >> http://lists.common-lisp.net/**cgi-bin/mailman/listinfo/gsll-**devel<http://lists.common-lisp.net/cgi-bin/mailman/listinfo/gsll-devel> >> > >

1 0

[GSLL-devel] How to use transpose keywords with matrix-product?
by Ravi Desai 29 Jul '11

29 Jul '11

Hello all, I'm unable to find out how to use the :trans, :notrans, and :conjtrans keywords when using the matrix-product function. Could someone help me out with this? Thanks

2 1

[GSLL-devel] Re Installation problems
by Laurence Aitchison 29 Jul '11

29 Jul '11

I am trying to install GSLL in SBCL. When I type (ql:quickload "gsll-tests") I get the following error message (CFFI::FL-ERROR "Unable to load foreign library (~A).~% ~A" #:LIBRARY-818 "Error opening shared object \"libffi.so\": libffi.so: cannot open shared object file: No such file or directory.") I have just installed the latest version of libffi, and libffi.so exists and is located at: /usr/local/lib/libffi.so Thank you! Laurence

2 1

[GSLL-devel] Bug in dirichlet-log-pdf?
by Nils Bertschinger 29 Jul '11

29 Jul '11

Hi, it appears that dirichlet-log-pdf is broken: CL-USER> (gsll:dirichlet-log-pdf #m(1.0d0 2.0d0) #m(0.3d0 0.7d0)) 0.0d0 CL-USER> (gsll:dirichlet-log-pdf #m(2.0d0 1.0d0) #m(0.7d0 0.3d0)) -0.35667494393873245d0 Obviously, the Dirichlet distribution is permutation invariant, so both results should be the same! By the way, both values are wrong. A straight forward implementation of the formula for the Dirichlet distribution gives: (defun dirichlet-log-pdf (alphas ps) (let ((norm (loop for a across alphas sum (gsll:log-gamma a) into num sum a into denom finally (return (- num (gsll:log-gamma denom)))))) (- (reduce #'+ (map 'vector #'(lambda (p a) (* (- a 1.0d0) (log p))) ps alphas)) norm))) CL-USER> (dirichlet-log-pdf #(1.0d0 2.0d0) #(0.3d0 0.7d0)) 0.33647223662121384d0 CL-USER> (dirichlet-log-pdf #(2.0d0 1.0d0) #(0.7d0 0.3d0)) 0.33647223662121384d0 CL-USER> as it should be. Maybe I'm doing something wrong, when calling the gsl library function. Otherwise, I suspect there is a problem with passing arrays to the GSL. Broken memory layout? I'm using the following version of SBCL: CL-USER> (lisp-implementation-version) "1.0.45.0.debian" CL-USER> (software-version) "2.6.35-28-server" CL-USER> gsl:*gsl-version* "1.14" and my GSLL is from gsll-20101207-git as packaged in quicklisp. Any ideas of what is going on here? Thanks, Nils

2 2

[GSLL-devel] matrix-product: transpose keyword resulting in memory fault
by Ravi Desai 17 Jul '11

17 Jul '11

Hello all, I needed to do a matrix-vector multiplication where the matrix would have to be transposed (I checked the dimensions). I tried to use the transpose functionality in the matrix-product function itself. (matrix-product v (f1 pattern-index) v0 1.0 1.0 :trans :notrans) where 'v' is a vector of weights and (f1 pattern-index) returns a vector, and v0 was an empty vector I'd made just prior to this line in the let definition. It compiled without any errors, but would say corruption error, and memory fault at 1 (pc=...., sp=...), and that the integrity of this image is possibly compromised, continuing with fingers crossed. when I removed the line defining v0 and placed the following instead, (matrix-product (matrix-transpose v) (f1 pattern-index)) The thing compiled perfectly and ran without any errors. To be quite honest I prefer using it this way since its more descriptive of whats going on. Thanks

1 0

Re: [GSLL-devel] How to use transpose keywords with matrix-product?
by Ravi Desai 17 Jul '11

17 Jul '11

Hello all, I found when trying to use it with SLIME and typing in in a file that the set of possible arguments showed up on emacs. I couldn't find that in the documentation of the function itself. I'm guessing that the TransA and TransB options can only be used when everything else is defined. Thanks

1 0