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

Antik split systems
by Liam Healy 04 Dec '15

04 Dec '15

I am planning to make a major branch merge for Antik soon which affects Antik and GSLL users. A couple years ago I reorganized Antik to split it up into separate systems so that, for example, if you just want to use GSLL, it will not load some parts of Antik that GSLL doesn't need. GSLL also split into two branches so that it could track that Antik branch. I made all updates in the new branches, so the master branches have not seen any progress in the last couple years. This weekend, prior to the Quicklisp December release, I plan to merge these branches into their respective master branches. I am sending this email because of the user-visible changes. On the current master branch of Antik, some global changes were made, notably to the reader, which meant that loading Antik affected completely different systems, which is undesirable. So, the readtable now must be changed manually if you want to use the Antik reader macros #m (grids), #d (dates and times), #_ (physical dimension quantities). Also, *read-default-float-format* was set to 'double-float and this no longer happens. As a convenience, I have defined (set-reader) (set-reader-in-file) to restore the old behavior. They must be invoked manually however, in each REPL or file (respectively) in which you want the desired reader behavior. In Antik, I've started to rewrite the documentation using Sphinx. However, this effort has just begun (and there are some problems with Sphinx I need to fix), so the texinfo documentation is still there. All the changes related to the new branch are in the Sphinx documentation. Thanks, Liam

1 0

Slight recode of gsll/data/permutations.lisp (Re: Bugfix for gsll/data/permutations.lisp)
by Mirko Vukovic 28 Nov '15

28 Nov '15

Liam, I am not trying to bludgeon this issue to death nor convince the you that I am right and you are wrong. In a nutshell, I started on this path because of exception errors when trying to load GSLL on Windows (SBCL or CCL) using GSL compiled MinGW64. I traced this to the initialization of the permutation structures and the size of the requested memory (see recent post on CFFI mailing list). I decided to rewrite a small part of permutation.lisp to use GSL's code to directly initialize and query permutation structure. The attached patch contains the rewrite and some minor edits. All the permutation and qrpt tests pass. The patch should be considered a proof-of-concept (or failure of concept). For me, this patch has cleared the exception errors that have started me on this trek. Even if it is not accepted, I learned something, and it was fun. And I can keep it on my GSLL so that it runs for me. Best, Mirko On Thu, Nov 26, 2015 at 10:24 AM, Liam Healy <lhealy(a)common-lisp.net> wrote: > Added thought: you can lookup any GSL (C) function to find the CL > equivalent by using gsl-lookup. So for example > > (gsl:gsl-lookup "gsl_permute_vector") > PERMUTE > T > > tells you #'permute is the function you want. If there is no > equivalent (there are some C functions with no interface in CL), you > will get NIL as the return value of this function. > > Liam > > > On Wed, Nov 25, 2015 at 11:20 PM, Mirko Vukovic <mirko.vukovic(a)gmail.com> > wrote: > > Thanks for the explanations - I missed the :generic and :method > specifiers. > > I'll study the macro-expansions. > > > > Sorry for the noise. > > > > Mirko > > > > On Tue, Nov 24, 2015 at 10:16 PM, Liam Healy <lhealy(a)common-lisp.net> > wrote: > >> > >> The original code looks right to me. > >> > >> You have taken the generic function and the associated foreign vector > >> methods #'permute and gratuitously renamed them #'permute-vector, > >> leaving the method for raw C pointer with the original name and no > >> generic function. Then you completely delete the generic function and > >> vector methods for #'permute-inverse for no apparent reason, leaving a > >> method for the raw C pointer only. > >> > >> There is no duplicated code here. There is certainly the equivalent of > >> gsl_permute_vector, it is the GRID:VECTOR-DOUBLE-FLOAT (second arg) > >> method of #'permute (which you renamed). > >> > >> I recommend macroexpansion as a way of seeing what's being defined. If > >> you use emacs, place the cursor on the defmfun line and do C-c C-m. > >> Then you will see all the generic functions and methods, and you will > >> see there is no error in the original code. > >> > >> On Tue, Nov 24, 2015 at 5:43 PM, Mirko Vukovic <mirko.vukovic(a)gmail.com > > > >> wrote: > >> > Because of a typo, GSLL did not have the equivalent of > >> > GSL_PERMUTE_VECTOR. > >> > > >> > There was also a section of duplicated code. > >> > > >> > This patch should fix these errors. > >> > > >> > NOTE: I did not test this patch - My GSLL system is not behaving > >> > super-cleanly on MSYS2 and GSL2.1. Proceed with care. > >> > > >> > Mirko > > > > >

1 0

Bugfix for gsll/data/permutations.lisp
by Mirko Vukovic 26 Nov '15

26 Nov '15

Because of a typo, GSLL did not have the equivalent of GSL_PERMUTE_VECTOR. There was also a section of duplicated code. This patch should fix these errors. NOTE: I did not test this patch - My GSLL system is not behaving super-cleanly on MSYS2 and GSL2.1. Proceed with care. Mirko

2 3

Trying to write the interface to gsl_multifit_fdfsolver_jac
by Mirko Vukovic 13 Nov '15

13 Nov '15

I am trying to write the interface to this gsl function: gsl_multifit_fdfsolver_jac (I believe introduced in gsl 2.0) that is necessary for nonlinear-least-squares.lisp (It is returns the Jacobian) See documentation <https://www.gnu.org/software/gsl/manual/html_node/Computing-the-covariance-…> and example usage: gsl_multifit_fdfsolver_jac(s, J); gsl_multifit_covar (J, 0.0, covar); This is my failed attempt after looking at gsll code and documentation (defmfun jacobian (solver) "gsl_multifit_fdfsolver_jac" (((mpointer solver) :pointer) (*jacobian :pointer) :definition :function :inputs (solver) :outputs (*jacobian) :c-return :success-failure :return ((make-foreign-array-from-mpointer *jacobian)) :documentation "Query the solver for the Jacobian, a pointer") Can someone tell me how to make this work? Thanks, Mirko

1 0

GSL 2.x news
by Mirko Vukovic 11 Nov '15

11 Nov '15

This email is related to recent releases of GSL and (not GSLL): GSL 2.0 and 2.1 have been released (see http://www.gnu.org/software/gsl/ and http://git.savannah.gnu.org/cgit/gsl.git/tree/NEWS) 2.0 is now distributed by msys2 on Windows. 2.0 brought new functionality, changes to the interface and some new bugs. As far as I know gsll has not been updated yet (and that is not a critique). 2.0 compiles on Windows on CCL and SBCL with the exception of two issues: - solve-minimize-fit\solver-struct__grovel.c - There is no class named GSLL::BASIS-SPLINE. I am not sure if these errors are related to the bugs discussed in this email thread: http://lists.gnu.org/archive/html/help-gsl/2015-11/msg00006.html) On SBCL 1.30 the test summary is: Unit Test Summary | 3722 assertions total | 3658 passed | 64 failed | 19 execution errors | 0 missing tests #<TEST-RESULTS-DB Total(3722) Passed(3658) Failed(64) Errors(19)> All execution errors have to do with exception_access_violation. Mirko

1 0

A question about numerical-integration.lisp
by Alasdair McAndrew 02 Nov '15

02 Nov '15

Hello, I am using the excellent gsll package (in the first instance), to provide an interface to quadpack for the mathematics system FriCAS. I'm slowly going through calculus/numerical-integration.lisp one function at a time and writing each one into FriCAS as I go. This means writing functions in FriCAS's own language SPAD which interface with gsll. This is made harder by me being a lisp newbie. However, I've come to integration-QAGP, and I've tried to run the test command (having first defined the function it calls). However, all I get is errors. I'm using SBCL in emacs-slime, and I have installed gsll with quicklisp, so that I can call an integration routine with * (gsll:integration-qng (lambda (x) (exp (- (* x x)))) 0.0 1.0) But the commands * (defun integration-test-f454 (x) (* (expt x 3) (* (log (abs (* (- (expt x 2) 1.0d0) (- (expt x 2) 2.0d0))))))) * (gsll:integration-QAGp 'integration-test-f454 (grid:copy-to (vector 0.0d0 1.0d0 (sqrt 2.0d0) 3.0d0)) 0.0d0 1.0d-3 1000) just produces a long list of errors. Is "grid" a standard library, or does it need to be loaded first? (Told you I was a newbie...) Thanks Alasdair -- [image: http://www.facebook.com/alasdair.mcandrew] <http://www.facebook.com/alasdair.mcandrew> [image: https://plus.google.com/+AlasdairMcAndrew/posts] <https://plus.google.com/+AlasdairMcAndrew/posts> [image: https://www.linkedin.com/pub/alasdair-mcandrew/a/178/108] <https://www.linkedin.com/pub/alasdair-mcandrew/a/178/108> [image: https://twitter.com/amca01] <https://twitter.com/amca01> [image: http://numbersandshapes.net] <http://numbersandshapes.net>

3 6

[Gsll-devel] [patch] fix gsl-config-pathname
by Leo Liu 03 Aug '15

03 Aug '15

GSLL fails to load on osx with homebrew-installed GSL. Could someone install this patch? Thanks. --- a/gsll-20140211-git/init/init.lisp 2014-04-28 11:48:43.000000000 +0800 +++ b/gsll-20140211-git/init/init.lisp 2014-04-28 12:15:59.000000000 +0800 @@ -63,7 +63,11 @@ (read-line s))) #+unix (defun gsl-config-pathname (pn) - (merge-pathnames pn (pathname (format nil "~a/" (gsl-config "--prefix")))))) + (flet ((libpath () + (let ((s (gsl-config "--libs-without-cblas"))) + ;; example output: -L/usr/local/Cellar/gsl/1.16/lib -lgsl + (pathname (format nil "~a/" (subseq s 2 (- (length s) 6))))))) + (merge-pathnames pn (libpath))))) (cffi:define-foreign-library libgslcblas (:darwin #+ccl #.(ccl:native-translated-namestring Diff finished. Mon Apr 28 12:17:35 2014

3 3

run-tests results on OS X mavericks
by Munawar Cheema 02 Aug '15

02 Aug '15

I installed gsll from quicklisp. I had to manually give values for the libgsl and libgslclas to get the install to work. Just to be sure it was installed correctly I ran the following: CL-USER> (in-package :gsl) #<Package "GSLL"> GSL> (lisp-unit:run-tests) Unit Test Summary | 3983 assertions total | 3964 passed | 19 failed | 6 execution errors | 0 missing tests #<TEST-RESULTS-DB Total(3983) Passed(3964) Failed(19) Errors(6)> GSL> -- Are these results consistent with expectations or should I be getting no failures or execution errors? -- Kind Regards Munawar

3 3

New location for Antik and GSLL git repositories
by Liam Healy 16 Apr '15

16 Apr '15

The git repositories and development pages for Antik and GSLL are now on gitlab.common-lisp.net under the "antik" group: https://gitlab.common-lisp.net/groups/antik If you clone/pull directly from the git repositories, update the URL as the old locations have been deleted. Those who get the projects exclusively from Quicklisp are unaffected by this change. Liam

1 1

[Gsll-devel] Recurring callbacks
by Eugene Zhemchugov 18 Jan '15

18 Jan '15

I was trying to evaluate a recurring integral when I ran into what I think a design flaw in the callbacks implementation. Consider the following code: (integration-qag (lambda (x) (integration-qag (lambda(y) (* (sin x) y)) 0d0 1d0 :gauss41)) 0d0 pi :gauss41) The result of the integration should be 1, but I get 4.93. Digging through the code, I figured out that callbacks are implemented in the following way: 1. During the GSLL compilation, a callback is defined via the CFFI:DEFCALLBACK macro. 2. The GSLL wrapper function assigns the function to be called to a dynamic variable and calls the underlying GSL function. 3. The GSL function calls the CFFI callback. 4. The CFFI callback calls the value stored in the dynamic variable. The relevant part of the GSLL wrapper function is generated by BODY-EXPAND. Here is a simplified version of the code generated for INTEGRATION-QAG: (defun integration-qag (function ...) (declare (special integration-qag-dynfn0)) (setf integration-qag-dynfn0 function) (foreign-funcall "gsl_integration_qag" ...)) The problem here is that if the function passed as a parameter to the GSLL function calls the same GSLL function again, as in the example above, the dynamic variable value is overwritten during the second call, and then the first GSL function ends up calling the second callback all the time. Seemingly, this can be fixed by binding instead of assigning the dynamic variable, i.e., by making BODY-EXPAND to generate code like this: (defun integration-qag (function ...) (let ((integration-qag-dynfn0 function)) (declare (special integration-qag-dynfn0)) (foreign-funcall "gsl_integration_qag" ...))) In this case the value of INTEGRATION-QAG-DYNFN0 is restored upon leaving the LET form. However, having patched the code this way, I broke callbacks introduced via the DEFMOBJECT macro. In this case the assignment happens in the REINITIALIZE-INSTANCE method specialized on this object, which is usually called in the construction phase. As an example, consider NONLINEAR-LEAST-SQUARES-EXAMPLE. The assignment is made during the call to MAKE-NONLINEAR-FDFFIT, however the callback is executed during the call to ITERATE. I think that the binding should be performed in ITERATE, as soon as it is needed. ITERATE receives an instance of an object subclassed from CALLBACK-INCLUDED, hence it has access both to the functions to be called (the FUNCALLABLES slot of CALLBACK-INCLUDED) and the callback specification (the CBINFO slot). But I don't know what would be the best way to implement it. One possibility is to loop through function parameters in BODY-EXPAND and figure out which are instances of CALLBACK-INCLUDED. That will work only for methods, and only for the specialized parameters. Is it sufficient? Other approaches I can think of would require changing the DEFMFUN interface to specify which parameters should be used to construct the bindings. What solution would you propose? I can make a patch for it if you don't think that the changes are too drastic to trust not an upstream developer. You may find my partial patch in the attachement.

2 1