Liam M. Healy [Thu, 3 Jan 2013 23:21:01 +0000]
Improved docstring for #'linear-fit
Liam M. Healy [Sat, 2 Mar 2013 21:13:56 +0000]
Fix flysheet to indicate dependence on new version of CFFI
Liam M. Healy [Wed, 2 Jan 2013 03:40:40 +0000]
Improved docstrings
Liam M. Healy [Mon, 31 Dec 2012 22:58:25 +0000]
Fix typo in docstring
Liam M. Healy [Sun, 18 Nov 2012 16:27:24 +0000]
Substitute for dim0 and dim1 in funcallables when scalarsp=NIL
Previously, the funcallable form made by #'make-funcallable-form (for
e.g. ODE solvers) when scalarsp=T involved substituting the symbols
'dim0 and 'dim1, but when scalarsp=NIL, would not do the substitution.
This caused an error when those unevaluated and unbound symbols
appeared in the lambda. By altering #'faify-form with an additional
'dimension-values argument and calling
(value-from-dimensions argspec dimension-values)
instead of
(parse-callback-argspec argspec 'dimensions),
this problem is fixed.
Liam M. Healy [Sun, 19 Feb 2012 15:42:07 +0000]
Rename dependent system cffi-libffi, fix Darwin path
Liam M. Healy [Fri, 13 Jan 2012 17:12:31 +0000]
Foreign type name :sizet from CFFI replaces 'cffi:sizet
Liam M. Healy [Tue, 3 Jan 2012 23:05:55 +0000]
Optional index arrays for sort-smallest-index, sort-largest-index
Arguments simplified in sort-smallest-index, sort-largest-index;
output array is optional and may be specified by size. Docstrings
made accurate. Previous versions selected only the first n part of
the unsorted array. Other functions in this file could benefit from
similar changes, but the changes have not been made because of the
existence of tests.
Liam M. Healy [Fri, 25 Nov 2011 03:58:02 +0000]
Fix special function return sf-result and sf-result-e10 type specifications
Liam M. Healy [Sun, 30 Oct 2011 14:35:32 +0000]
Use (:struct foo) syntax to eliminate style warnings from CFFI
Liam M. Healy [Sun, 30 Oct 2011 04:44:08 +0000]
Foreign structure with CFFI's new syntax
Specify the foreign structures with CFFI's new syntax, (:struct foo),
and fix #'creturn-st so that it will not take :struct (or any other
keyword) as a return variable name. GSLL now compiles and loads
without error or warning, but it has not been tested.
Liam M. Healy [Sat, 29 Oct 2011 23:35:40 +0000]
Fix sf-result* definitions, eliminate ffexpand, fsbv:object
Fix sf-result and sf-result-e10 translation methods, and use in
complex-with-error and values-with-errors. Some usage in gamma.lisp
updated, but there will be a need to replace sf-result with (:struct
sf-result) in all declarations for many special functions.
I removed ffexpand and everything it depended on because there is no
reason to make a side defcfun for call-by-value functions, and if FSBV
is missing CFFI already emits an error. All usage of fsbv:object and
any thing from the FSBV package is now gone.
Liam M. Healy [Mon, 24 Oct 2011 02:21:19 +0000]
Eliminate fsbv:object for built-in types
Eliminate fsbv:object for built-in types; started conversion of
special function result (sf-result*) structures but this requires a
modification to CFFI. Simulated annealing structure called by value.
Use cffi-fsbv's sizet definition; this imposes a dependence on that
system. No compilation attempted yet.
Liam M. Healy [Mon, 24 Oct 2011 00:48:20 +0000]
Eliminate defenumeration
Liam M. Healy [Mon, 24 Oct 2011 00:33:42 +0000]
Merge branch 'experimental' into cffi-fsbv
Conflicts:
init/funcallable.lisp
linear-algebra/blas2.lisp
linear-algebra/blas3.lisp
solve-minimize-fit/linear-least-squares.lisp
special-functions/coulomb.lisp
Liam M. Healy [Tue, 18 Oct 2011 03:11:33 +0000]
Library path for Darwin + CCL per Yuan MEI
Liam M. Healy [Wed, 28 Sep 2011 17:36:47 +0000]
Improve wording and grammar in docstrings
Liam M. Healy [Mon, 12 Sep 2011 22:03:52 +0000]
Use generic functions for histogram
Define methods for existing generic functions for histogram:
grid:dimensions (replacing #'bins), set-zero (replacing #'reset).
Also, add a comment that the functions for 2D histogram have not been
defined as a "to be done" reminder.
Liam M. Healy [Fri, 26 Aug 2011 14:37:40 +0000]
Move asdf-system-connections into :defsystem-depends-on
Liam M. Healy [Fri, 26 Aug 2011 02:37:34 +0000]
Restore connections in asd file
Apparently asdf-system-connections requires must be loaded _before_
the system is loaded; trying to make it a dependent system seems to
cause it to drop the main system directory name in the path. To make
this load on quicklisp, put an explicit load of
asdf-system-connections in the asd file.
Liam M. Healy [Tue, 23 Aug 2011 13:57:21 +0000]
Moved connected system definitions to a new file gsll-connections.lisp
soemraws [Sat, 20 Aug 2011 13:12:52 +0000]
Merge branch 'master' of git://repo.or.cz/gsll
Liam M. Healy [Sat, 20 Aug 2011 03:40:32 +0000]
Change shadowing of 'acceleration
CCL gets a conflict between the type antik:acceleration and the class
gsll:acceleration, so remove the symbol from the shadowing list of
GSLL and add it to antik::*antik-user-shadow-symbols*.
soemraws [Sat, 20 Aug 2011 00:38:57 +0000]
Added dimensionless physical constants from GSL
Liam M. Healy [Fri, 19 Aug 2011 14:16:52 +0000]
Remove reference to gsll-tests
Liam M. Healy [Fri, 19 Aug 2011 13:51:47 +0000]
Test ntuple in histogram by writing the data to a temporary file, then reading the file
Liam M. Healy [Fri, 19 Aug 2011 03:08:56 +0000]
Results of test matrix-product-hermitian as single-floats
Liam M. Healy [Wed, 17 Aug 2011 05:03:13 +0000]
Refer to Antik in documentation, delete reference to GSD
Liam M. Healy [Wed, 17 Aug 2011 04:52:46 +0000]
Fix test failures due to creation of wrong type of grid from #m macro
Liam M. Healy [Wed, 17 Aug 2011 03:03:28 +0000]
Merge branch 'antik'
Conflicts:
gsll.asd
ordinary-differential-equations/ode-system.lisp
random/dirichlet.lisp
Liam M. Healy [Wed, 10 Aug 2011 02:40:40 +0000]
Updated flysheet analytics code
Liam M. Healy [Fri, 29 Jul 2011 16:23:39 +0000]
Dirichlet PDF functions, vector length should be K, not K-1
For the Dirichlet PDF functions, the vector length should be K, the
first argument, not K-1. Reported by Nils Bertschinger.
Conflicts:
random/dirichlet.lisp
Liam M. Healy [Fri, 29 Jul 2011 16:15:58 +0000]
Dirichlet PDF functions, vector length should be K, not K-1
For the Dirichlet PDF functions, the vector length should be K, the
first argument, not K-1. Reported by Nils Bertschinger.
Liam M. Healy [Thu, 26 May 2011 21:56:22 +0000]
Use grid:aref instead of grid:gref
Liam M. Healy [Wed, 4 May 2011 14:26:13 +0000]
Clarify with-ode-integration used for scalar dependent variables
Liam M. Healy [Sat, 23 Apr 2011 22:12:47 +0000]
New arglist for grid:ensure-foreign-array; replace #'dot calls with #'grid:inner
Liam M. Healy [Fri, 8 Apr 2011 13:20:36 +0000]
Export 'next step for ODE
Liam M. Healy [Sun, 27 Feb 2011 01:45:11 +0000]
Add :description and :long-description to the gsll.asd file
Liam M. Healy [Tue, 22 Feb 2011 17:06:27 +0000]
Merge branch 'antik' of ssh://repo.or.cz/srv/git/gsll into antik
Conflicts:
histogram/ntuple-example.dat
removed
Liam M. Healy [Tue, 22 Feb 2011 17:05:28 +0000]
Remove ntuple-example.dat
Liam M. Healy [Sat, 19 Feb 2011 18:09:29 +0000]
Renamed grid:ensure-foreign-array from make-foreign-array-or-default
Liam M. Healy [Sat, 12 Feb 2011 22:27:55 +0000]
Define methods for #'grid:inner instead of #'dot
Liam M. Healy [Sun, 30 Jan 2011 16:39:49 +0000]
Remove invert-matrix to Antik, optional argument for LU-decomposition
The user-friendly ("higher") functions layered on GSLL are now in
Antik's math-high system, so the antik directory has been removed, and
linear-algebra.lisp is enhanced and in Antik. #'LU-invert now takes
'inverse as an optional argument, with the sensible default.
Liam M. Healy [Sun, 30 Jan 2011 15:18:21 +0000]
Define gsll-tests as a system connection
Define gsll-tests as a system connection, rather than a separate ASDF
system. This means that if lisp-unit and GSLL are loaded (in either
order), then the tests will automatically be available.
Liam M. Healy [Sun, 30 Jan 2011 02:59:24 +0000]
In actual-array-class, don't replace real class names
In actual-array-class, if 'category is not one of 'vector, 'matrix or
'both, then assume it's actually a class name and just return it
without trying to make a class name. This fixes a problem with
e.g. elt+ where the second argument specializer is a class name of the
scalar.
Liam M. Healy [Tue, 25 Jan 2011 17:12:36 +0000]
Remove dependence on foreign-array-tests, which is obsolete
Liam M. Healy [Thu, 13 Jan 2011 15:29:18 +0000]
Antik user package symbol shadowing and package use
Define symbols to be shadowed and packages to be used for convenience
and conflict resolution in the Antik user packages.
Liam M. Healy [Wed, 12 Jan 2011 14:15:28 +0000]
Single-float tests have explicit single-float numbers (finish)
Liam M. Healy [Wed, 12 Jan 2011 05:51:43 +0000]
Don't import grid:element-type, grid:foreign-array, and grid:matrix
Don't import grid:element-type, grid:foreign-array, and grid:matrix,
but rather refer to them with package prefix explicitly.
Superficially checked but requires more careful check.
Liam M. Healy [Wed, 12 Jan 2011 00:13:03 +0000]
Don't import 'grid:dimensions, use alexandria:make-keyword
Liam M. Healy [Mon, 10 Jan 2011 23:34:03 +0000]
Explicit package reference to grid:dimensions
Explicit package reference to grid:dimensions but there's an error
because if the symbol isn't imported many more tests fail.
Liam M. Healy [Mon, 10 Jan 2011 23:01:48 +0000]
Explicit package home in grid:foreign-pointer
Liam M. Healy [Mon, 10 Jan 2011 23:01:17 +0000]
Single-float tests have explicit single-float numbers (continued)
Liam M. Healy [Mon, 10 Jan 2011 15:35:47 +0000]
Shadow symbols from antik
With the export of unit and dimension names, there are conflicts with
gsl:psi, gsl:acceleration, gsl:knots. There is no actual conflict in
the usage because Antik physical quantities only needs to use these
symbols as symbols, there is no binding. So they are now
shadowing-imported into gsll.
Added explicit "d0" to numbers in polynomial to insure they are read
as double-floats.
Liam M. Healy [Fri, 7 Jan 2011 23:12:59 +0000]
Single-float tests have explicit single float numbers (start)
Change literal floats in tests from an unspecified precision to
explicit single floats, so that if *read-default-float-format* is
something other than 'single-float the tests will still pass. About
half the failures now fixed. Function #'cdot is returning a pointer
instead of a numerical value.
Liam M. Healy [Sat, 1 Jan 2011 21:17:41 +0000]
Import symbols from gsll into antik-user
Import symbols from the gsll package into the antik-user package,
managing conflicts.
Liam M. Healy [Sat, 1 Jan 2011 16:49:35 +0000]
Move invert-matrix to antik/linear-algebra.lisp
Liam M. Healy [Sat, 1 Jan 2011 16:31:40 +0000]
Methods for arithmetic functions on foreign-arrays
Depend on Antik instead of GSD. Add new methods for Antik functions
on foreign-arrays. This permits algebra-style computation on arrays,
e.g.,
(in-package :antik-user)
(setf grid:*default-grid-type* 'grid:foreign-array)
(expt (* (rotate-3d :z #_15_deg) (rotate-3d :z #_45_deg)) -1)
#m((0.5 0.
8660254037844386 0.0) (-0.
8660254037844385 0.
5000000000000001 -0.0)
(0.0 0.0 1.0))
(rotate-3d :z #_-60_deg)
#m((0.
5000000000000001 0.
8660254037844386 0.0)
(-0.
8660254037844386 0.
5000000000000001 0.0) (0.0 0.0 1.0))
Liam Healy [Sun, 19 Dec 2010 21:14:09 +0000]
Use fsbv:object instead of cffi:mem-aref
Liam Healy [Sun, 19 Dec 2010 20:58:25 +0000]
Eliminate scref and use of grid:dcref
Liam Healy [Sun, 19 Dec 2010 16:43:39 +0000]
Use new macros from FSBV
Use new macros define-equivalent-type, defconvert, and defenumeration
from FSBV.
Liam M. Healy [Thu, 16 Dec 2010 17:24:59 +0000]
Use fsbv:object to replace fsbv:convert-from-pointer
Liam Healy [Sun, 12 Dec 2010 05:27:34 +0000]
Define cl-convert-form in terms of fsbv:convert-from-pointer
Liam Healy [Sat, 4 Dec 2010 22:47:22 +0000]
Use fsbv:defconvert to define sf-result and sf-result-e10
Use fsbv:defconvert to define sf-result and sf-result-e10, and use
#'fsbv:object instead of #'val, #'err, #'e10, which have been
eliminated.
Liam Healy [Sat, 4 Dec 2010 15:03:36 +0000]
Introduce values-unless-singleton for return values in defmfun expansion
New function values-unless-singleton will wrap the forms in 'values if
there are more than one, otherwise it just returns the form. This is
used in the return value(s) in the defmfun expansion (by way of
body-expand) so that if the returned form returns multiple values,
they all come back from the defmfunned function (note that (values
(values ...)) removes only returns the first value.
Liam Healy [Sun, 28 Nov 2010 03:53:34 +0000]
Eliminate calls to grid:complex-to-cl
Eliminate calls to grid:complex-to-cl because it has been eliminated;
rely on fsbv:object to do the conversion. Use new function
#'fsbv:converter-defined-p.
Liam Healy [Sat, 27 Nov 2010 18:59:27 +0000]
Use fsbv:object for conversion if available; fix results of dot, cdot tests
In cl-convert-form, use fsbv:object on foreign structs if FSBV has
been loaded. Previously, the conversion of (complex single-float) was
wrong because it assumed numbers were double-float; the test results
for dot and cdot have been fixed.
Liam Healy [Sat, 27 Nov 2010 17:16:58 +0000]
Added comments for maref
Liam Healy [Thu, 25 Nov 2010 05:11:20 +0000]
New macros #'maref and #'(setf maref)
New macros #'maref and #'(setf maref) get and set elements of the GSL
array directly from the mpointer. These are not used by users, but in
the make-funcallable-form expansion of callbacks when scalarsp = T is
specified. This fixes an error in the argument order in the (setf
get-value) generic function that this replaces.
Liam M. Healy [Wed, 24 Nov 2010 20:27:01 +0000]
New macro access-value
New macro access-value to call gsl_*_set and gsl_*_get, as yet unused.
It will eventually replace get-value and (setf get-value) as used in
reference-foreign-element, but has the arguments in the right order
and because it's a macro, expands immediately to a
cffi:foreign-funcall instead of doing a generic function dispatch on
an eql spcializer of the class name.
Liam Healy [Sun, 14 Nov 2010 22:59:54 +0000]
Revert the histogram change
Revert the last histogram change because GSD has been changed so that
the gref compiler macro does not take effect except for foreign-array
types.
Liam Healy [Sun, 14 Nov 2010 00:33:44 +0000]
Definition of grid:gref* for histogram
Define method grid:gref* for histograms; it was a definition of
grid:gref, but with the compiler macro that turns that into
grid:gref*, we need it to be a definition of grid:gref* instead.
Removed the definition of +foreign-array-pointer+ and
+foreign-array-type+ to the grid package.
Liam M. Healy [Tue, 12 Oct 2010 21:48:37 +0000]
Instructions to use quicklisp and how to send patches
Liam Healy [Sun, 3 Oct 2010 22:15:44 +0000]
Add tests for rank-1-update
Liam Healy [Sun, 3 Oct 2010 21:40:54 +0000]
Merge branch 'master' of ssh://repo.or.cz/srv/git/gsll
James Wright [Wed, 29 Sep 2010 23:51:11 +0000]
Fixed EBADLEN error in `rank-1-update' and `conjugate-rank-1-update'
Sumant Oemrawsingh [Fri, 1 Oct 2010 20:13:01 +0000]
Added non-radix-2 tests for radix-2-sized vectors
Effectively, non-radix-2 algorithm tests on power-of-2 vector sizes are added.
The result is, that some tests fail, and the cause at this point is not yet
clear. The tests that fail, are all related to non-radix-2 algorithms working
on power-of-2 vector sizes.
Only tests that are missing, are the so-called bitreverse tests.
Sumant Oemrawsingh [Wed, 15 Sep 2010 21:18:57 +0000]
Changed off-stride checks to be only one assertion
Previously, each element that was checked off-stride would count as an
assertion.
Sumant Oemrawsingh [Wed, 15 Sep 2010 20:54:49 +0000]
Fixed offset for vector initialisation to better match GSL's tests.
Sumant Oemrawsingh [Wed, 15 Sep 2010 20:48:55 +0000]
Fixed FFT tests, both on- and off-stride.
For stride > 1 elements, the results are compared only on-stride, while
(for complex only) it is verified that off-stride elements are left alone.
Tests are still disabled due to time it takes to complete tests.
Tests are not yet complete wrt GSL's FFT tests.
Liam Healy [Sat, 4 Sep 2010 21:09:10 +0000]
Added function #'off-stride to make duplicate array with only the off-stride elements
Liam Healy [Sat, 4 Sep 2010 12:54:05 +0000]
Use vector/length for all versions of GSL in testing FFT
Since the function vector/length no longer relies on GSL for doing the
division, it works for all versions of GSL and whether or not FSBV is
loaded. This fixes a failure for those versions where
:initial-contents was getting an array and not a list.
Sumant Oemrawsingh [Thu, 2 Sep 2010 22:07:35 +0000]
Expected number of assertions fail consistently in fast-fourier-transform.
There was one case left where even out-of-stride elements were scaled as well.
Tests are still not enabled.
Sumant Oemrawsingh [Wed, 1 Sep 2010 21:11:21 +0000]
Fix for random errors in tests
The test functions did not properly initialize newly created vectors to
contain unique values. This caused comparison of elements that are out of the
stride to randomly succeed. The new test functions properly initialize the
vectors just like is done in the GSL test functions. The result is that
comparison of elements that are out of stride should fail consistently.
This automatically means that the comparison should only check elements which
are on stride, just like is done in GSL. This is not fixed yet, so the
fast-fourier-transform unit tests are still left disabled for now.
Liam Healy [Mon, 23 Aug 2010 02:18:03 +0000]
Fix FFT to actually do something, but comment out because of failures
The FFT tests were testing nothing of consequence, just that NIL was
equal to NIL, because it was looking at the result of (reset-urand)
and then three non-existent multiple values. This is now fixed, so
that fft-*-result-check actually binds the results of the
calculations. Also, single-float tests are referenced to
+sgl-epsilon+ and not +dbl-epsilon+. However, for
(all-fft-test-forms 9 3 (64 99)), many tests fail:
FAST-FOURIER-TRANSFORM: 235 assertions passed, 55 failed.
so the tests are commented out.
Liam Healy [Sat, 21 Aug 2010 21:32:45 +0000]
New size-vector-scalar replaces size-vector-real
New function size-vector-scalar replaces size-vector-real, and gives
the size as a double float if the array is real, and as the element
type if it is complex, in order to be acceptable to the elt* interface
to GSL's _scale routines.
Liam M. Healy [Thu, 19 Aug 2010 20:43:59 +0000]
Fix simulated annealing for new grid:copy
Liam Healy [Tue, 17 Aug 2010 03:11:18 +0000]
Improved finding of libraries for OS X, by Leo
Liam Healy [Sat, 14 Aug 2010 22:35:21 +0000]
Add FFT tests
Add FFT tests; these are a subset of the GSL tests. For some reason
the tests take a lot longer in GSLL, so instead of testing all sizes
1...99, we test only 1...9 and then 64 and 99 for stride 1 only.
Liam M. Healy [Fri, 13 Aug 2010 17:56:16 +0000]
Put :unix at the end of load-foreign-library list
Put :unix at the end of load-foreign-library list so that more
specific unices like :darwin get seen first.
Liam M. Healy [Fri, 13 Aug 2010 15:40:05 +0000]
Grovel file depends on init
Liam Healy [Fri, 13 Aug 2010 02:52:43 +0000]
Invert the imaginary part in create-complex-matrix
The imaginary part of the elements in create-complex-matrix should be
inverted to match GSL's create_complex_matrix. Now all LU tests
pass.
Liam M. Healy [Wed, 11 Aug 2010 14:05:34 +0000]
Workaround CLISP bug
CLISP does not recognize that an uninterned symbol in different
parts of a progn is the same;
http://sourceforge.net/tracker/?func=detail&aid=836838&group_id=1355&atid=101355
Since this bug was "long-standing" the better part of a decade ago, we
will assume it now has tenure and change the progn to let.
Liam Healy [Tue, 10 Aug 2010 02:27:04 +0000]
Add dependencies on mathematical
Liam Healy [Sun, 8 Aug 2010 01:50:45 +0000]
Replace 'pi with 'dpi, the double-float version
The standard declares that 'pi is an implementation-dependent
long-float, but GSL requires double float, so the constant 'dpi has
been defined as (coerce pi 'double-float). This should make CLISP at
the least happier.
Liam M. Healy [Thu, 22 Jul 2010 21:47:38 +0000]
Updated user information
Liam M. Healy [Wed, 21 Jul 2010 17:35:37 +0000]
Update status.text, last element of list
Update status.text mostly be removing obsolete information without
replacing it with current information. Use alexandria:lastcar for the
last element of list so I do not use private definitions (non-loaded
source).
Liam M. Healy [Tue, 20 Jul 2010 22:02:16 +0000]
Name conditions generic-failure-* differently
Name conditions generic-failure-1 and generic-failure-2 because CCL
doesn't like the duplicate differently. Withdraw the
conditionalization away of +nan+, +positive-infinity+, and
+negative-infinity+ for CCL because the latest version
"Version 1.6-dev-r13994M-trunk (LinuxX8664)"
can handle them.
Liam M. Healy [Mon, 19 Jul 2010 16:38:28 +0000]
CCL don't define +nan+, +positive-infinity+, +negative-infinity+
In CCL, +nan+, +positive-infinity+, +negative-infinity+ are not
defined because of a problem with CCL loading these definitions.
Liam Healy [Mon, 19 Jul 2010 02:54:19 +0000]
Merge branch 'convert'
Trivial conflicts:
data/array-structs.lisp
init/init.lisp