mirrors/ecl
1
Fork 0
mirror of https://gitlab.com/embeddable-common-lisp/ecl.git synced 2026-05-01 13:40:50 -07:00
Code Issues Projects Releases Packages Wiki Activity
4600 commits 18 branches 54 tags 44 MiB
Commit graph

91 commits

Author SHA1 Message Date
Juan Jose Garcia Ripoll
fb92bca134 Fixed typo in prepare_ratio_to_float() 2010-12-07 20:11:07 +01:00
Juan Jose Garcia Ripoll
a995a89c1d Fix detection of NaN/inf for platforms without fenv.h 2010-12-02 22:46:40 +01:00
Juan Jose Garcia Ripoll
3801f5dfe3 Added a new file, impl/math_fenv.h, grouping all definitions related to floating point exceptions and their handling. This is completed with impl/math_fenv_msvc.h with inline definitions of the C99 fenv.h functions for Microsoft's compiler 2010-11-09 00:44:12 +01:00
Juan Jose Garcia Ripoll
3b92fb53f0 Removed the possibility of building ECL without bignums 2010-11-06 16:42:39 +01:00
Juan Jose Garcia Ripoll
cf29412c64 Removed init_numbers(), replacing it with static constants in all_symbols.d 2010-11-06 00:06:06 +01:00
Juan Jose Garcia Ripoll
0270afbb05 Rounding routines ratio_to_double, ratio_to_long_double, were missing code for the situation when the argument is not a fixnum 2010-10-03 15:54:29 +02:00
Juan Jose Garcia Ripoll
655809781a Rounding of bignums in ecl_to_double() did not work (GMP truncates). 2010-09-27 00:35:02 +02:00
Juan Jose Garcia Ripoll
321356dadf cl_core is now mostly filled statically in main.d 2010-08-21 16:31:42 +02:00
Juan Jose Garcia Ripoll
e83bd3dfb4 Unboxed type SHORT-FLOAT removed from the core. 2010-08-17 19:47:14 +02:00
Juan Jose Garcia Ripoll
0e3b1c389f Replace some more uses of mpz_set_si/ui with _ecl_big_set_fixnum/index 2010-08-17 19:47:13 +02:00
Juan Jose Garcia Ripoll
14cf3f7029 Replace more uses of cl_cos/sin/... with ecl_cos/sin/.. 2010-06-20 16:57:21 +02:00
Juan Jose Garcia Ripoll
445d269d6e Use static constants for rational and complexes 2010-06-04 16:39:51 +02:00
Juan Jose Garcia Ripoll
fe56a0c5a9 In number.d we use ecl_def_ct* instead of allocating the floating point numbers at boot time. 2010-06-04 15:46:09 +02:00
Juan Jose Garcia Ripoll
01e227c402 Replaced many uses of type_of() with specialized macros (ECL_BIT_VECTOR_P, ECL_BASE_STRING_P, etc) 2010-02-28 01:09:13 +01:00
Juan Jose Garcia Ripoll
89b9a561ea Replaced ecl_fixnum_in_range() with explicit checks in the functions, which are cheaper and, thanks to ecl_unlikely, faster. 2010-02-27 16:23:21 +01:00
Juan Jose Garcia Ripoll
86c211a6a1 Optimize ecl-inl.h for small cons and use of ecl_unlikely. New macros ECL_CONSP, ECL_LISTP, ECL_ATOM, ECL_SYMBOLP 2010-02-27 09:47:05 +01:00
Juan Jose Garcia Ripoll
871104a0cc Replaced FEtype_error_real with FEwrong_type_*_arg 2010-02-25 17:37:03 +01:00
Juan Jose Garcia Ripoll
d72d2685bd Fixed typo. 2009-12-13 14:38:55 +01:00
Juan Jose Garcia Ripoll
e85ae8d20c Fixed typo in #ifdef that only affected builds with --enable-long-double 2009-12-13 13:42:56 +01:00
Juan Jose Garcia Ripoll
f96a987819 ecl_to_long_double appeared before ratio_to_long_double 2009-12-13 12:08:53 +01:00
Juan Jose Garcia Ripoll
19943c0a22 ECL now has its own rounding routine for converting rationals to floats. 2009-12-05 01:47:41 +01:00
Juan Jose Garcia Ripoll
484fb955ab Renamed all remaining internal functions that work with bignums. 2009-08-28 22:31:39 +02:00
Juan Jose Garcia Ripoll
44443c10b8 Prefixed the big_register* functions with _ecl_ 2009-08-28 20:06:40 +02:00
Juan Jose Garcia Ripoll
3d778f4974 Bignum registers can not use statically allocated memory if we use GMP's own allocation/deallocation routines. This implies additional cleanup when deallocating an environment, and allowing for larger registers to decrease the pressure on the garbage collector. 2009-08-28 13:05:36 +02:00
Juan Jose Garcia Ripoll
3919ccdadc Changed the routines that manipulate bignums so that they use bignum registers and free them when finished -- this should allow ECL work without changing GMP's memory functions. 2009-08-28 02:03:30 +02:00
Juan Jose Garcia Ripoll
3556aae864 Fixed typo in number.d (CL_FIXNUM_BITS -> FIXNUM_BITS). 2009-08-21 22:07:43 +02:00
Juan Jose Garcia Ripoll
2ca580b13c New routines to convert lisp objects to 'int', 'long' and 'bool'. 2009-07-15 15:19:08 +02:00
Juan Jose Garcia Ripoll
eb864ae55a On platforms where it is not possible to trap FPE, explicitely check the status of the exceptions around basic operations (sin, cos, exp, etc). 2009-06-14 21:06:20 +02:00
Juan Jose Garcia Ripoll
0b0e026664 Create constants in cl_core for variants of -0.0 and use them in compiled code. IMAGPART now produces signed zeros for negative real numbers. 2009-06-14 11:48:04 +02:00
Juan Jose Garcia Ripoll
cbffb985ff long_double_to_integer now has the expected accuracy. 2009-06-14 01:47:37 +02:00
Juan Jose Garcia Ripoll
849302a89b RATIONAL works now more accurately with long-floats 2009-06-13 22:54:50 +02:00
Juan Jose Garcia Ripoll
13763bae70 Implement sign comparisons using signbit(), which defaults to (x==y) where it is not available. Since floats may have extra unused bits (long double), we cannot implement eql using bit comparisons 2009-06-13 22:50:28 +02:00
Juan Jose Garcia Ripoll
5008fdd45f Big patch for introducing NaNs and infinities. 2009-06-04 15:55:13 +02:00
Juan Jose Garcia Ripoll
42fdda5e1c Whenever isfinite is not defined, use finite(). Additionally, in Solaris, load ieefp.h before using this macro. 2009-05-20 19:29:16 +02:00
Juan Jose Garcia Ripoll
1b558faa95 Fixed two typos in ecl_to_int32_t() 2009-04-02 15:22:41 +02:00
Juan Jose Garcia Ripoll
e282487f0d Added support for long long and unsigned long, and [u]intN_t in the FFI. 2009-03-02 23:56:39 +01:00
Juan Jose Garcia Ripoll
fb7ef0c3b4 No need to divide when there is no common denominator 2009-03-01 23:19:47 +01:00
Juan Jose Garcia Ripoll
157120712e The functions for ecl_make_[u]int64 are only needed for platforms with 32 bit words. 2009-03-01 21:28:24 +01:00
Juan Jose Garcia Ripoll
ba9540e220 New specialized array types for 16, 32 and 64 bits large integers 2009-03-01 20:36:08 +01:00
Juan Jose Garcia Ripoll
e16ff91913 Detect and provide conversion routines for C specialized integer types ([u]int{8,16,32,64}_t) 2009-03-01 15:41:36 +01:00
Juan Jose Garcia Ripoll
cbe49afc5b Unify the sources for windows and posix threads 2008-10-11 11:38:15 +02:00
Juan Jose Garcia Ripoll
5bda4b4f61 All allocation functions are now in the ECL library and they now have the ecl_* prefix. Old names are kept in macro form. 2008-10-10 22:36:15 +02:00
Juan Jose Garcia Ripoll
3f1419fe35 New flag --with-signed-zero (Juanjo). RUN-PROGRAM now returns exit code of program (Josh Elsasser). 2008-10-05 00:04:15 +02:00
jjgarcia
140f0611a7 The reader and the printer understand negative zeros. 2008-09-28 20:14:00 +00:00
jjgarcia
860d38f2bb Upgrading of imaginary part in a complex number did not work if the real part had type long-float 2008-08-02 18:19:39 +00:00
jjgarcia
ebefa6ed13 Increased precision of PI 2008-08-02 16:24:34 +00:00
jjgarcia
37923e4ac5 Fixed ATAN for builds without long floats. PI has to be of type LONG-FLOAT 2008-08-02 15:25:58 +00:00
jjgarcia
0c86faece3 Use isfinite/isnan as macros, like C99 prescribes. Needed for long double support in OS X 2008-07-25 14:59:31 +00:00
jjgarcia
90e936d85c There is not finitel(), so just use coercion. 2008-06-28 08:22:14 +00:00
jgarcia
5ef5d32e74 Enforce indent and mode conditions on files 2008-01-26 10:30:41 +00:00