| 1 |
Revision history for Perl module Math::MPFR |
| 2 |
|
| 3 |
4.47 |
| 4 |
- Accommodate Math::MPC overloading of Math::MPFR objects |
| 5 |
- Wrap C's strlen (_strlen), API SvCUR (_SvCUR), API SvCUR_set (_Sv_CUR_set), |
| 6 |
and API SvLEN (_SvLEN) functions. (These wrappers are documented in the POD, |
| 7 |
but are not exported.) |
| 8 |
- Fiddle about with the internals of Rmpfr_fpif_export_mem. |
| 9 |
- Add Rmpfr_cmp_str() which compares the value of a Math::MPFR object with the |
| 10 |
value of a string at infinite precision. |
| 11 |
|
| 12 |
4.46 |
| 13 |
- Add Rmpfr_fpif_import_mem() and Rmpfr_fpif_export_mem(), new in 4.3.0-dev. |
| 14 |
- Allow atonv() to receive octal arguments - via new function oct2bin. |
| 15 |
- Add mpfrtoa_subn. (Tests added to mpfrtoa.t.) |
| 16 |
- In Makefile.PL, exclude "-lquadmath" from the "LIBS" argument passed to |
| 17 |
WriteMakefile() if ($Config{osname} =~/linux/i && $Config{nvtype} ne "__float128" |
| 18 |
&& !$config{gnulibc_version}). |
| 19 |
We do this in an attempt to allow Math::MPFR to build on linux systems where |
| 20 |
libc is musl, not glibc. For an example of the problem see: |
| 21 |
https://www.cpantesters.org/cpan/report/65ddb9fa-a453-11f0-9829-bde118873113 |
| 22 |
- In Makefile.PL make allowance for some cygwin build options. See (eg): |
| 23 |
https://www.cpantesters.org/cpan/report/41ed0a6e-6f74-1014-a0b5-c5df354b4780 |
| 24 |
|
| 25 |
4.45 |
| 26 |
- Fix typo in MPFR.xs - s/WE_HAVE_BENDIAN/MPFR_HAVE_BENDIAN/g |
| 27 |
See https://github.com/sisyphus/math-mpfr/issues/11. Thanks @gregoa. |
| 28 |
|
| 29 |
4.44 |
| 30 |
- Add subnormalize_float32, subnormalize_float16 and subnormalize_bfloat16. |
| 31 |
- Add Rmpfr_get_BFLOAT16/Rmpfr_set_BFLOAT16, Rmpfr_get_FLOAT16/Rmpfr_set_FLOAT16 |
| 32 |
and Rmpfr_get_FLT/Rmpfr_set_FLT functions. |
| 33 |
- Ensure that overloaded <=> operations return one of 1, -1, 0, and undef. |
| 34 |
Required some amendments to MPFR.xs, and some rewriting of parts of some |
| 35 |
test scripts. |
| 36 |
Issue raised by https://rt.cpan.org/Ticket/Display.html?id=168076 . |
| 37 |
- Add _PREC_MIN() to Prec/Prec.xs. This allows MPFR.pm to set the constant |
| 38 |
MPFR_PREC_MIN to the appropriate value. |
| 39 |
|
| 40 |
4.43 |
| 41 |
- More changes to t/_Float16.t and t/bfloat16.t needed to address what 4.42 |
| 42 |
was intended to fix. |
| 43 |
|
| 44 |
4.42 |
| 45 |
- Fix bug in t/_Float16.t and t/bfloat16.t. Fixes (eg): |
| 46 |
https://www.cpantesters.org/cpan/report/0e7804d2-5c3b-11f0-943b-9bf76d8775ea |
| 47 |
|
| 48 |
4.41 |
| 49 |
- Fix bug in t/new.t. Fixes (eg): |
| 50 |
https://www.cpantesters.org/cpan/report/2138de96-805b-1014-9104-da8c93be5a9b |
| 51 |
- Add Rmpfr_set_bfloat16, Rmpfr_get_bfloat16 and Rmpfr_buildopt_bfloat16_p. |
| 52 |
- Add unpack_float32, unpack_float16 and unpack_bfloat16. |
| 53 |
|
| 54 |
4.39 |
| 55 |
- Rewrite parts of "OPERATOR OVERLOADING" and "FORMATTED OUTPUT" sections in |
| 56 |
MPFR.pod. |
| 57 |
- Make correction and expansion in decimalize() documentation in MPFR.pod |
| 58 |
- Add new test file t/overload_op_eq.t |
| 59 |
|
| 60 |
4.38 |
| 61 |
- Enable the Rmpf_*_uj and Rmpfr_*_sj functions (except for Rmpfr_get_uj and |
| 62 |
Rmpfr_get_sj) for all perl configurations. Also enable Rmpfr_pown. |
| 63 |
- Relocate math_mpfr_include.h conditional defining of "mpfr_exp_t" to below |
| 64 |
the #inclusion of mpfr.h. |
| 65 |
|
| 66 |
4.37 |
| 67 |
- Fix bug in t/cross_class_mpq.t that caused failures like: |
| 68 |
http://www.cpantesters.org/cpan/report/cfa1541e-f6d4-11ef-a0db-2a526e8775ea |
| 69 |
|
| 70 |
4.36 |
| 71 |
- Fix bug in t/atonv.t - fixes https://rt.cpan.org/Ticket/Display.html?id=160048. |
| 72 |
Thanks Slaven Rezic. |
| 73 |
- Add new function q_fmod_fr() |
| 74 |
- Simplify "OPERATOR OVERLOADING" in MPFR.pod |
| 75 |
|
| 76 |
4.35 |
| 77 |
- Add Rmpfr_get_float16, Rmpfr_set_float16 and Rmpfr_buildopt_float16_p. |
| 78 |
All 3 are new in mpfr-4.3.0. |
| 79 |
- Overload '%' operator to perform fmod(). |
| 80 |
- Accommodate overloading of Math::GMP objects. |
| 81 |
- Alter overloading of left shift and right shift operations to be compatible |
| 82 |
with perl and Math::BigFloat |
| 83 |
- Fix https://rt.cpan.org/Public/Bug/Display.html?id=158699 (introduced in 4.25). |
| 84 |
Thanks SREZIC. |
| 85 |
|
| 86 |
4.34 |
| 87 |
- Fix http://www.cpantesters.org/cpan/report/b5e0d3db-6deb-1014-9c2b-9eec45d75318 |
| 88 |
Thanks @twata_1. |
| 89 |
- Sanitize Rmpfr_*printf handling of "%a" formatting. |
| 90 |
- Add new utility function nv2mpfr(). |
| 91 |
|
| 92 |
4.33 |
| 93 |
- Add new (in MPFR-4.3.0) function Rmpfr_trigamma and new test script t/trigamma.t |
| 94 |
- Add dragon_test() as a replacement for the (now) deprecated nvtoa_test(). |
| 95 |
|
| 96 |
4.32 |
| 97 |
- Workaround a bug in some Win32 builds. |
| 98 |
See https://github.com/StrawberryPerl/Perl-Dist-Strawberry/issues/226 |
| 99 |
|
| 100 |
4.31 |
| 101 |
- Reduce the testing done in t/LongDouble.t to simply test the correctness of |
| 102 |
Rmpfr_set_LD() and Rmpfr_get_LD(). |
| 103 |
- Add log_2(),log_10(), sind(), cosd(), tand() and tangent() functions. |
| 104 |
- Optionally allow (via setting of $Math::MPFR::PERL_INFNAN) infs and nans to be |
| 105 |
stringified by nv2a() to the same as perl provides. Previously they have been |
| 106 |
stringified by nv2a() strictly as "Inf", "-Inf" and "Nan" - which could differ |
| 107 |
from perl's presentation. See "DISPLAYING INFS AND NANS" in the documentation. |
| 108 |
|
| 109 |
4.29 |
| 110 |
- Add new (in MPFR-4.3.0) function Rmpfr_compound and new test script t/compound.t. |
| 111 |
- In math_mpfr_include.h, re-specify MATH_MPFR_NV_MAX in hex, as 32-bit Windows |
| 112 |
perl-5.39.7 (nvtype of 'double') was apparently unable to evaluate the decimal |
| 113 |
form correctly. |
| 114 |
- Silence pointless compilation warnings generated by MPFR.xs. |
| 115 |
|
| 116 |
4.28 |
| 117 |
- Correct the testing for the presence of the Windows subnormal bug in t/nvtoa2.t. |
| 118 |
- Small tweaks to Rmpfr_cmp_IV and Rmpfr_set_IV. |
| 119 |
- Avoid monitoring PL_markstack_ptr in MPFR.xs. |
| 120 |
- In Rmpfr_deref2, replace 1st arg of '0' in mpfr_get_str() call with 'NULL'. |
| 121 |
|
| 122 |
4.27 |
| 123 |
- Remove stray '}' from Rmpfr_add_z |
| 124 |
- Simplify overloading of '++' and '--' |
| 125 |
|
| 126 |
4.26 |
| 127 |
- Alter t/out_str.t and t/printf.t to accommodate sisyphus' MSVC-built perls. |
| 128 |
|
| 129 |
4.25 |
| 130 |
- work on fixing bugs in nvtoa() for DoubleDoubles. (Use _mpfrtoa instead of _nvtoa.) |
| 131 |
- Fix typos in pod. Thanks Vincent Lefevre - https://github.com/sisyphus/math-mpfr/pull/6. |
| 132 |
- Alter the Makefile.PL (and some test scripts) to accommodate sisyphus' MSVC-built perls. |
| 133 |
|
| 134 |
4.24 |
| 135 |
- add nvtoa_test() to facilitate validation of results returned by nvtoa() and mpfrtoa(). |
| 136 |
- mpfrtoa() now takes a second optional arg that specifies the minimum normal precision. |
| 137 |
If the precision of the Math::MPFR object is less than the specified value, then the |
| 138 |
the object is deemed to hold a subnormalized value. |
| 139 |
|
| 140 |
4.23 |
| 141 |
- Fix bug in overload_spaceship() handling of NaN. If the first arg was non-nan |
| 142 |
and thesecond arg was a Math::MPFR nan object, then the spaceship operator |
| 143 |
returned 0 instead of undef. (Added a test for this to t/overload.t.) |
| 144 |
- Fix overload_dec() to handle DEAL_WITH_NANFLAG_BUG_OVERLOADED macro. |
| 145 |
See https://github.com/sisyphus/math-mpfr/pull/5 |
| 146 |
|
| 147 |
4.22 |
| 148 |
- Add overloading of '<<' and '>>'. (Left shifts and right shifts don't make a lot of |
| 149 |
sense in floating-point environments. These overloads simply call mul_2ui or mul_2si |
| 150 |
or (respectively) div_2ui or div_2si, thereby altering the value of the exponent, |
| 151 |
but having no effect on the mantissa. |
| 152 |
- Work around perl bug in https://github.com/Perl/perl5/issues/19550. |
| 153 |
(Affected only perl-5.35.10.) |
| 154 |
|
| 155 |
4.21 |
| 156 |
- Fix bug re inconsistent settings of POK and NOK flag, which manifests itself as (eg): |
| 157 |
http://www.cpantesters.org/cpan/report/dc17e330-900b-11ec-bfc9-d1f1448276d4 |
| 158 |
It's a bug that rarely bites, and applies mainly (but not only) to perl-5.18.x and |
| 159 |
earlier. |
| 160 |
|
| 161 |
4.19 |
| 162 |
- Fix https://github.com/sisyphus/math-decimal64/pull/1 (also aplies to Math::MPFR) |
| 163 |
Thanks to @hiratara |
| 164 |
- Deal with deprecations of mpfr_root() and mpfr_grandom(). These 2 functions |
| 165 |
will now croak() if Math::MPFR has been built against mpfr-4.0.0 or later. |
| 166 |
- Replace four instances of mpz_cmp_ui(z, 0) with less expensive mpz_sgn(z) |
| 167 |
- Fix doubletoa() handling of negative zero. (Was hanging.) |
| 168 |
- Remove references to _MSC_VER |
| 169 |
- Replace sv_2mortal(newSVpv()) with more efficient newSVpvn_flags() where |
| 170 |
applicable. This is done via the MORTALIZED_PV() macro in math_mpfr_include.h |
| 171 |
|
| 172 |
4.18 |
| 173 |
- Allow anytoa() to (optionally) take just one (Math::MPFR object) arg, instead of |
| 174 |
the two args originally mandated. |
| 175 |
- Simplify Makefile.PL. |
| 176 |
- Replace XS defines NV_IS_FLOAT128 and NV_IS_LONG_DOUBLE with |
| 177 |
USE_QUADMATH and USE_LONG_DOUBLE (which are already defined by perl headers). |
| 178 |
- Replace XS defines NV_IS_DOUBLE and NV_IS_53_BIT with the condition NVSIZE == 8 |
| 179 |
(which is already defined in the perl headers). |
| 180 |
- Add Rmpfr_set_IV. |
| 181 |
- Utilize Rmpfr_cmp_IV and Rmpfr_cmp_NV in the overloaded comparison subs. |
| 182 |
- Remove the *_UV functions and adapt the corresponding *_IV functions to handle the case |
| 183 |
that SvUOK() is true. (This is how it should always have been.) |
| 184 |
- Add Rmpfr_init_set_IV, Rmpfr_init_set_NV and Rmpfr_init_set_float128. |
| 185 |
- Add Rmpfr_cmp_float128. |
| 186 |
- Rmpfr_cmp_IV, Rmpfr_set_IV, Rmpfr_init_set_IV, Rmpfr_cmp_NV, Rmpfr_set_NV |
| 187 |
and Rmpfr_init_set_NV accept only scalars that are IOK or (respectively) NOK as their |
| 188 |
2nd argument. |
| 189 |
- Add POK_flag(), IOK_flag(), NOK_flag (@EXPORT_OK only) |
| 190 |
|
| 191 |
4.17 |
| 192 |
- In Rmpfr_set_default_prec(), Rmpfr_set_prec() and Rmpfr_set_prec_raw(), check |
| 193 |
that the argument is not less than 2 if mpfr version is less than 4.0.0. |
| 194 |
(This change was actually applied to Rmpfr_set_default_prec in Math-MPFR-4.16.) |
| 195 |
- Remove all occurrences of "scalar reverse" in demos\doubledouble.p and |
| 196 |
t\subnormal_doubles.t. |
| 197 |
- Add q_add_fr, q_sub_fr, q_mul_fr, q_div_fr and fr_cmp_q_rounded functions. |
| 198 |
- Add Rmpfr_cmp_IV, Rmpfr_cmp_UV, Rmpfr_cmp_sj and Rmpfr_cmp_uj (though the MPFR |
| 199 |
library currently contains neither mpfr_cmp_sj nor mpfr_cmp_uj). |
| 200 |
- Fix Rmpfr_get_str_ndigits and Rmpfr_get_str_ndigits_alt so that they don't change |
| 201 |
the inexflag from unset to set. (This was fixed for mpfr_get_str_ndigits in |
| 202 |
mpfr versions > 4.1.0.) |
| 203 |
- Add new function mpfrtoa. |
| 204 |
- Fix typo in Rmpfr_get_str_ndigits code: |
| 205 |
https://github.com/sisyphus/math-mpfr/commit/1f47b34b4d6b289bef9a40994f95ff57da3c6952#commitcomment-50313325 |
| 206 |
Thanks @trizen. |
| 207 |
- Add new functions Rmpfr_acosu, Rmpfr_acospi, Rmpfr_asinu, Rmpfr_asinpi, |
| 208 |
Rmpfr_atanu, Rmpfr_atanpi, Rmpfr_atan2u, Rmpfr_atan2pi, Rmpfr_powr, Rmpfr_pown, |
| 209 |
Rmpfr_pow_uj, Rmpfr_pow_sj, Rmpfr_log2p1, Rmpfr_log10p1, Rmpfr_compound_si, |
| 210 |
Rmpfr_exp2m1 and Rmpfr_exp10m1. |
| 211 |
- Add Rmpfr_pow_IV and Rmpfr_pow_UV |
| 212 |
- Alter decimalize() such that, if a second argument is provided, the function returns |
| 213 |
the projected number of significant digits. This does not break backward compatibility. |
| 214 |
(See the POD for elaboration.) |
| 215 |
- Add anytoa() function. |
| 216 |
- Use POSIX in demos/doubledouble.p to fix a (rarely encountered) bug. |
| 217 |
- Refactor IV/UV/NV/PV handling inside overload subs. |
| 218 |
|
| 219 |
4.16 |
| 220 |
- Workaround mpfr library assertion failure in decimalize() could occur when |
| 221 |
precision is 1 && mpfr library version is less than 4.0.2. |
| 222 |
|
| 223 |
4.15 |
| 224 |
- Make some changes in test files relating to bareword filehandles and indirect |
| 225 |
object syntax. |
| 226 |
- Enhance the Makefile.PL's capability of detecting that mpfr/gmp libraries are |
| 227 |
not going to be found when needed. |
| 228 |
- Add decimalize() and check_exact_decimal(). |
| 229 |
- For mpfr-4.2.0 and later only: add Rmpfr_sinu, Rmpfr_cosu, Rmpfr_tanu, |
| 230 |
Rmpfr_sinpi, Rmpfr_cospi, Rmpfr_tanpi. |
| 231 |
- Add Rmpfr_fmod_ui - new in mpfr-4.2.0, but also implemented here in Math::MPFR |
| 232 |
when built against earlier versions of the mpfr library. |
| 233 |
- Move code out of MPFR.xs and into (newly created) grisu3.c. |
| 234 |
|
| 235 |
|
| 236 |
4.14 |
| 237 |
- Add numtoa() and atonum(). |
| 238 |
- #include <inttypes.h> at beginning of grisu3.h to address: |
| 239 |
https://www.cpantesters.org/cpan/report/0d8a8dfd-6bf4-1014-a7c0-8323ad4a5811 |
| 240 |
- Fix https://www.cpantesters.org/cpan/report/3622b742-4202-11ea-8c53-8f701f24ea8f |
| 241 |
by amending Makefile.PL |
| 242 |
- Add some debug statements throughout nvtoa() in MPFR.xs - relies on NVTOA_DEBUG=1 |
| 243 |
being provided as a Makefile.PL command line argument . |
| 244 |
- Correct test 5 of t/numtoa.t for DoubleDouble nvtype. |
| 245 |
- Add wrapping of mpfr_total_order_p (previously missed) which is new in 4.1.0. |
| 246 |
- Add wrapping of mpfr_cmpabs_ui (previously missed) which is new in 4.1.0. |
| 247 |
- In atonv() pre-processing replace " & " with " && " as the former (which was a |
| 248 |
mistake, anyway) is not guaranteed to yield the correct result. |
| 249 |
|
| 250 |
4.13 |
| 251 |
- Fix bug in t/rndna.t and t/rndna2.t (exposed by: |
| 252 |
http://www.cpantesters.org/cpan/report/90030d18-8f4f-11e9-a14c-c69b77725194 ) |
| 253 |
- Change stringification of 'inf' & 'nan' from '@Inf@' & '@NaN@'to 'Inf' & 'NaN, |
| 254 |
except for Rmpfr_deref2(), which will still retain the actual inf or nan |
| 255 |
string returned by mpfr - ie '@Inf@' & '@NaN@'. |
| 256 |
- Remove definition of MAXIMUM_ALLOWABLE_BASE as it is always 62 now that we |
| 257 |
no longer support versions of MPFR older than 3.0.0. |
| 258 |
- Accommodate extended range of allowable bases for mpfr_out_str with MPFR-4.1.0 |
| 259 |
and later. |
| 260 |
- Cleanup of some code called by nvtoa(). |
| 261 |
- Fix bug in _get_exp_and_bits() for DoubleDoubles. Integer values > 2 ** 105 |
| 262 |
were being misrepresented by nvtoa(). |
| 263 |
- Improve the private functions _d_bytes, _ld_bytes, _dd_bytes, _f128_bytes |
| 264 |
- Replace the private functions _d_bytes_fr, _ld_bytes_fr, _dd_bytes_fr, |
| 265 |
_f128_bytes_fr with the single function _bytes_fr. |
| 266 |
- Remove mpfr_min_inter_base() and mpfr_max_orig_base() as both are of no value. |
| 267 |
|
| 268 |
4.12 |
| 269 |
- Fix bug in nvtoa() that had values with negative single digit significands && |
| 270 |
exponents less than -4 appearing as "-x.e-EXP" instead of "-xe-EXP" |
| 271 |
- Add doubletoa() function. (Uses grisu3 algorithm.) |
| 272 |
|
| 273 |
4.11 |
| 274 |
- Optimizations to nvtoa() |
| 275 |
|
| 276 |
4.09 |
| 277 |
- Hopefully fix bugs in t/nvtoa.t reported by (eg): |
| 278 |
http://www.cpantesters.org/cpan/report/844b433e-2463-11e9-b3bd-8a932d9bfb17 |
| 279 |
and |
| 280 |
http://www.cpantesters.org/cpan/report/608800de-245b-11e9-b3bd-8a932d9bfb17 |
| 281 |
- Add t/nvtoa2.t for additional testing of nvtoa function\ |
| 282 |
- Fix nvtoa() for Double-Double long doubles. (Previous to fix, strings being |
| 283 |
returned by nvtoa() for this nvtype were not always the minimum required |
| 284 |
length.) |
| 285 |
|
| 286 |
4.08 |
| 287 |
- In nvtoa(), Work around mpfr_get_str() bug that was not fixed until |
| 288 |
mpfr-4.0.2 |
| 289 |
|
| 290 |
4.07 |
| 291 |
- Fix bug in t/atonv.t. (Thanks Dana Jacobsen) |
| 292 |
- Add nvtoa() |
| 293 |
- Alter the way that atonv() is called. |
| 294 |
- Emit deprecation warnings in Rmpfr_grandom and Rmpfr_root if mpfr |
| 295 |
version >= 4.0 |
| 296 |
|
| 297 |
4.06 |
| 298 |
- Fix bug (https://bugs.debian.org/918673) in t/subnormal_doubles.t |
| 299 |
|
| 300 |
4.05 |
| 301 |
- Add atodouble() function |
| 302 |
- Fix atonv to correctly handle subnormal double-doubles |
| 303 |
- Add Rmpfr_dot() and Rmpfr_get_str_ndigits. (New in 4.1.0.) |
| 304 |
- Correct Rmpfr_sum to return signed int, not unsigned value. |
| 305 |
- Add Rmpfr_set_DECIMAL128 and Rmpfr_get_DECIMAL128. (New in 4.1.0.) |
| 306 |
|
| 307 |
4.04 |
| 308 |
- Support only mpfr-3.0.0 and later |
| 309 |
- Check that first arg given to atonv() is of correct precision. |
| 310 |
- In MPFR.xs - replace NULL FILE* arguments with stdin/stdout in |
| 311 |
mpfr_out_str/mpfr_inp_str. (Thanks Vincent Lefevre.) |
| 312 |
|
| 313 |
4.03 |
| 314 |
- Values less than DENORM_MIN were automatically denormalized to 0 ... but should have been |
| 315 |
denormalized to DENORM_MIN when greater than DENORM_MIN/2. |
| 316 |
- Attend to various other issues involving handling of subnormals. |
| 317 |
- Remove Math::Float128 and Math::Decimal64 from PREREQ_PM. |
| 318 |
- Add atonv(). |
| 319 |
|
| 320 |
4.02 |
| 321 |
- mpfr_subnormalize() in _d_bytes(), _ld_bytes() and _f128_bytes(). (These 3 functions are |
| 322 |
used by Math::NV.) |
| 323 |
4.01 |
| 324 |
- Automatically build Math::MPFR with __float128 support if mpfr_buildopt_float128_p() returns |
| 325 |
true && 'F128=0' has not been provided as an argument to Makefile.PL. |
| 326 |
- Automatically build Math::MPFR with _Decimal64 support if mpfr_buildopt_decimal_p() returns |
| 327 |
true && 'D64=0' has not been provided as an argument to Makefile.PL. |
| 328 |
- Make sure that _Float128 is a known type when MPFR_WANT_FLOAT128 is defined |
| 329 |
(because it appears in mpfr.h in mpfr-4.1.0). |
| 330 |
- Add '-lquadmath' to LIBS in Makefile.PL (as some Cygwin builds apparently want it). |
| 331 |
- Prototype those subs specified in 'use subs'. |
| 332 |
- Alter Prec.xs, Random.xs and V.xs to accommodate perl.h's defining (as of 5.28.7) of |
| 333 |
INTMAX_C and UINTMAX_C. |
| 334 |
(Replaced '#include <mpfr.h>' with '#include "../math_mpfr_include.h" |
| 335 |
|
| 336 |
4.0 |
| 337 |
- Add Rmpfr_free_cache2 and Rmpfr_free_pool. |
| 338 |
- Add MPFR_RNDF rounding mode |
| 339 |
- Fix bug in Rmpfr_cmp_NV and add the tests (test 8 in t/nan_cmp.t) that would have |
| 340 |
detected it. The bug affected only quadmath builds of perl, and only then if |
| 341 |
CAN_PASS_FLOAT128 was not defined. |
| 342 |
- Improve performance of Rmpfr_set_NV and Rmpfr_cmp_NV for quadmath builds of perl |
| 343 |
where CAN_PASS_FLOAT128 is not defined. |
| 344 |
- Add Rmpfr_rootn_ui (mpfr-4.0.0 and later only). |
| 345 |
- Add tests to Rmpfr_get_q.t to check that Rmpfr_get_q canonicalizes the rational. |
| 346 |
(Trizen pointed out that, with 4.0.0-RC1, mpfr_get_q did NOT canonicalize.) |
| 347 |
|
| 348 |
3.36 |
| 349 |
- Add Rmpfr_rec_root, provided by Vincent Lefevre: |
| 350 |
(See https://sympa.inria.fr/sympa/arc/mpfr/2016-12/msg00032.html) |
| 351 |
Re-arranged as an XSub, and some special case handling added. |
| 352 |
- Add Rmpfr_beta (mpfr-4.0.0 and later only). |
| 353 |
- Rename t/new_in_3.2.0.t to t/new_in_4.0.0.t |
| 354 |
- Add Rmpfr_q_div and Rmpfr_z_div and use them in overload_div to replace the existing |
| 355 |
code (that was rounding twice). |
| 356 |
- Add overloading of comparisons with mpq_t and mpz_t |
| 357 |
- use mpfr library of version of mpfr_get_q (in preference to homegrown rendition) if |
| 358 |
MPFR_VERSION_MAJOR >= 4. |
| 359 |
- Workaround mpfr_get_float128() failure to round subnormals reliably. (Workaround was removed |
| 360 |
soon after - when the mpfr_get_float128 problem was fixed in mpfr source.) |
| 361 |
|
| 362 |
3.35 |
| 363 |
- In Makefile.PL define -DNV_IS_DOUBLE if appropriate (for use in math_mpfr_include.h as part of |
| 364 |
fix for __float128 handling in nvtype=double DEBUGGING builds of perl). |
| 365 |
- Tighten the conditions under which _win32_infnanstring() is called. (Now called only if OS is |
| 366 |
MS Windows && $] < 5.022.) |
| 367 |
- Add Rmpfr_round_nearest_away() |
| 368 |
- Move POD from MPFR.pm to MPFR.pod |
| 369 |
|
| 370 |
3.34 |
| 371 |
- eval{} Rmpfr_z_sub in t/flags.t (test 23). |
| 372 |
- Remove -DUSE_LONG_DOUBLE (from Makefile.PL). |
| 373 |
- Add _is_infnanstring to MPFR.xs (for use with overloaded string operations). |
| 374 |
- Add Rmpfr_cmp_NV. |
| 375 |
|
| 376 |
3.33 |
| 377 |
- Document Rmpfr_grandom as deprecated as of mpfr-3.2.0. |
| 378 |
- Add the following functions which will be available when mpfr-4.0.0 (or higher) is installed: |
| 379 |
Rmpfr_fmodquo, Rmpfr_fpif_export, Rmpfr_fpif_import, Rmpfr_flags_clear, Rmpfr_flags_set, |
| 380 |
Rmpfr_flags_test, Rmpfr_flags_save, Rmpfr_flags_restore, Rmpfr_rint_roundeven, Rmpfr_roundeven, |
| 381 |
Rmpfr_nrandom, Rmpfr_erandom, Rmpfr_fmma, Rmpfr_fmms, Rmpfr_log_ui and Rmpfr_gamma_inc |
| 382 |
- Add support for IEEE 754 long doubles to bytes() function, and rewrite bytes() documentation. |
| 383 |
- Remove library functions (gmp_v and Rmpfr_get_version) from Math::MPFR::V, as we don't want any |
| 384 |
library functions in that module. |
| 385 |
- Handle the setting of the NaN flag with mpfr_add/sub_ui/si functions for versions of mpfr (3.1.4 |
| 386 |
and earlier only). The bug should be fixed in later versions of mpfr. |
| 387 |
- Replace SvNV/SvUV/SvIV with SvNVX/SvUVX/SvIVX where appropriate. (In typemap, too.) |
| 388 |
- Add Rmpfr_buildopt_float128_p and Rmpfr_buildopt_sharedcache_p |
| 389 |
- t/_1basic.t now ouptuts the values of the various (available) buildopt functions. |
| 390 |
|
| 391 |
3.32 |
| 392 |
- Corrections to fixes to t/DoubleDouble2.t and t/bytes.t. (Same problem existed with |
| 393 |
some addtional architectures, too.) |
| 394 |
Version 3.31 will be marked for removal to backpan. |
| 395 |
|
| 396 |
3.31 |
| 397 |
- Fixes to t/DoubleDouble2.t and t/bytes.t (which were failing on machines that had the |
| 398 |
"double-double" type, but only when perl's nvtype was double). |
| 399 |
|
| 400 |
3.30 |
| 401 |
- Correction to the negative zero bug workaround in Rmpfr_fits_uintmax_p (which |
| 402 |
had previously been returning false for (-0, MPFR_RNDD). |
| 403 |
- Remove Rmpfr_set_str_binary and Rmpfr_print_binary. (Obsolete in mpfr library.) |
| 404 |
- Fix t/LongDouble.t for change to Math::LongDouble output precision (ticket #111813) |
| 405 |
|
| 406 |
3.29 |
| 407 |
- Add _d_bytes XSub. |
| 408 |
- Simplify demos/doubledouble.p |
| 409 |
- Allow _d_bytes, _ld_bytes, and _f128_bytes to determine the base of |
| 410 |
the given argument. |
| 411 |
- Add _d_bytes_fr, _dd_bytes_fr, _ld_bytes_fr and _f128_bytes_fr |
| 412 |
- Add bytes() function. |
| 413 |
- Amend typedef of __float128 alignment (Windows only) |
| 414 |
- Add Rmpfr_get_q (even though MPFR does not implement mpfr_get_q) |
| 415 |
- define mp_bitcnt_t to unsigned long int for pre-gmp-5.0.0 |
| 416 |
|
| 417 |
3.28 |
| 418 |
- Fix mistake in test 5 in t/bytes.t. |
| 419 |
|
| 420 |
3.27 |
| 421 |
- Replace mpfr_set_default_prec() calls in _ld_bytes and _f128_bytes with |
| 422 |
appropriate mpfr_init2() calls. |
| 423 |
|
| 424 |
3.26 |
| 425 |
- Re-phrase misleading message in t/use64bitint.t. (Message implied |
| 426 |
that absence of 'long long int' implied that ivsize was 32-bit.) |
| 427 |
- Add Rmpfr_print_rnd_mode - previously missed. |
| 428 |
- Rewrite rounding value check as CHECK_ROUNDING_VALUE macro |
| 429 |
- Add _ld_bytes and _f128_bytes XSub (for use of Math::NV) |
| 430 |
|
| 431 |
3.25 |
| 432 |
- Catch the error of giving too large a "size" argument to mpfr_sum.t and add a |
| 433 |
check for this to t/sum.t |
| 434 |
- Amend the #definition of LNGAMMA_BUG to cater for the fact that mpfr-3.1.3 |
| 435 |
did *not* contain a fix for the lngamma bug. |
| 436 |
|
| 437 |
3.24 |
| 438 |
- Rename the Rgmp_randstate* functions to Rmpfr_randstate*, and move them into |
| 439 |
their own (Math::MPFR::Random) module. |
| 440 |
- Standardise numeric treatment of strings that contain non-numeric characters. |
| 441 |
Add Math::MPFR::NNW and Math::MPFR::nnumflag() & friends. |
| 442 |
- Correction to t/native_float128.t |
| 443 |
- Use NVtoD64 to assign NV to Math::Decimal64 object in D64_LD.t - as |
| 444 |
Math::Decimal64::new() is no longer capable of making that assignment. |
| 445 |
|
| 446 |
3.23 |
| 447 |
- Rename Rmpfr_get_decimal64, Rmpfr_set_decimal64, Rmpfr_get_float128 & Rmpfr_set_float128 |
| 448 |
to Rmpfr_get_DECIMAL64, Rmpfr_set_DECIMAL64, Rmpfr_get_FLOAT128 & Rmpfr_set_FLOAT128 |
| 449 |
- Provide Rmpfr_get_float128 & Rmpfr_set_float128 which pass the __float128 value as an NV |
| 450 |
- Ensure that when casting an IV/NV to an mpfr_t, the mpfr_t has enough precision to |
| 451 |
accurately represent the value. |
| 452 |
- No longer define USE_64_BIT_INT if ivtype is 'long'. (Thus had to tweak some test files.) |
| 453 |
- Rename USE_64_BIT_INT #define to MATH_MPFR_NEED_LONG_LONG_INT |
| 454 |
- Define IVSIZE_BITS (in XS file) if MATH_MPFR_NEED_LONG_LONG_INT is defined - for use in |
| 455 |
initialization of mpfr_t in some of the overload functions. |
| 456 |
- Add FLOAT128_DIG |
| 457 |
|
| 458 |
3.22 |
| 459 |
- Add some metadata to Makefile.PL. |
| 460 |
- Fix the LNGAMMA_BUG workaround in mpfr.c, and change test 69 in t/test1.t to |
| 461 |
test for correct behaviour of mpfr_lngamma(). |
| 462 |
- Add Rmpfr_set_NV - and use it in new(). |
| 463 |
- Remove Rmpfr_sprintf_ret and Rmpfr_snprintf_ret. |
| 464 |
- No longer include inttypes.h on the basis of USE_LONG_DOUBLE being defined. |
| 465 |
(Was harmless ... but didn't make much sense.) |
| 466 |
- Define PERL_NO_GET_CONTEXT. |
| 467 |
- Tweaks to Math::MPFR::gmp_v() and Math::MPFR::V::gmp_v() |
| 468 |
|
| 469 |
3.21 |
| 470 |
- Fix errors in t/decimal64_conv.t |
| 471 |
|
| 472 |
3.19 |
| 473 |
- Rewrite _wrap_mpfr_sprintf(), _wrap_mpfr_snprintf(), _wrap_mpfr_sprintf_rnd() |
| 474 |
and _wrap_mpfr_snprintf_rnd(). Their first arg has been changed from a char* |
| 475 |
to an SV*, and they take an additional argument that specifies the size of |
| 476 |
the buffer (1st arg) into which the resultant string is written. |
| 477 |
- Change Rmpfr_sprintf_ret() and Rmpfr_snprintf_ret(). The first arg (buffer) |
| 478 |
has been removed, and an additional argument that specifies the size of the |
| 479 |
buffer into which the resultant string is written needs to be provided. |
| 480 |
- Fix spelling error in doc. Thanks Salvatore Bonaccorso. (Ticket #89710) |
| 481 |
- Add Rmpfr_set_float128 and Rmpfr_get_float128. These provide access to the |
| 482 |
'__float128' data type (for compilers that support the type) via the |
| 483 |
Math::Float128 module. |
| 484 |
- Simplify querying of Decimal64 support in the mpfr library being used |
| 485 |
- Add MPFR_DBL_DIG, MPFR_LDBL_DIG, mpfr_max_orig_len, mpfr_min_inter_prec, |
| 486 |
mpfr_min_inter_base and mpfr_max_orig_base. |
| 487 |
- Have lngamma(-0) be +Inf instead of NaN. (Bug in mpfr 3.1.2 and earlier.) |
| 488 |
|
| 489 |
3.18 |
| 490 |
- Depending upon the rounding mode values, the mpfr library may incorrectly return |
| 491 |
false for mpfr_fits_u*_p(x) for -1.0 < x < 0.0 or for -0.5 < x <0.0. (This |
| 492 |
should be fixed in versions of mpfr later than 3.1.2.) |
| 493 |
Have the Rmpfr_fits_u*_p() functions return correct result irrespective of the |
| 494 |
version of mpfr being used - and have the Math:MPFR test suite test for this. |
| 495 |
- The 'P' type specifier (precision) wasn't working correctly on big-endian machines |
| 496 |
when mp_prec_t size is less than IV size. Add a prec_cast function to fix this. |
| 497 |
- Add an _mp_sizes function (not exportable) that returns the sizes of mpfr_exp_t, |
| 498 |
mpfr_prec_t and mpfr_rnd_t. |
| 499 |
- Add Math::MPFR::_ivsize (not exportable). |
| 500 |
- Add Math::MPFR::_nvsize (not exportable). |
| 501 |
- Correct the Rmpfr_*printf functions so that they croak with an appropriate error |
| 502 |
message if a rounding arg is supplied but the last arg is not a Math::MPFR object. |
| 503 |
|
| 504 |
3.17 |
| 505 |
- Apply patch to MPFR.xs from Daniel Kahn Gillmor that eliminates |
| 506 |
"null argument where non-null required" warnings. |
| 507 |
- Fix typo in croak() message in MPFR.xs (also caught by Daniel). |
| 508 |
|
| 509 |
3.16 |
| 510 |
- Add Rmpfr_set_LD() and Rmpfr_get_LD(). These provide access to the |
| 511 |
'long double' data type (for perls whose nvtype is double) via the |
| 512 |
Math::LongDouble module. |
| 513 |
|
| 514 |
3.15 |
| 515 |
- Add _Decimal64-mpfr_t conversion |
| 516 |
|
| 517 |
3.14 |
| 518 |
- Fix pod. (The ' ' symbol was causing pod.t failures.) Also took care of |
| 519 |
some other typos/oversights in the documentation. |
| 520 |
|
| 521 |
3.13 |
| 522 |
- Add Math::MPFR::V |
| 523 |
- Add pre/post increment/decrement operations (overload_add and overload_inc) |
| 524 |
- Allow (cross-class) overloading of basic math operations with Math::GMPz, |
| 525 |
Math::GMPq and Math::GMPf objects. |
| 526 |
|
| 527 |
3.12 |
| 528 |
- Remove 'not' from overloads. Apparently achieved nothing anyway, and 5.16 |
| 529 |
warns that it's invalid. |
| 530 |
|
| 531 |
3.11 |
| 532 |
- Add t/tls.t and t/tls_flags.t (to test the thread local storage capabilities, |
| 533 |
where applicable). |
| 534 |
3.10 |
| 535 |
- Add the following functions (conditional upon building against |
| 536 |
mpfr-3.1.0 or later): |
| 537 |
Rmpfr_set_divby0, Rmpfr_clear_divby0, Rmpfr_divby0_p, Rmpfr_buildopt_tune_case, |
| 538 |
Rmpfr_frexp, Rmpfr_grandom, Rmpfr_z_sub, Rmpfr_buildopt_gmpinternals_p. |
| 539 |
- Add Rgmp_randinit_mt. |
| 540 |
- In Rmpfr_get_d_2exp and Rmpfr_get_ld_2exp, replace sv_setuv (was incorrect) |
| 541 |
with sv_setiv. |
| 542 |
- Replace New/Newz with Newx/Newxz on perls that support the new symbols (Newx/Newxz). |
| 543 |
3.02 |
| 544 |
- Rmpfr_min_prec now available only when building against mpfr-3.0.0 or later. (The |
| 545 |
mpfr_min_prec function was not available until 3.0.0.) |
| 546 |
3.01 |
| 547 |
- Remove the (outdated) link to http://www.loria.fr/projets/mpfr/mpfr-current/mpfr.html |
| 548 |
in the "DESCRIPTION" section of the documentation. (Thanks Hugh Myers.) |
| 549 |
- In the documentation, elaborate on the way that Rmpfr_get-version() and |
| 550 |
MPFR_VERSION & friends determine their values. (Thanks Vincent Lefevre.) |
| 551 |
- Correct Rmpfr_min_prec(), and remove Rmpfr_max_prec(). |
| 552 |
3.0 |
| 553 |
- Corresponding to release of mpfr-3.0.0. (Still needs only 2.4.0, |
| 554 |
but 3.0.0 needed to take advantage of all features.) |
| 555 |
- Add MPFR_* rounding modes (including MPFR_RNDA for builds against |
| 556 |
mpfr-3.0.0) |
| 557 |
- Add Rmpfr_buildopt_tls_p() and Rmpfr_buildopt_decimal_p() (mpfr-3.0.0 only) |
| 558 |
- Add the following functions, specific to mpfr-3.0.0: Rmpfr_regular_p, |
| 559 |
Rmpfr_set_zero, Rmpfr_digamma, Rmpfr_ai, Rmpfr_set_flt, Rmpfr_get_flt, |
| 560 |
Rmpfr_urandom and Rmpfr_set_z_2exp. |
| 561 |
- Add Rmpfr_get_z_2exp (mpfr_get_z_exp is now renamed to mpfr_get_z_2exp). |
| 562 |
- Add gmp_randstate_t* to the Math::MPFR typemap. |
| 563 |
- Change handling of NaNs (wrt overloaded comparison operators). |
| 564 |
The overloaded spaceship operator now returns undef when NaNs are involved |
| 565 |
(more in keeping with perl behaviour), and the erange flag is now |
| 566 |
set whenever a comparison involving NaNs is performed (more in |
| 567 |
keeping with the mpfr library behaviour). This is irrespective |
| 568 |
of whether the NaN is a Math::MPFR object or a perl NV. |
| 569 |
The other comparison operators now also set the erange flag when |
| 570 |
a NaN is involved (for compatibility with the mpfr library). |
| 571 |
- Add Rmpfr_snprintf and Rmpfr_snprintf_ret. |
| 572 |
- Add Rmpfr_inits, Rmpfr_inits2 and Rmpfr_clears functions. |
| 573 |
- The overload_copy sub now preserves the precision of the copied |
| 574 |
object (instead of taking on default precision). |
| 575 |
2.03 |
| 576 |
- Needs version 2.4.0 (or later) of the mpfr library. |
| 577 |
- Add the following functions (new in mpfr-2.4.0): |
| 578 |
Rmpfr_add_d, Rmpfr_sub_d, Rmpfr_d_sub Rmpfr_mul_d, Rmpfr_div_d, |
| 579 |
Rmpfr_d_div, Rmpfr_rec_sqrt, Rmpfr_sinh_cosh, Rmpfr_li2, Rmpfr_zeta_ui, |
| 580 |
Rmpfr_modf, Rmpfr_fmod, Rmpfr_printf, Rmpfr_fprintf, Rmpfr_sprintf, |
| 581 |
Rmpfr_sprintf_ret |
| 582 |
- Fix overload_sqrt() so that it returns a NaN for negative operands |
| 583 |
instead of croaking, as was originally the case. |
| 584 |
- Remove (the need for) INLINE.h |
| 585 |
- Add RMPFR_PREC_MIN and RMPFR_PREC_MAX. |
| 586 |
- Add Rmpfr_sum, Rmpfr_get_d_2exp, Rmpfr_get_ld_2exp, Rmpfr_dim, |
| 587 |
Rmpfr_fits_ushort_p, Rmpfr_fits_sshort_p, Rmpfr_fits_uint_p, |
| 588 |
Rmpfr_fits_sint_p, Rmpfr_fits_uintmax_p, Rmpfr_fits_intmax_p, |
| 589 |
Rmpfr_fits_IV_p, Rmpfr_fits_UV_p Rmpfr_add_si Rmpfr_get_UV |
| 590 |
Rmpfr_get_IV Rmpfr_get_NV |
| 591 |
|
| 592 |
2.02 |
| 593 |
- Add TRmpfr_out_str/TRmpfr_inp_str (which wraps mpfr_out_str/mpfr_inp_str |
| 594 |
more Truly than Rmpfr_out_str/Rmpfr_inp_str). |
| 595 |
- Rmpfr_out_str now has the option to take a prefix string. |
| 596 |
- Remove the tests in the Makefile.PL that test for the presence of |
| 597 |
a suitable compiler, GMP library and MPFR library. |
| 598 |
|
| 599 |
2.01 |
| 600 |
- Needs version 2.3.0 (or later) of the mpfr library. |
| 601 |
- Add Rmpfr_j0, Rmpfr_j1, Rmpfr_jn, Rmpfr_y0, Rmpfr_y1, |
| 602 |
Rmpfr_yn, Rmpfr_lgamma, Rmpfr_remainder, Rmpfr_remquo, |
| 603 |
Rmpfr_fms, Rmpfr_signbit, Rmpfr_setsign, and |
| 604 |
Rmpfr_copysign(all new in 2.3.0). Add also Rmpfr_get_patches |
| 605 |
(documented for the first time in 2.3.0). |
| 606 |
- Add Rmpfr_init_set_ld. |
| 607 |
- Add Rmpfr_deref2 to the list of exportable functions, and |
| 608 |
document its usage in the POD section. |
| 609 |
- Rmpfr_deref2 now has the mpfr library allocate (and free) memory |
| 610 |
for the mantissa. |
| 611 |
- Add Rmpfr_hypot (previously missed). |
| 612 |
- Rmpfr_out_str now takes an optional fifth argument (a string that |
| 613 |
will be appended to the mpfr_out_str output). And the output |
| 614 |
buffer is now flushed each time Rmpfr_out_str is called. |
| 615 |
- Rewrite new() - to make it a little tidier. |
| 616 |
- Add Rmpfr_set_uj and Rmpfr_set_uj_2exp to @EXPORT_TAGS. (Previously |
| 617 |
overlooked.) |
| 618 |
- Add Rmpfr_integer_string() - mainly to provide a simple means of |
| 619 |
getting 'sj' and 'uj' values on a 64-bit perl where the MPFR |
| 620 |
library does not support mpfr_get_uj and mpfr_get_sj functions - |
| 621 |
which happens with libraries built with Microsoft Compilers. |
| 622 |
- Fix the rendering of Inf and NaN values. (Now matches the |
| 623 |
mpfr_out_str output for Inf and NaN.) |
| 624 |
- Change the format returned by Rmpfr_get_str. (Hence the output of |
| 625 |
Rmpfr_get_string and Math::MPFR::overload_string have changed |
| 626 |
from earlier versions of Math::MPFR.) |
| 627 |
- Fix some of the overloaded subroutines as regards their handling |
| 628 |
of NaN. |
| 629 |
|
| 630 |
1.11 |
| 631 |
- Fix bug in Rmpfr_get_str. (Sometimes printed out one less digit |
| 632 |
than it ought.) |
| 633 |
- Remove mpfr.info from distribution. Refer instead to: |
| 634 |
http://www.mpfr.org/mpfr-current/mpfr.html |
| 635 |
- Have 'make clean' remove comp.exe, check.exe and version.exe. |
| 636 |
- Add support for perls built with -Duselongdouble |
| 637 |
(Rmpfr_*_ld functions, except Rmpfr_init_set_ld() which is not |
| 638 |
available in unpatched 2.2.1). |
| 639 |
- Add Rmpfr_cmp_d, Rmpfr_exp10 and the Rmpfr_*_uj functions (which |
| 640 |
had somehow been earlier overlooked). |
| 641 |
- Fix handling of unsigned long longs on perls built with -Duse64bitint. |
| 642 |
- Allow assignment of mpz, mpq, and mpf types using new() |
| 643 |
|
| 644 |
1.10 |
| 645 |
- No longer assign Exporter and DynaLoader to @Math::MPFR::ISA. |
| 646 |
- Add support for perls built with -Duse64bitint. This involved the |
| 647 |
inclusion of the Rmpfr_set_sj, Rmpfr_set_sj_2exp and Rmpfr_get_sj |
| 648 |
functions - as well as some alterations to the overload subs. |
| 649 |
|
| 650 |
1.09 |
| 651 |
- Changes to the format returned by Rmpfr_get_str and overload_string. |
| 652 |
- Add a new() method/function for creation of mpfr_t objects. |
| 653 |
- Fix Rmpfr_out_str() and Rmpfr_inp_str() - were segfaulting with |
| 654 |
MS compilers. |
| 655 |
|
| 656 |
|
| 657 |
1.08 |
| 658 |
- As of this version, MPFR-2.2.x is required. |
| 659 |
- Add new functions made available with the release of MPFR-2.2.0: |
| 660 |
Rmpfr_set_overflow Rmpfr_set_underflow Rmpfr_set_inexflag |
| 661 |
Rmpfr_set_erangeflag Rmpfr_set_nanflag Rmpfr_erfc Rmpfr_atan2 |
| 662 |
Rmpfr_pow_z Rmpfr_subnormalize Rmpfr_const_catalan Rmpfr_sec |
| 663 |
Rmpfr_csc Rmpfr_cot Rmpfr_root Rmpfr_eint Rmpfr_get_f Rmpfr_sech |
| 664 |
Rmpfr_csch Rmpfr_coth Rmpfr_lngamma RMPFR_VERSION_NUM |
| 665 |
- Add constants (actually implemented as subroutines): |
| 666 |
MPFR_VERSION MPFR_VERSION_MAJOR MPFR_VERSION_MINOR |
| 667 |
MPFR_VERSION_PATCHLEVEL MPFR_VERSION_STRING |
| 668 |
- Remove Rmpfr_add_one_ulp and Rmpfr_sub_one_ulp (no longer supported |
| 669 |
by the MPFR library). |
| 670 |
- Redefine '__gmpfr_default_rounding_mode' in MPFR.xs as it's no |
| 671 |
longer in mpfr.h. (Been moved to mpfr-impl.h which I don't want to |
| 672 |
include.) |
| 673 |
- Rewrite of overload_atan2 - now that mpfr_atan2 is available in |
| 674 |
MPFR-2.2.0. (Also fixes a bug in overload_atan2). |
| 675 |
|
| 676 |
1.07 |
| 677 |
- New() was allocating more memory than necessary in many instances. |
| 678 |
Rewrote the XS code (in those instances) to fix that. |
| 679 |
- Add Rmpfr_strtofr |
| 680 |
|
| 681 |
|
| 682 |
1.06 |
| 683 |
- Modify the Makefile.PL checks added in version 1.05 to cater for |
| 684 |
those building with a Microsoft compiler |
| 685 |
- Alter Rmpfr_set_str and Rmpfr_init_set_str so that a base of 0 |
| 686 |
can be specified - in which case the base is determined from the |
| 687 |
input string |
| 688 |
- Added the following functions: |
| 689 |
Rmpfr_get_ui Rmpfr_get_si Rmpfr_fits_ulong_p Rmpfr_fits_slong_p |
| 690 |
- Fix a bug in 'Rmpfr_deref2()' - used by 'overload_print()' |
| 691 |
- Provide more complete testing with 4 new test files - namely |
| 692 |
test4.t, trig.t, set_str.t and exceptions.t. |
| 693 |
|
| 694 |
|
| 695 |
1.05 |
| 696 |
- Add some checks to Makefile.PL to ensure that anyone trying to |
| 697 |
build this module has an up-to-date MPFR library. This is mainly |
| 698 |
an attempt to stop cpan testers who don't have an up-to-date MPFR |
| 699 |
from testing the module and subsequently reporting "FAIL" or "NA". |
| 700 |
|
| 701 |
|
| 702 |
1.04 |
| 703 |
- Math::MPFR-1.04 to 1.07 require MPFR-2.1.x. |
| 704 |
- Added the following functions, not available with versions earlier |
| 705 |
than 2.1.0 : |
| 706 |
Rmpfr_set_ui_2exp Rmpfr_set_si_2exp Rmpfr_get_z Rmpfr_si_sub |
| 707 |
Rmpfr_sub_si Rmpfr_mul_si Rmpfr_si_div Rmpfr_div_si Rmpfr_sqr |
| 708 |
Rmpfr_cmp_z Rmpfr_cmp_q Rmpfr_cmp_f Rmpfr_zero_p Rmpfr_free_cache |
| 709 |
Rmpfr_get_version Rmpfr_get_emin_min Rmpfr_get_emin_max |
| 710 |
Rmpfr_get_emax_min Rmpfr_get_emax_max Rmpfr_clear_erangeflag |
| 711 |
Rmpfr_erangeflag_p Rmpfr_rint_round Rmpfr_rint_trunc |
| 712 |
Rmpfr_rint_ceil Rmpfr_rint_floor |
| 713 |
|
| 714 |
1.03 |
| 715 |
- Replace 'SvROK()' with 'sv_isobject()' in the overload functions. |
| 716 |
- Make 'get_refcnt()' and 'get_package_name()' non-exportable. |
| 717 |
|
| 718 |
1.02 |
| 719 |
- Make overloading work with strings. |
| 720 |
- Add 'Rmpfr_dump'. |
| 721 |
- Add 'Math::MPFR::gmp_v' (not exportable). |
| 722 |
|
| 723 |
1.01 |
| 724 |
- Eradicate segfault that was occurring when 'overload_add', |
| 725 |
'overload_sub', 'overload_mul', 'overload_div' and 'overload_pow' |
| 726 |
croaked due to invalid argument(s). |
| 727 |
|
| 728 |
0.07 |
| 729 |
- Replace 'Rmpfr_round_prec' with 'Rmpfr_prec_round' in keeping |
| 730 |
with changes to the MPFR library. |
| 731 |
- Remove 'Rmpfr_random' (deprecated). |
| 732 |
- Add 'Rmpfr_cbrt', 'Rmpfr_cmpabs', ', 'Rmpfr_greater_p', |
| 733 |
'Rmpfr_greaterequal_p', 'Rmpfr_less_p', 'Rmpfr_lessequal_p', |
| 734 |
'Rmpfr_lessgreater_p', 'Rmpfr_equal_p', 'Rmpfr_unordered_p', |
| 735 |
'Rmpfr_inp_str', 'Rmpfr_frac', 'Rmpfr_integer_p', |
| 736 |
'Rmpfr_nexttoward', 'Rmpfr_nextabove', 'Rmpfr_next_below', |
| 737 |
'Rmpfr_min', 'Rmpfr_max', 'Rmpfr_get_exp', 'Rmpfr_set_exp' |
| 738 |
'Rgmp_randinit_default', 'Rgmp_randinit_lc_2exp', |
| 739 |
'Rgmp_randinit_lc_2exp_size', 'Rgmp_randseed', 'Rgmp_randseed_ui', |
| 740 |
'Rgmp_randclear'. |
| 741 |
- Change module name from 'Math::GnuMPfr' to 'Math::MPFR'. |
| 742 |
|
| 743 |
0.06 |
| 744 |
- Add 'Rmpfr_gamma()', 'Rmpfr_zeta()', and 'Rmpfr_erf()'. |
| 745 |
- Add overloading for 'atan2', 'cos', 'sin', 'log', 'exp', and 'int'. |
| 746 |
|
| 747 |
0.05 |
| 748 |
- Requires mpfr-2.0.3 |
| 749 |
- Add 'Rmpfr_out_str()' |
| 750 |
- Add 'Rmpfr_get_default_rounding_mode()' |
| 751 |
- Rename 'Rmpfr_set_str_raw()' to 'Rmpfr_set_str_binary()' |
| 752 |
in keeping with changes to the MPFR library. |
| 753 |
- Alter 'Rmpfr_check_range()' in keeping with change to |
| 754 |
'mpfr_check_range()' in the MPFR library. |
| 755 |
- Introduce operator overloading. |
| 756 |
|