cpython

History

Stefan Krah 4d3e0a695a Improve Underflow handling in the correct-rounding loop. The case for Underflow to zero hasn't changed: _mpd_qexp() internally uses MIN_EMIN, so the result would also underflow to zero for all emin > MIN_EMIN. In case digits are left, the informal argument is as follows: Underflow can occur only once in the last multiplication of the power stage (in the Horner stage Underflow provably cannot occur, and if Underflow occurred twice in the power stage, the result would underflow to zero on the second occasion). Since there is no double rounding during Underflow, the effective work precision is now 1 <= result->digits < prec. It can be shown by a somewhat tedious argument that abs(result - ex) < ulp(result, result->digits). Therefore the correct rounding loop now uses ulp(result, result->digits) to generate the bounds for ex in case of Underflow.		2012-05-31 20:01:05 +02:00
..
libmpdec	Improve Underflow handling in the correct-rounding loop. The case for	2012-05-31 20:01:05 +02:00
tests	Raise InvalidOperation if exponents of zeros are clamped during exact	2012-04-05 15:21:58 +02:00
ISSUES.txt	Issue #7652 : Integrate the decimal floating point libmpdec library to speed	2012-03-21 18:25:23 +01:00
README.txt	Issue #7652 : Integrate the decimal floating point libmpdec library to speed	2012-03-21 18:25:23 +01:00
_decimal.c	Issue #14478 : Cache the hash of a Decimal in the C version.	2012-04-10 16:27:58 +02:00
docstrings.h	Issue #7652 : Integrate the decimal floating point libmpdec library to speed	2012-03-21 18:25:23 +01:00

README.txt


About
=====

_decimal.c is a wrapper for the libmpdec library. libmpdec is a fast C
library for correctly-rounded arbitrary precision decimal floating point
arithmetic. It is a complete implementation of Mike Cowlishaw/IBM's
General Decimal Arithmetic Specification.


Build process for the module
============================

As usual, the build process for _decimal.so is driven by setup.py in the top
level directory. setup.py autodetects the following build configurations:

   1) x64         - 64-bit Python, x86_64 processor (AMD, Intel)

   2) uint128     - 64-bit Python, compiler provides __uint128_t (gcc)

   3) ansi64      - 64-bit Python, ANSI C

   4) ppro        - 32-bit Python, x86 CPU, PentiumPro or later

   5) ansi32      - 32-bit Python, ANSI C

   6) ansi-legacy - 32-bit Python, compiler without uint64_t

   7) universal   - Mac OS only (multi-arch)


It is possible to override autodetection by exporting:

   PYTHON_DECIMAL_WITH_MACHINE=value, where value is one of the above options.


NOTE
====
 
decimal.so is not built from a static libmpdec.a since doing so led to
failures on AIX (user report) and Windows (mixing static and dynamic CRTs
causes locale problems and more).