Commit Graph

129 Commits

Author SHA1 Message Date
Stefan Krah 5d0d2e2b04 Explain the strategy to avoid huge alignment shifts in _mpd_qadd() in detail. 2012-04-18 18:59:56 +02:00
Stefan Krah ed4b21ff4f Cosmetic change: initialize digits to 1 (redundant). 2012-04-18 18:45:22 +02:00
Stefan Krah bc771e9b19 Remove redundant finalization of the result. 2012-04-18 18:25:37 +02:00
Stefan Krah aecaf0b663 Fix comments and whitespace. 2012-04-18 18:08:20 +02:00
Stefan Krah 6369f77d20 Support mythical ones' complement machines. 2012-04-18 17:57:56 +02:00
Stefan Krah 140893cbaa The previous code is correct, but hard to verify: The libmpdec documentation
rightfully states that an mpd_t with a coefficient flagged as MPD_CONST_DATA
must not be in the position of the result operand. In this particular case
several assumptions guarantee that a resize will never occur in all possible
code paths, which was the reason for using MPD_CONST_DATA and saving an
instruction by omitting the initialization of tmp.alloc.

For readability, tmp is now flagged as MPD_STATIC_DATA and tmp.alloc
is initialized.
2012-04-18 17:48:34 +02:00
Stefan Krah ec766a6179 1) Remove claim of an input invariant that is only true for static mpd_t.
Resizing is used _inside_ libmpdec functions, and it is permitted to
   change x->alloc several times while setting x->len at the end of the
   function. Therefore, for dynamic mpd_t x->alloc can _temporarily_ drop
   below x->len. Of course the final result always has x->len <= x->alloc.

   For static mpd_t this cannot happen, since resizing to a smaller
   coefficient is a no-op.

2) Remove micro optimization in mpd_switch_to_dyn(): Previously only the
   valid initialized part of the existing coefficient up to x->len was
   copied to the new dynamic memory area. Now copying does the same as
   realloc() and the entire old memory area is copied.

   The rationale for this change is that it is no longer needed to memorize
   the explanation given in 1).
2012-04-10 23:11:54 +02:00
Stefan Krah 7b544ca08d Fix stale comment. 2012-04-10 23:08:29 +02:00
Stefan Krah cc74b6ab14 Issue #14478: Cache the hash of a Decimal in the C version. 2012-04-10 16:27:58 +02:00
Stefan Krah e37f8b29fc Issue #14520: Add __sizeof__() method to the Decimal object. 2012-04-09 21:27:20 +02:00
Stefan Krah f69aef747a Resize the coefficient to MPD_MINALLOC also if the requested size is below
MPD_MINALLOC. Previously the resize was skipped as a micro optimization.
2012-04-09 20:47:57 +02:00
Stefan Krah dd159ce606 Speed up _decimal by 30-40% for numerical workloads by improving the cache
locality for regularly sized coefficients.
2012-04-09 20:24:57 +02:00
Stefan Krah 94ef3e4c1d Use the MPD() accessor macro. 2012-04-09 19:20:46 +02:00
Stefan Krah dc36efa368 1) Fix comment.
2) Assert that the source operand is not special. Prevent resulting assert
   failure (harmless) by initializing flags before calling mpd_qshiftr_inplace.

3) Save a couple of instructions (mpd_zerocoeff already sets digits and len).
   Reorder initialization to match the order in the mpd_t struct.
2012-04-07 15:57:59 +02:00
Stefan Krah 4771cca817 Whitespace. 2012-04-05 16:15:25 +02:00
Stefan Krah 871b96bd5a Reduce array size. 2012-04-05 16:07:22 +02:00
Stefan Krah a6169484c2 Formatting. 2012-04-05 15:48:59 +02:00
Stefan Krah ff3eca0cc3 Allow printing a leading '-' and the maximum number of exponent digits
rather than raising RuntimeError (allocated space is sufficient for the
additional character).
2012-04-05 15:46:19 +02:00
Stefan Krah 0774e9b9f5 Raise InvalidOperation if exponents of zeros are clamped during exact
conversion in the Decimal constructor. Exact here refers to the
representation and not to the value (clamping does not change the value).
2012-04-05 15:21:58 +02:00
Stefan Krah 91c0274bc4 Improve comments. 2012-04-02 20:51:08 +02:00
Stefan Krah 5100171d81 Clear the context flags if a context is initialized from the DefaultContext. 2012-04-02 15:02:21 +02:00
Stefan Krah 41e031004b Fix Overflow exception in the bignum factorial benchmark that is due to
the recent change of the default value for context.Emax.
2012-04-01 23:25:34 +02:00
Stefan Krah 0e41981cd5 Use abort() rather than exit() to appease tools like rpmlint. abort() is used
in libmpdec to prevent undefined behavior if an invalid context is used. This
cannot occur for the _decimal module since user input for the context is
validated.
2012-03-30 14:12:20 +02:00
Stefan Krah fe17b2bc77 Raise MemoryError instead of InvalidOperation/MallocError for compatibility
with decimal.py. The standard specifies InsufficientStorage (MallocError) as
a sub-condition of InvalidOperation. This allows a calculation to continue
with NaN results when allocation fails.
2012-03-25 18:59:21 +02:00
Stefan Krah c64150bcac Fix formatting after removing tabs. 2012-03-23 16:34:41 +01:00
Stefan Krah cd9e1d0205 Whitespace. 2012-03-23 16:22:05 +01:00
Stefan Krah b6405efd1b Use the same exception hierarchy as decimal.py. FloatOperation now also
inherits from TypeError. Cleanup in module initialization to make repeated
import failures robust.
2012-03-23 14:46:48 +01:00
Stefan Krah 7cc5521d40 Whitespace. 2012-03-21 20:21:20 +01:00
Stefan Krah 1919b7e72b Issue #7652: Integrate the decimal floating point libmpdec library to speed
up the decimal module. Performance gains of the new C implementation are
between 12x and 80x, depending on the application.
2012-03-21 18:25:23 +01:00