Re: [Boost-bugs] [Boost C++ Libraries] #9233: Suboptimal shift by multiples of 8 bits

Subject: Re: [Boost-bugs] [Boost C++ Libraries] #9233: Suboptimal shift by multiples of 8 bits
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2016-02-26 10:19:04

#9233: Suboptimal shift by multiples of 8 bits
-------------------------------+----------------------------
  Reporter:  psiha             |      Owner:  johnmaddock
      Type:  Bugs              |     Status:  closed
 Milestone:  To Be Determined  |  Component:  multiprecision
   Version:  Boost 1.54.0      |   Severity:  Optimization
Resolution:  wontfix           |   Keywords:
-------------------------------+----------------------------

Comment (by psiha):

 Replying to [comment:3 johnmaddock]:
 > Fixed in
 https://github.com/boostorg/multiprecision/commit/cb1a41835f4566987b96c7a7034088e22d83f77b


 Thanks John :)
 I no longer use/need this functionality so I didn't test but from a
 cursory skim it looks good...except that maybe the (little) endian checks
 are not really necessary: if you make them in only one place - where you
 decide on the underlying/raw/limb layout - if you order the limbs in
 'machine endianess' (I'm guessing) the rest of the code can be endianness
 agnostic (and more readily assume correct byte ordering/use
 memcpy/move/set more easily)...

 > It turns out that memmove is (very slightly) faster than moving whole
 limbs so that's the preferred method when available.

 It's also smaller (one call instead of a loop for every instantiation;)

 > Limbs are half the size of the largest available integers - so 32-bit
 for msvc as there's no 128-bit integer support (unlike mingw for example).

 Thanks for the explanation ;)

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/9233#comment:4>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.