This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Seeming g++-specific "feature deficit"...
- From: "Linda A. Walsh" <gcc at tlinx dot org>
- To: Jonathan Wakely <jwakely dot gcc at gmail dot com>
- Cc: libstdc++ <libstdc++ at gcc dot gnu dot org>
- Date: Sat, 04 Oct 2014 12:07:27 -0700
- Subject: Re: Seeming g++-specific "feature deficit"...
- Authentication-results: sourceware.org; auth=none
- References: <542F6960 dot 2050205 at tlinx dot org> <alpine dot DEB dot 2 dot 11 dot 1410040836130 dot 3813 at stedding dot saclay dot inria dot fr> <542FEFDA dot 1040205 at tlinx dot org> <CAH6eHdSEQq3y2mWjxTyk4M9EaBA_RhpMXxGj6Vh7WSEHJ08EvQ at mail dot gmail dot com>
Jonathan Wakely wrote:
As I'm pretty sure I've said before, you seem to be trying to use
valarray where it isn't suitable.
----
I didn't get that it was designed to be separate
from the STL...
Some of the features like "sum", are very useful in my
application, I could use vector for that if I wanted
to implement my own class on top of vector... ;-/
After a few years of what works under my belt I'm sure this
will all seem obvious, but after a few months... eh... not so
much.
I'm also doing math on the arrays -- mult add div average...
admittedly, I didn't check, but it didn't seem vectors were
designed for that, which was why I have err'd toward
the side of valarrays...the lack of just the 'sum' function
was a major "downer", from the standpoint that I wanted to
use the builtin (not that it is hard to write an addon), but
I thought the builtin numeric ops on valarray were pretty cool.
I've been frustrated at the division in applicable functions --
though I entirely understand the need for alignment in the valarrays,
still wish they would have gone with a more orthogonal implementation
with clear caveats about where usage of such features impact
performance.
I hope you'll appreciate my attempts to come up to speed,
despite the many wrong dead-ends I've already gone down and
backtracked on.
Sadly, my ability to get things done has not been so much due to
any brilliance but the ability to make (and correct) mistakes faster
than most (though that rate has gotten slower w/age).