|Re: [eigen] State of eigen support for small integers(16 bit, signed/unsigned)|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] State of eigen support for small integers(16 bit, signed/unsigned)
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Thu, 20 Aug 2009 10:42:48 -0400
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=7uGevQZt8wUz+Y64YPjA7vD5+IcWNcL7U7S2gQdPvxU=; b=uymCL8MJPOVt03cA7FD7vWB8bY3rqZ36YYqvRt+aqBJRhyLjtKEHfvsGSpVW10lRWI tZrhDJqXIl3L79Hrgllwl5xdG/e64BKexdYIhRWE91PdECeQWYmiHWWuJ8shPylOWPXw q5MNu/0t/fAfqCu6johBqKPN5OZbJOtvBMhMk=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:content-transfer-encoding; b=A9Oe+I/vx2SHEIpfHJmhhENbYpS0dY4oH6zba7MusIVx5G2nhSybm/VnMDkL5GBOZH mn2ENfoG8+VRHuSvQgh7EQiomD4HjCZLzL1ad+j+U9Q5sCGhdNLCwUfBS2dB6c7EKzlW Wfy8rhvUVH/9768NMdUwLCiAaMu8KFNLyzJTk=
2009/8/20 Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>:
> In data giovedì 20 agosto 2009 15:45:34, Benoit Jacob ha scritto:
>> Yes, most welcome. When I look at forums linking to our website, this
>> is a frequently wanted feature. The fact that people on this list
>> asked for it several times, finishes proving that it's a wanted
>> feature. Vectorization is most welcome here too -- the prospect of 8x
>> boosts on int16 and 16x boosts on int8 is very interesting...
> I'm quite curious : why would people need 8-bit or 16-bit integers nowadays ?
> I can only think of embedded platforms, but that does not seem to be the
> matter here, is it ?
Actually it is the matter for many Eigen users! And I believe that was
brought up in an earlier discussion on small int types.
> Or is it for the speed ? kind of ... trading precsion for
> speed, that's it ?
If you can replace int32 by int8, then not only you divide memory
usage by 4, but if you have vectorization then you can also multiply
the speed by 4, as 16 int8's fit in a single packet of 128 bits. That
makes small ints attractive on any platform where we have
vectorization, regardless of memory usage.