Re: [eigen] Re: vectorization for ints not working

[Rohit Garg <rpg.314@xxxxxxxxx>]

On Sat, Sep 12, 2009 at 1:31 PM, Gael Guennebaud
<gael.guennebaud@xxxxxxxxx> wrote:
> yes this is because I did not expect such kind of use of Map for small
> fixed size vectors. Actually why don't you use an Eigen vector for
> "bytes" ? Then you won't need to use Map.

Good idea. I think there might be a way to do this. Actually, this
code is to be wrapped to python via swig and  I wanted to avoid
letting swig run amok eigen's headers, so it was done this way.

> Anyway, to make Map more usable what we have to do is to add a
> FirstAligned option to Map so that Eigen knows your data is
> aligned.... The ForceAlign option has a different meaning.

What does ForceAligned mean then? I though it was to let eigen know
that the pointer is aligned. :|

>
> cheers,
> gael.
>
> On Sat, Sep 12, 2009 at 8:24 AM, Rohit Garg <rpg.314@xxxxxxxxx> wrote:
>> PS: this is eigen's unstable branch, but using code checked out a
>> month or so ago.
>>
>> On Sat, Sep 12, 2009 at 11:50 AM, Rohit Garg <rpg.314@xxxxxxxxx> wrote:
>>> I had a simple piece of code involving matrix4i, and it is not
>>> compiling with vectorization. This is on fedora 11, 64 bit, gcc 4.4.1.
>>> To compile, use make eigen. I am interested in vectorization of the
>>> code in the addressesGen function, which enclosed in asm directives.
>>> There are clear signs of eigen unrolling the loop, but I'd like my
>>> vectorization please. :)
>>>
>>> To compile, do make eigen.
>>>
>>> --
>>> Rohit Garg
>>>
>>> http://rpg-314.blogspot.com/
>>>
>>> Senior Undergraduate
>>> Department of Physics
>>> Indian Institute of Technology
>>> Bombay
>>>
>>
>>
>>
>> --
>> Rohit Garg
>>
>> http://rpg-314.blogspot.com/
>>
>> Senior Undergraduate
>> Department of Physics
>> Indian Institute of Technology
>> Bombay
>>
>>
>>
>
>
>



-- 
Rohit Garg

http://rpg-314.blogspot.com/

Senior Undergraduate
Department of Physics
Indian Institute of Technology
Bombay