| Re: [eigen] Eigen on an arm core |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Eigen on an arm core
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Sat, 8 Aug 2009 20:12:28 +0200
- 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=7+OcJ0TtVSLtvjUjWXzhYhXe9788ebxGzN3JlhK/c0U=; b=MzE3ZYWWD+BuGswdvD1YyGH0PdcamC9fN3x+0FW2B7JIP6SbB81PMz3WUSifNDYHmt FqysyWyI1lg7pC9Gum6pcwpUOTG5bbWSAri29g1mM+1TyvlP2efv6fTWHEWft74ysq9h u8mau3ipVoUt97XETjR0m3IQqTrzabPsuCt/c=
- 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=V2pWDNLd3Rh43yWpTKYvLh40UeB8SELrOfGj+sqoyLJAU9ncwmP9fOG0Bfjnb8v/zy gcJSMNoefJ56ofQCFSercdVyT2A1TXOedcY3rT/tjOzcC/GthfrHeXvmZL/K06tALWax iwRV3rYJ7olQP9BSavUAiMJcJhYuySLVXHh8A=
2009/8/8 Daniel Stonier <d.stonier@xxxxxxxxx>:
> Ok, that seems to have sidestepped the problem completely - thanks and
> sorry for the inconvenience!
:)
>
> Just out of curiosity, how much of a boost does vectorisation give you
> with the SSE capabilities on intel cores?
It depends a lot on what you're doing. It can be up to 4x and even
slightly above (i've seen 4.7x). On the other hand, it can bring no
boost at all in other cases.
Benoit
>
>
> 2009/8/9 Daniel Stonier <d.stonier@xxxxxxxxx>:
>> I just got pulled into a project they have been hammering away at for
>> a while. Unfortunately its a bit of chaos. They're not really clued
>> into programming on linux and are trying to do something non-trivial.
>> Lots of fun!
>>
>> Seems as though they're doing the msproject style and bundling every
>> library and its kitchen sink with the project, so the eigen is a bit
>> old.
>>
>> Toolchain is a generic gcc toolchain for arm developed by samsung.
>> Version is 2.0.0.
>>
>> I'll update now and let you know how it goes.
>>
>> Cheers,
>> Daniel.
>>
>> 2009/8/9 Benoit Jacob <jacob.benoit.1@xxxxxxxxx>:
>>> With Eigen >= 2.0.1 and the GCC compiler, indeed we disable all
>>> special alignment code, so we also disable this assertion, that's why
>>> you don't have this issue.
>>>
>>> As long as we don't do vectorization on a certain platform, there's no
>>> need to bother at all about alignment. That's why the solution to
>>> David's problem is as simple as: either upgrade to Eigen >= 2.0.1 (if
>>> he's using GCC) or tell us what compiler you're using and how to port
>>> the platform check.
>>>
>>> FYI the platform check is at Eigen/src/Core/util/Macros.h line 44 (in
>>> the 2.0 branch). We need to know how to detect your compiler, and how
>>> to detect the CPU architectures listed there, on this compiler.
>>>
>>> Cheers,
>>> Benoit
>>>
>>> 2009/8/8 Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>:
>>>>
>>>>
>>>> Hello,
>>>>
>>>> I can't help you there, but I often test my project on ARM (and MIPS, sparc32,
>>>> ppc32, x86, amd64) and i don't have this kind of problem. I'm using eigen
>>>> trunk or 2.0.3, gcc 4.3.2/4.4.3 . Maybe we don't use the same features of
>>>> eigen though.
>>>>
>>>> "Testing" here means "compile and do some unit tests"
>>>>
>>>> regards,
>>>> Thomas
>>>>
>>>> In data sabato 08 agosto 2009 19:13:31, Daniel Stonier ha scritto:
>>>>> We threw eigen at an arm core recently running a SLAM algorithm and
>>>>> ignorantly discovered we were completely unknowing about alignment
>>>>> issues!
>>>>
>>>> --
>>>> Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>
>>>> http://www.freehackers.org/thomas
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>
>
>