| Re: [eigen] intiial ARM NEON results |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] intiial ARM NEON results
- From: Rohit Garg <rpg.314@xxxxxxxxx>
- Date: Sat, 27 Feb 2010 19:26:53 +0530
- 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 :from:date:message-id:subject:to:content-type :content-transfer-encoding; bh=CKmvVsWj7jKs8/5JFijbyWSCJ4SQhyngaITBeYv5wGg=; b=R+fCDlnHqZzc2x9u2zwvRbLCfqqqFM1Zqq2nW39dODIkG08ca11gffxEtaVfWUEv/1 GkKXB0EJ1w17dNBdXyoNgMpSBL7j/G4RC0eiYkWIFYWxoJcod+AokbABx7J1ROY29Pe8 BEYeRDGwTNIneMtaMQ3/mq4pI7Cjjn53AeMAI=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type:content-transfer-encoding; b=Yml+kkVE5QjkHPD7AMTrKCfLjbF1LurSQEoovcnqAN4ztGoU69pTAZSY0RaA//9Xzt Sr1frQse615CsFNIfWQ55SVsKDJ2JTwn4lgFsV91PW6CPXpqg58vvM3aTAtp9LTz/iwu /MaMMJ30k6MRKaMpkvD/RBoLYMG/IbLxcwIs4=
On Sat, Feb 27, 2010 at 7:17 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
> sounds good! Looking forward to the patch :)
> Benoit
>
> 2010/2/26 Konstantinos Margaritis <markos@xxxxxxxx>:
>> On Saturday 27 February 2010 00:37:28 Konstantinos Margaritis wrote:
>>> Hi everyone,
>>>
>>> I haven't finished yet, there are 4-5 functions to finish in the ARM NEON
>>> port, but here is a benchmark that I managed to compile and run:
Good job! With increasing proliferation of ARM based devices having
NEON support, this could give Eigen on such devices a big push.
>>>
>>> $ PATH=~/native/bin/:$PATH g++ -O3 -DNDEBUG -DREPEAT=100 -I../
>>> benchVecAdd.cpp -o benchVecAdd -lrt $ ./benchVecAdd
>>
>> er stupid me, that's obviously a mistake, I copied from a previous run. Here are the correct results:
>>
>> $ PATH=~/native/bin/:$PATH g++ -O3 -DNDEBUG -DREPEAT=1000 -I../ benchVecAdd.cpp -o benchVecAdd -lrt
>> $ ./benchVecAdd
>>
>> 7.73s 0.0192771 GFlops
>> $ PATH=~/native/bin/:$PATH g++ -O3 -DNDEBUG -DREPEAT=1000 -mfpu=neon -mfloat-abi=softfp -I../ benchVecAdd.cpp -o
>> benchVecAdd -lrt
>> h$ ./benchVecAdd
>> 2.93s 0.0508572 GFlops
>>
>> That's ~2.64x faster!
>>
>> Konstantinos
>>
>>
>>
>
>
>
--
Rohit Garg
http://rpg-314.blogspot.com/
Senior Undergraduate
Department of Physics
Indian Institute of Technology
Bombay