[Eeglablist] Installing CUDAICA on Windows 10 (2021 update)

[周云晖]

Hi Makoto,

Yes, CUDAICA for Windows use double precision number, it is specified in “config.h”. I remember the algorithm may get numerically unstable in rare cases if we use single precision mode.

Best,
Yunhui

> -----原始邮件-----
> 发件人: "Makoto Miyakoshi via eeglablist" <eeglablist at sccn.ucsd.edu>
> 发送时间: 2021-11-16 09:29:12 (星期二)
> 收件人: "EEGLAB List" <eeglablist at sccn.ucsd.edu>
> 抄送: 
> 主题: Re: [Eeglablist] Installing CUDAICA on Windows 10 (2021 update)
> 
> Dear Yunhui,
> 
> Thank you for your comment, Yunhui. It is an honor to discuss the CUDA
> installation with you!
> 
> > Actually to run CUDAICA for Windows, you don't need to install Microsoft
> Visual Studio at all. Only NVIDIA CUDA and Intel MKL is needed. Visual
> Studio is only required if you want to build the exe file from source.
> 
> That is very good to know! I updated the section of the Wiki accordingly.
> Please let me know if the description is closer to your recommendation.
> https://sccn.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code#By_using_CUDAICA_.2811.2F15.2F2021_updated:_Thank_you_Ugo_and_Yuhui.21.29
> 
> I used your .exe file and did not build it. I would not be able to make
> CUDAICA work if I need to build it myself because I could not find any
> detailed instructions to do so for a real naive person like me. One of my
> collaborators could do it all by himself, but it is because he is
> exceptionally skilled. Most of standard users would find it much easier to
> download your pre-build .exe file. I deeply appreciate your contribution to
> the community, Yunhui.
> 
> May I ask a question?
> CUDA works much faster for single precision than double precision. I know
> runica() uses double precision because the algorithm includes matrix
> inversion. Does it mean CUDAICA also uses double precision? Are we using
> CUDA on its slower side? Just curious.
> 
> Makoto
> 
> 
> 
> On Mon, Nov 15, 2021 at 4:46 PM 周云晖 <yhzhou17 at fudan.edu.cn> wrote:
> 
> > Hi, I am the maintainer of CUDAICA for Windows.
> >
> > I have looked at the link (
> > https://sccn.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code#By_using_CUDAICA_.2811.2F10.2F2021_added.29)
> > for the instruction of installing CUDAICA for Windows.
> >
> > Actually to run CUDAICA for Windows, you don't need to install Microsoft
> > Visual Studio at all. Only NVIDIA CUDA and Intel MKL is needed. Visual
> > Studio is only required if you want to build the exe file from source.
> >
> > I have also noticed that Intel has re-packaged MKL into oneAPI toolkit. In
> > our lab we sticked to Intel MKL 2020.4 so the Github instruction is not
> > updated. Sorry for that, but I will update the instruction in the following
> > few days.
> >
> > There is also a bug that CUDAICA for Windows cannot read binary files
> > larger than 4GB. This has been fixed a few days ago. If someone has run
> > into this issue before, you may try the updated version.
> >
> > Best,
> >
> > Yunhui
> >
> >
> > > -----原始邮件-----
> > > 发件人: "Makoto Miyakoshi via eeglablist" <eeglablist at sccn.ucsd.edu>
> > > 发送时间: 2021-11-11 14:02:10 (星期四)
> > > 收件人: "EEGLAB List" <eeglablist at sccn.ucsd.edu>, ugob at siu.edu
> > > 抄送:
> > > 主题: [Eeglablist] Installing CUDAICA on Windows 10 (2021 update)
> > >
> > > Dear list members,
> > >
> > > I summarized the steps to install cudaica() which uses GPU computation to
> > > calculate infomax ICA (Raimondo et al., 2012). The result from the speed
> > > comparison between runica() and cudaica() was not as dramatic as x25
> > > reported by the original paper, probably because Tjerk's smart hack alone
> > > already gave x4-5 speed up to runica(). Still, using a relatively cheap
> > > GTX1660 (the pre-COVID price range is $250), I confirmed x4-5 speed up
> > > compared with runica(). The detailed instruction can be found in the
> > > following link.
> > >
> > >
> > https://sccn.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code#By_using_CUDAICA_.2811.2F10.2F2021_added.29
> > >
> > > WARNING: The installation was difficult.
> > >
> > > Makoto
> > > _______________________________________________
> > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > To unsubscribe, send an empty email to
> > eeglablist-unsubscribe at sccn.ucsd.edu
> > > For digest mode, send an email with the subject "set digest mime" to
> > eeglablist-request at sccn.ucsd.edu
> >
> >
> >
> >
> >
> >
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu