Installing Nvidia CUDA on Ubuntu 14.04 for Linux GPU Computing | Hacker News Reader

Installing Nvidia CUDA on Ubuntu 14.04 for Linux GPU Computing

quantstart.com

47 pointsshogunmike14 comments

Would this be significantly easier if there were just a Dockerfile with all this stuff in it and then the instructions would be, "docker build -t cuda . && docker run -t -i cuda /bin/sh"? Or maybe make it a trusted docker image and developers could just pull/run it?

Later: perhaps not. Graphics drivers and the like rely on an interactive terminal and an X11 system, AFAICT.

I think this could actually be done if the container is run in "--privileged" mode.

I think it is more than that. The installation has steps that fail because of terminal requirements at install-time:

    Unpacking x11-common (from .../x11-common_1%3a7.6+12ubuntu2_all.deb) ...
    debconf: unable to initialize frontend: Dialog
    debconf: (TERM is not set, so the dialog frontend is not usable.)
    debconf: falling back to frontend: Readline
    debconf: unable to initialize frontend: Readline
    debconf: (This frontend requires a controlling tty.)
    debconf: falling back to frontend: Teletype

Have a look at the github repo:

https://github.com/hughdbrown/docker-cuda

I thought HFT was moving toward FPGA? Does anyone know where/how GPU-HPC is used?

You're right. High-frequency trading is already using FPGA and ASICs.

GPU-HPC is still used heavily (within finance) for derivatives pricing techniques, via Monte Carlo and Finite Difference methods.

First of all, HFT is by far not the only place for HPC. GPU/Intel MIC HPC usage is still growing in scientific computing. Then, when we look at the financial world, there's still lots of interest in GPGPU, but rather on the more heavy weight analysis side. Not everything needs to be done in single digit milliseconds, there's decisions that can still be done in mere seconds where you'd rather have a more general purpose capable approach that's relatively easy to program - which is where GPUs come in, since their programmability becomes increasingly easy. I know the company that released the first full blown F# port of CUDA, mainly for the financial sector: https://quantalea.net/. They can even integrate that stuff with Excel so the bankers can have tons of calculations in their spreadsheets and still have the graphs update in realtime when they use their pretty sliders ;-).

Gah, just make nvidia support for ubuntu that works properly

Indeed, I had quite a bit of trouble getting it to work prior to this. The main issue seemed to be compatibility with Ubuntu packaged Nvidia drivers of various versions. It's certainly not as straightforward as it could be.

Installing the same version of the same package on Gentoo ~amd64:

    emerge nvidia-cuda-sdk

If only it was this easy in Ubuntu :-)

Well, its that easy if you can live with slightly old version. There is CUDA package in official repo IIRC.

Yep, just apt-get install nvidia-cuda-toolkit. It's currently v5.5.22 (a mid-2013 version). Debian started working on a packaging of v6.x in the "experimental" repository a few weeks ago, which will probably migrate to regular Debian and Ubuntu releases once it's tested a bit.

Well it won't be in Ubuntu before 15.04 so it keeps consistent in lagging a year behind.

I actually tried this a few months ago. I had some issues trying to run the typical vector addition "Hello world!" examples on my system.

I'm not 100% sure but I also think that by installing the nvidia-cuda-toolkit package it leaves out the CUDA samples. These contain the deviceQuery and bandwidthTest scripts necessary to check that CUDA is functional.

Admittedly I have two consumer cards in SLI, so that may have affected the install. It could also have been incompatibility between the actual Nvidia display drivers and the various dependencies. It's a bit messy!

Reply on news.ycombinator.com