Installation on TX1 or TX2?


#1

I know these aren’t ideal systems for MapD, but any reasons it won’t work? I wanted a test install to play with and I already have an Nvidia TX1 dev kit.


Hello. Could you help me . About Platform
#2

Hi,

Thanks for looking into MapD

We have also been working with TX2 on MapD recently :slight_smile:

It should also work on the TX1 but I don’t know that we have checked it recently.

I think it should build fine there

Regards


#3

I’ll be updating the readme in the next few days, but in general you’ll want to follow the Ubuntu instructions for installing the dependencies at https://github.com/mapd/mapd-core#ubuntu-1604-1610 . When Folly fails to build, open its Makefile and remove the -msse4.2 compiler flag. Everything else should then build successfully.

You will also need to update LD_LIBRARY_PATH to add:

LD_LIBRARY_PATH=/usr/lib/jvm/java-8-openjdk-arm64/jre/lib/aarch64/server:$LD_LIBRARY_PATH

#4

Thanks for the follow up.

With twice the memory and bandwidth, the TX2 is certainly a better fit. Hope to try one soon.

I can think of several possible uses for MapD on portable GPU’s (e.g. robotics).

Are you allowed to say what your application is?


#5

Finally updated my TX1 to the 3.0 jetpack to try out MapD.
I noticed the binary download (with Immerse) is x86 only.

If I build MapD from source for the TX1, will I be able to use your Immerse UI?


#6

Hi @ian, yes right now you’ll have to build from source for ARM/Jetson. Immerse should work out of the box but we’re working through some issues related to backend rendering (i.e. for pointmap/scatterplots/choropleths) on the Jetson so only frontend-rendered (i.e. d3) charts currently work on that platform.

Once ARM is fully supported we will likely provide a community edition build for it as well.


#7

That’s good to know

I’m actually interested in using it with an AR headset and the GPU rendering sounded very interesting. Please let me know when the issues have been resolved.

Is there documentation on the GPU rendering? For example, guarantee of high frame rates, or output of large graphs (node-edge) via various layout algorithms, 3D render, etc… I can dig into the source, but wondered if I overlooked any documentation.


#8

Ian,

Keep an eye on our website for news on the newer version

Just a reminder the render code is not in the open source code.

More detailed documentation around our render process is being finalized presently, and will go up on docs.mapd.com when it is available

We do not currently support any kind of graph layout algorithms currently, but we are working with Graphistry in the GOAI project which may produce the kind of layout you thinking about.

regards