Fork me on GitHub
#graalvm
<
2020-04-19
>
sb05:04:50

I would like to use lib-python and repl via graalvm on aarch64. If that isn’t possible I use Golang or C++ but hopefully that is possible solve.

sb05:04:27

In the past I used not the official way compile things via substratevm.. but changed lot of things..

sogaiu07:04:12

to clarify, i had success with producing aarch64 binaries of clj-kondo and bb. the compilation succeeded on android via proot, but it also worked in a qemu aarch64 environment (-M virt). qemu building takes much longer -- in the case of bb it took 2 hours in qemu vs 20 mins on android.

sogaiu07:04:04

iiuc, rpi 3 should be possible as well with an appropriate distribution, but iiuc memory is limited there.

sogaiu07:04:31

may be that should be "should be possible to run binaries" -- building may be another story.

sb08:04:04

@sogaiu did you use java11? Yesterday for me on Ubuntu doesn’t work with official graalvm image :/ - before this I successfully compiled with substratevm/old way.

sb08:04:14

I added extra 4gb swap and I have 4gb. But didn’t finish within 3-4 hours. I try again..

sb08:04:25

Maybe qemu better idea..

sogaiu08:04:57

not quite sure i follow -- what i used was: https://github.com/graalvm/graalvm-ce-builds/releases/download/vm-19.3.1/graalvm-ce-java11-linux-aarch64-19.3.1.tar.gz (so that's the aarch64 build for linux that is tied to java11)

sogaiu08:04:08

to get bb to work some changes were necessary -- those should be summarized here: https://github.com/borkdude/babashka/issues/241#issuecomment-610908212

🍻 4
sogaiu08:04:53

assuming the tutorial is followed, my qemu invocation looked like this:

qemu-system-aarch64 \
  -smp 4 \
  -M virt \
  -m 8192 \
  -cpu cortex-a53 \
  -kernel vmlinuz-4.9.0-12-arm64 \
  -initrd initrd.img-4.9.0-12-arm64 \
  -append 'root=/dev/vda3' \
  -drive if=none,file=hda.raw,id=hd \
  -device virtio-blk-pci,drive=hd \
  -netdev user,id=mynet \
  -device virtio-net-pci,netdev=mynet \
  -nographic

👍 4
sogaiu08:04:20

the vda3 part will obviously depend on how partitioning is done -- my value is different from what was in the tutorial

sogaiu08:04:49

note that i allocated 8gb of memory

👍 4
sogaiu08:04:58

...and 4 cores

sogaiu08:04:00

changing the -Xmx value in bb's build script seemed to help too

sb08:04:46

Thanks!! I check all things and redo now!! 👍

sogaiu08:04:24

good luck 🙂

sb18:04:58

@sogaiu works!! 🎉

👍 4
sb18:04:15

sbThat was slow really on jetson nano