search

Geontech / meta-redhawk-sdr

REDHAWK SDR Layer for Yocto/OpenEmbedded -based deployments
http://geontech.com/getting-started-with-meta-redhawk-sdr/
GNU Lesser General Public License v3.0
9 stars 6 forks source link

sse2neon - better machine support #59

Open btgoodwin opened 4 years ago

btgoodwin commented 4 years ago

The COMPATIBLE_MACHINE = "(arm|aarch64)" usage in this recipe will not work for devices like the Ettus E310 since those values aren't in the MACHINEOVERRIDES. I'm sure this extends to other machines as well. I'm open to other suggestions, but this would be more flexible since in those cases, OVERRIDES will almost certainly have arm or aarch64 in the list (otherwise the dependency chain for rh.DataConverter would not have reached over to those patches and dependencies lists):

COMPATIBLE_MACHINE_arm = "${MACHINE}"
COMPATIBLE_MACHINE_aarch64 = "${MACHINE}"

We would need to apply this as far back as rocko.

btgoodwin commented 4 years ago

@jschindehette @ajung123 Thoughts?

rodrigo455 commented 4 years ago

the presence of neon in TUNE_FEATURES, isn't it relevant?

btgoodwin commented 4 years ago

It is. Are you suggesting something even more generic like this?

COMPATIBLE_MACHINE = "${@bb.utils.contains('TUNE_FEATURES', 'neon', '${MACHINE}', '' ,d)}"

edit: I'm not sure if simply '1' is good enough there, or even if the syntax of this is correct now.

rodrigo455 commented 4 years ago

I've seen this in cpuburn-neon recipe:

COMPATIBLE_MACHINE = "(${@bb.utils.contains("TUNE_FEATURES", "neon", "${MACHINE}", "Invalid!", d)})"

but now it is different in more recent releases of meta-openembedded (in warrior branch it is renamed to cpuburn-arm), which appears to follow your idea

COMPATIBLE_MACHINE ?= "(^$)"
COMPATIBLE_MACHINE_armv7a = "(.*)"
COMPATIBLE_MACHINE_armv7ve = "(.*)"
COMPATIBLE_MACHINE_aarch64 = "(.*)"

you could also check the ne10 recipe

btgoodwin commented 4 years ago

It probably has to do with parsing vs. post-parsing (override syntax) as to how successful the in-line python actually is. So override syntax it is then?

rodrigo455 commented 4 years ago

from my side... I would say so. it is already better than the current approach indeed

btgoodwin commented 4 years ago

~Alright, I'll make this part of the 2.2.6 update unless there's a pressing need. A couple of us are chipping through Zeus compatibility right now, which may see us move to gnu++11 (vs. gnu++98) as well as a few other patches.~

I pushed zeus-next but it doesn't have much testing yet. The patch is d4474443bcbed6ed047de2d1cf92072f93505ab7; need to propagate it through the rest of the branches.