[release-0.5] Backport #19022 #21708
Conversation
|
last time around this caused build issues which is why I was leaving this alone on release-0.5 |
|
please cherry-pick with |
staticfloat
force-pushed
the
sf/backport_19022
branch
from
7737162 to
b4498fa
Compare
I'm not sure exactly what build issues you're referring to, but I think it's worth the effort so that we can support non-NEON ARM architectures. I've redone my cherry-picking, thanks for the pointers. |
|
still missing a reference to the pr # in the commit messages |
staticfloat
force-pushed
the
sf/backport_19022
branch
from
b4498fa to
cab5345
Compare
|
Sorry, missed that. Done. |
|
those commit shas also don't make much sense, those aren't linking to the commit that's on master |
* Allow `cpu_target` to specify a generic arch, matching the behavior on x86
* Detect the CPU arch version with `uname`
* Require `armv6`
Close #13270 (`armv5` is not supported)
Fix #18042
* Remove warning about generic arch since it's not really useful
Fix #17549
* Require at least the same ARM arch version and profile the C code is compiled with
(cherry picked from commit 760bc41 and PR #18100)
staticfloat
force-pushed
the
sf/backport_19022
branch
from
cab5345 to
ea99033
Compare
|
Rarely have I been so confused by a git mess. One more try. |
|
Closing this as 0.6.0 is out now. |
This backports some fixes for ARMv7 support. It makes it easier for Julia to autodetect the proper version, and fixes the binaries so that they can run on ARM processors that don't have NEON support. If I download the beta 0.5.2 binary (6accda38c4) from this thread and try to run it on a scaleway server with no NEON support, it
SIGILL's. With this change, we can compile withJULIA_CPU_TYPE=generic(which is what the buildbots are doing anyway) and it works okay.