On 3/25/26 20:46, Alexander Kanavin wrote:
CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know
the content is safe.
On Wed, 25 Mar 2026 at 03:26, hongxu via lists.openembedded.org
<[email protected]> wrote:
Add a cpu-optimization option for disable-optimization [1], disabled by default,
otherwise import numpy failed on qemux86-64 with kvm
[snip]
import numpy as np
...
RuntimeError: NumPy was built with baseline optimizations:
(X86_V2) but your machine doesn't support:
(X86_V2).
[snip]
This error is odd. The default tune is v3, qemu guest does support
both v2 and v3, so perhaps you could check why numpy selects v2, and
claims that machine doesn't support that, and whether we can configure
it to match?
+PACKAGECONFIG[cpu-optimization] =
"-Ddisable-optimization=false,-Ddisable-optimization=true"
Reading https://numpy.org/doc/stable/reference/simd/build-options.html,
I do not think setting disable_optimization to true by defailt is
correct. Currently, numpy will auto-detect the capabilities of target
machine, and configure itself accordingly. With this change, that auto
detection is compeltely turned off.
Please explain the rationale for this change.
In our case, we use `-cpu kvm64' other than `-cpu Skylake-Client
-machine q35,i8042=off'
QB_CPU_KVM:intel-x86-64 = "-cpu kvm64"
//Hongxu
Alex
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#233962):
https://lists.openembedded.org/g/openembedded-core/message/233962
Mute This Topic: https://lists.openembedded.org/mt/118495599/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
