ChanServ changed the topic of #aarch64-laptops to: Linux support for AArch64 Laptops (Asus NovaGo TP370QL - HP Envy x2 - Lenovo Mixx 630 - Lenovo Yoga C630)
<steev>
oh, yeah, that's more so for ubuntu's support than debian's
<jhovold>
Changes since the previous branch include:
<jhovold>
- fix module loading regression in -rc4 that broke display
<jhovold>
- camera pmic support
<jhovold>
- efivars support (update to v4)
<jhovold>
- soundwire reset support (update to v2)
echanude has joined #aarch64-laptops
<jhovold>
- drop unused (?) edp panel entry
svarbanov has quit [Ping timeout: 480 seconds]
<clover[m]>
Wow, does this mean the webcam working?
<konradybcio>
no
<steev>
quinine: not yet, afaik
echanude_ has joined #aarch64-laptops
echanude has quit [Ping timeout: 480 seconds]
<danielt>
jhovold: Pretty much trivial in the grand scheme of things but whilst making sure I am up to date with config changes I spotted that johan_defconfig sets CONFIG_QCOM_QSEECOM=y and I don't think this exist the kernel anymore...
<steev>
correct, the new option is QCOM_SCM_QSEECOM
<quinine>
Qualcomm says the x13s' NPU performance is 5x that of Intel's 12t i5, so maybe it can be used to do something interesting.
<clover[m]>
In windows I think its just used for Webcam stuff like auto eye contact
<exeat>
Not even remotely an expert in "cloud-class" "generative AI experiences" -- but wouldn't opencl work to accelerate that stuff?
<quinine>
opencl doesn't seem to have enough ecology, not sure if that's because API limits performance. I know most people are using cuda.
<qzed>
mostly ecosystem reasons afaik: opencl should (at least in its latest versions) be performance-wise as "unlimited" as cuda
<qzed>
nvidia doesn't (? didn't at least last I got down to that abstraction-level, might have changed a bit) really care about opencl (so not sure if there's a remotely as well optimized compiler as there is for cuda)
<qzed>
and AMD really needs to play catch-up on the software-side...
<qzed>
so all the software developed around cuda because nvidia pushed that hard
<qzed>
but nowadays for deploymend you have onnx, which should be mostly backend agnostic and also work on accelerators like NPU
<jhovold>
danielt: thanks, i verified that the new symbol is selected by Kconfig, but forgot to remove the old one from the defconfig
<jhovold>
clover[m]: not yet, but the pmic is needed to power the camera sensors
<HdkR>
(Also every driver vendor stopped working on OpenCL since 2.0 was a nightmare. 3.0 changed that)
<steev>
jhovold: oh nice, i'd seen the initial push a while back, and looked through it
<steev>
i figured you didn't announce it, for reasons
<jhovold>
yeah, that was an even more preliminary version, the pmic driver is still going to need to be reworked
<steev>
i read through the email thread and saw that :(
<jhovold>
i plan on helping out with that, as soon i can find some time...
<jhovold>
as*
echanude has joined #aarch64-laptops
echanude_ has quit [Ping timeout: 480 seconds]
<clover[m]>
HdkR: tried any more wwan slot hacks lately or is that totally futile?
baozich has quit [Ping timeout: 480 seconds]
<HdkR>
clover[m]: Doesn't matter when it's vendor locked.
baozich has joined #aarch64-laptops
<clover[m]>
would a WWAN enabled model be needed to do something like put an SSD in there?
<clover[m]>
or is it even more locked than that
<HdkR>
It's a PCIe vendor id lock, the port is still there regardless
<emily[m]>
I think someone successfully reflashed an ssd to report a whitelisted id