Just tried, latest ISO boots up OK.
On ASUS Vivobook S15 S5507QAD, which is x1e78100, 32GB variant. @xsl1ck Used usb-c stick, maybe give some another flash drive a try.
I got the T14s in mid December (assembled to specification, ie no pre-made). I directly did updates via Lenovo Vantage (don’t recall all that was updated though) before installing Ubuntu (and extracting firmware).
dmidecode says this about Memory Array Mapped Address:
I bought Thinkpad Yoga week ago with idea to check that. Right now, the answer is no, using Hexagon SDK is not possible with what is publicly available to my knowledge, or if there are plans to make it possible.
In addition Hexagon SDK doesn’t support x1e-80-100 and binaries for aarch64 are unavailable. Would be cool to get hexagon-clang build form armv7. But, oh well Hexagon+HVX llvm backend is upstream and you can compile your code with upstream clang just fine using SDK headers in most cases.
Hexagon development on Dell XPS 13/ThinkPad T14s which in itself is pretty compelling in comparison arm+hexagon cross and adb push ways of doing things.
In theory someday yes and FastRPC (Remote Procedure Calls to Hexagon ADSP/CDSP) could work same ways as in android, but some firmware bits are missing.
The kernel support and userspace code to get those part of puzzle seems to be mostly in the open or pretty close to make it possible. Probably not mini-dm for logging, this probably would need some work too.
I don’t know if Windows 11 firmware provided by qcom-firmware-extract is in any way similar to android one and even if it allows use of FastRPC. Or even which hardware units really are in this SoC. But marketing material seems to hint Hexagon DSP + HVX + HMX and maybe LPAI, since all this AI PC/NPU talk.
I’ve established so far it is likely Hexagon ISA v73 (might still be wrong), which wasn’t obvious and easy to find in itself.
I will be able to confirm that and what are capabilities once/if I’ll get to run code on ADSP/CDSP.
There is really no good public source of hard specs beside of calling it AI PC/NPU.
Based on this assumption, I think code signing would be required. With Hexagon v75 and up would be bit better in this regard.
Even if one does build FastRPC FastRPC github repo you’ll need signed/unsigned PD along with some support binaries likefastrpc_shell_<Domain> and few other bits.
Here is part of my attempt to use getserial from Hexagon SDK for code signing.
As expected:
Jan 30 18:47:28 prak-Yoga-Slim-7-14Q8X9 getserial[32416]: ../../src/fastrpc_apps_user.c:3554: Error 0x2: open_shell failed for domain 3 search paths used are /usr/lib/dsp/, /vendor/dsp/, /v>
So no Halide, TVM or your own Hexagon SDK apps for us on ubuntu-concept, unless we get some support from those who can (Qualcomm/linaro) in this area.
Also there are some device-tree fixes for cdsp for 6.14, so you’ll probably would need to pick up those yourself.
If we got that and hardware has required units, it would allow to access Vector Co-processor (ie HVX) and scalar DSP.
The HMX (tensor co-processor part) seems to be closed/not document/inaccessible outside of Qualcomm and usable only on pretty high level by AI SDK’s.
Maybe I’ll be able to get it working/hack my way using some android binaries for Hexagon v73 if I am lucky one day, but I don’t have my hopes up. Or it will turn out to be a gift for youtube watching on couch
I installed the latest iso image on my Thinkpad T14s. I can only get it to boot if there is a USB drive attached to the unit or if I reboot into Windows and then restart and select Ubuntu in the grub menu. If I boot Ubuntu from shutdown it will not get past the grub menu.
Any time frame when the display brightness for the low power IPS panel will get fixed. I see mention it was working previously, but not working on the latest iteration of the kernel?
Specs of Thinkpad:
CPU: X Elite 78-100
Memory: 32GB
SSD: 1TB
Display: 1920x1200 Low Power IPS
Keyboard: US
Hmm. I could only provide a separate dtb for PWM backlight control to be enabled. You would need to install it manually, and do a change in flash-kernel for it to work. Would that help?
Note sure who you were referring to, but I cannot copy anything from a USB-drive (on my T14s). I get a nice blue screen of death and a hard reset is the only thing that get the system up again. I can see the file system on the USB-drive, but that is all. Any further interaction and it is blue screen. Not sure if this may be file system type dependent, I haven’t tested that many different USB-drives.
Oddly enough, the install ISO was of course on a USB-drive.
As I am writing this, I am watching you @tobhe speaking at a Ubuntu conference (on YouTube). Thanks for all the work.
Ohh! That sucks. Well, I guess I haven’t stress tested my T14s that hard yet with other things (than USB-drive copying). Anything else that would be easy to test? That is, to see if the memory bug gets triggered.
Also, could you provide more details on how to perform the temporary fix you did?
Is there separate bug-tracking for display related things somewhere? Currently I have the following issues:
internal screen: no backlight control
one 4k screen: cannot boot with it plugged in (hang after grub); but plugging it in afterwards works. Sometimes the screen shows only 640x480, but unplugging and replugging fixes it.
another 4k screen: no problems booting with it or with resolutions.
third screen (3440x1440): if the screen turns off (e.g. after time in idle), it never turns back on. However, swaymsg -t get_outputs reports it as connected and even as focused when I move the mouse there. Unplugging and replugging the screen does not fix this. The only way to get the screen back on is killing the session (gdm does manage to bring it back online).
Apparently a tplg file is an ALSA Topology file, used by kernel drivers as a map between driver and different configurations of hardware. Or something like that.