Radeon DRM Gets New Information Ioctl Queries

[phoronix]

Phoronix: Radeon DRM Gets New Information Ioctl Queries

For those using the Gallium3D HUD, RadeonTOP, or other utilities, more data is being exposed with new patches for the Radeon DRM driver...

http://www.phoronix.com/vr.php?view=MTgwMTk

[agd5f]

I don't think there is a good standard cross-vendor way to expose this information. It's all VERY asic specific. Making it generic enough to support multiple vendors would really reduce the usefulness signficantly. Each vendor has different clock trees and voltage domains as well as different blocks that make up their GPU.

[dungeon]

Alex, what about those temperatures for APUs, i guess they are false and should not be exposed? At least for me sensors for radeon says 8.0?C here on Kabini

And also k10temp gives false reading, similar to that.

I use separete it87 it8603 for temperature reading, which is fine for the APU.

[curaga]

Two years and four months later, it's happening :P

Anyway I fully approve, and if I was still receiving dri-devel I'd post a reviewed-by. One thing, you mention a harvest register, but on quick gleaning I couldn't find any?

[agd5f]

Alex, what about those temperatures for APUs, i guess they are false and should not be exposed? At least for me sensors for radeon says 8.0?C here on Kabini

And also k10temp gives false reading, similar to that.

I use separete it87 it8603 for temperature reading, which is fine for the APU.

I don't think the OEMs always set up the internal thermal sensor properly on APUs. They often use OEM specific sensors which seems to be the case on your board.

[agd5f]

Two years and four months later, it's happening :P

Anyway I fully approve, and if I was still receiving dri-devel I'd post a reviewed-by. One thing, you mention a harvest register, but on quick gleaning I couldn't find any?

I didn't export any yet. It's a possible future use. We currently export the harvest information separately via the info ioctl.

[dungeon]

I don't think the OEMs always set up the internal thermal sensor properly on APUs. They often use OEM specific sensors which seems to be the case on your board.

Ah probably who knows , i tought to ask also why power reading does not work but read this now:

http://marc.info/?l=linux-kernel&m=141038422917691&w=2

I hoped they will make it, but it will be N/A . My gosh, how on Windows all that works , maybe i will read that BKDG to see what the hell happens there

[flim]

Alex, what about those temperatures for APUs, i guess they are false and should not be exposed? At least for me sensors for radeon says 8.0?C here on Kabini

And also k10temp gives false reading, similar to that.

I use separete it87 it8603 for temperature reading, which is fine for the APU.

Might be related: https://www.kernel.org/doc/Documentation/hwmon/k10temp

[jrch2k8]

Ah probably who knows , i tought to ask also why power reading does not work but read this now:

http://marc.info/?l=linux-kernel&m=141038422917691&w=2

I hoped they will make it, but it will be N/A . My gosh, how on Windows all that works , maybe i will read that BKDG to see what the hell happens there

well in windows happens too, if you don't install the drivers from the mobo manufacturer page. I had over time plenty of hardware(raid controllers, wifi cards, motherboards chipset from AMD/Intel, rs-232 controllers, etc.) that crashed windows / server using the chip maker driver instead the crappy ones in the motherboards site(lessons learned never use inboard SAS raid controllers if you don't want nasty data corruption from driver upgrades).

the main difference between windows and linux in this aspect, is that manufacturers include their crappy modifications in their crappy drivers where in linux devs have to figure out what ugly chinese 2 cents knock off they used outside the vendor specs

[dungeon]

well in windows happens too, if you don't install the drivers from the mobo manufacturer page. I had over time plenty of hardware(raid controllers, wifi cards, motherboards chipset from AMD/Intel, rs-232 controllers, etc.) that crashed windows / server using the chip maker driver instead the crappy ones in the motherboards site(lessons learned never use inboard SAS raid controllers if you don't want nasty data corruption from driver upgrades).

the main difference between windows and linux in this aspect, is that manufacturers include their crappy modifications in their crappy drivers where in linux devs have to figure out what ugly chinese 2 cents knock off they used outside the vendor specs

Thing is it87 sensor works for me, which one user write thanks to information he got from Asus, so nothing to figure out. But those plain AMD's neither work, radeon, k10temp and fam15h_power. And there are planty of documentation available But i am happier with Asus because on Gigabyte AM1 boards guess what - nothing works fine

Also it is not one OEM but two: Asus mobo and Acer laptop, both with 16h family and i see same sensor brokeness on linux.

But with it87 i am fine, only if they did not ditched 16h from fam15h_power (what the hell is N/A ) and if radeon allow sensor to be disabled in kernel config