08:24pepp: Venemo: I use https://lore.kernel.org/amd-gfx. eg https://lore.kernel.org/amd-gfx/20250204143113.6320-1-christian.koenig@amd.com/T/#
08:27pepp: for each patch there's a 'raw' link to download the email and you can then use "git am xxx" to apply iy
10:26Venemo: pepp: how do you find anything in there?
10:27Venemo: that is not the series I was looking for
10:30Venemo: okay, I think I found what I was looking for
10:30Venemo: what a tedious process
10:34Venemo: agd5f: the package power that is reported by amdgpu is definitely broken on Raphael. it says that it consumes about 10 W while the CPU is compiling the kernel on 32 threads
10:39Venemo: in fact it is reporting fewer watts consumed now than during idle
12:36tonyk: Venemo: you can use `b4 shazam <patch id>`
12:36tonyk: https://b4.docs.kernel.org/en/latest/maintainer/am-shazam.html
12:36tonyk: it will download the patch and apply to your tree
12:46Venemo: thanks tonyk
13:19Venemo: tonyk: how do you know the patch id?
13:20tonyk: Venemo: what patch to you want to apply?
13:21Venemo: the "Rework and fix queue reset for gfx7-gfx10" series from Christian
13:21tonyk: did you got it in your email? or someone sent you a link for it?
13:23Venemo: tonyk: https://lore.kernel.org/amd-gfx/ -> "Rework and fix queue reset for gfx7-gfx10"
13:24tonyk: ok
13:24Venemo: or https://lists.freedesktop.org/archives/amd-gfx/2025-February/119508.html
13:25tonyk: b4 accepts lore.kernel links, so you can use `b4 shazam https://lore.kernel.org/all/20250204143113.6320-1-christian.koenig@amd.com/`
13:25Venemo: does that apply the whole series or just a specific patch?
13:25tonyk: the message id is the last part of the link, so you can do also `b4 shazam 20250204143113.6320-1-christian.koenig@amd.com`
13:25tonyk: the whole seires
13:25Venemo: nice! thanks
13:26tonyk: you can use`b4 am 20250204143113.6320-1-christian.koenig@amd.com` to download the patches, and apply then individually
13:26tonyk: and if someone sends you a patch via email, your client has an option somewhere to show the message id
13:27Venemo: awesome
13:44tonyk: hwentlan_ agd5f: could you have a look on this new version for the async page flip? :) https://lore.kernel.org/lkml/20250127-tonyk-async_flip-v12-2-0f7f8a8610d3@igalia.com/
14:45MrCooper: lileo hwentlan_: out of curiosity, when using the KMS legacy cursor ioctl, can DC move the cursor ~anytime outside of vblank, or does it just use the latest position for the next refresh?
14:50Venemo: agd5f: is the series to enable the i2c buses supposed to be included and working in 6.14-rc1?
14:51agd5f: for RGB stuff? that will be in 6.15
14:55Venemo: yes, I didn't realize that
14:56Venemo: I thought because the patch was written a month ago, it would fit into 6.14
15:25Venemo: how is it decided what goes into which kernel release?
15:37agd5f: Venemo, most subsystems cut off new features around rc-5 or 6.
15:43Venemo: but 6.14 is only at rc1, and was not even there when the patch was written
15:43Venemo: or do you mean that they cut it off at rc-5 or 6 of the previous version?
15:45agd5f: yeah, previous version. E.g., feature cut off for 6.14 is ~6.13-rc6 timeframe
15:48Venemo: ah, I see
15:48Venemo: I guess except for bug fixes?
15:49agd5f: right
15:51Venemo: who decides whether something is a fix or a feature? eg. would the improved GPU reset be considered a fix, since it fixes a real problem (will fix a real problem once it's ready)?
15:54tonyk: I think usually the fixes means "patches that fixes things added during the merge window" or previous bugs
15:55tonyk: improvements are not considered fixes in most cases
16:59Venemo: tonyk: well, I guess that begs the question of what is a bug. I think bringing down the whole system for a GPU hang is a bug
17:08Ristovski: Venemo: do note that on APUs, the reported GPU power is actually the SoC power afaik
17:23DemiMarie: tonyk: I agree with Venemo on this.
17:31Venemo: Ristovski: yes, I want to see the SoC power, but it's showing me a number that makes no sense
17:32Venemo: Ristovski: or are you telling me that my 7950X can compile a kernel on 32 threads with only 10W?
18:54Venemo:
21:36johnny0: Venemo: your issue was noted near the 7950x's launch: https://forum.level1techs.com/t/ryzen-7000-linux-idle-power-whats-normal-high-cpu-ppt/190273/16
21:38johnny0: fwiw, single ccd zen4 chips do seem to report as expected
21:53agd5f: FWIW, I filed a ticket with the PMFW team. Will let you know what I hear back