[Bug 105622] i915_init takes over 100 ms

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=105622

--- Comment #9 from Paul Menzel <pmenzel+bugs.freedesktop at molgen.mpg.de> ---
Created attachment 138291
  --> https://bugs.freedesktop.org/attachment.cgi?id=138291&action=edit
Screenshot from HTML output of `bootgraph.py`

(In reply to David Weinehall from comment #8)
> From past experience I'd say that initcall_debug alone isn't a very useful
> means to measure this. There's a tool called pm-graph that's great for
> things like this.
> 
> https://github.com/01org/pm-graph
> 
> With that (in ftrace mode) I'm seeing roughly 35ms on the Broadwell I tested
> on. I can test on KBL too, if you believe it'll make a difference (the BDW
> was simply closest at hand).

I know about pm-graph and the script `bootgraph.py` – you told me about the
script (new name) `sleepgraph.py` in another report, if I remember correctly.
;-)

Anyway, the times of initcall_debug and `bootgraph.py` match. This time I I
measured it both show 50 ms, which is still quite a lot.

>  i915_init: start=4474.445, end=4525.887, length(w/o overhead)=50.175 ms, return=0

These 50 ms are spent in `generic_hdmi_init` of `snd_hda_codec_hdmi`.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20180322/941b5a2e/attachment.html>