Energy efficiency #3633
Energy efficiency #3633
Conversation
|
Again, some rough notes that you're free to pick from. It could be mentioned that the use of file descriptors to wait for events/data/input/etc... improves efficiency by allowing the CPU to idle to lower clock speeds and maybe turn off other cores (depends on other things running of course). Another thing that can be mentioned is that our usage of C++ is a deliberate choice to balance runtime performance and development speed versus other interpreted (Python) or compiled + GC langauges (Go, Java). |
| result, Mir has work in progress to exploit additional of hardware composition | ||
| options where they are available. | ||
|
|
||
| At the time of writing (October 2024) Mir uses hardware composition for cursors |
There was a problem hiding this comment.
Since we have versioned documentation now, does it make sense to say that? We should, instead, keep it up to date? And when we do, we can add a Since Mir… note?
| been update. That can avoid the need to regenerate parts of the display that | ||
| have not changed. | ||
|
|
||
| At the time of writing (October 2024) Mir has not implemented this optimisation. |
There was a problem hiding this comment.
AFAIU we do, just on the composition level (e.g. always update full surface, but only when it changed)?
There was a problem hiding this comment.
By "full surface" you mean display buffer?
There was a problem hiding this comment.
At least in the screencopy sense, we report on damage per client surface (not deeper than that).
Does damage even matter in the composition sense? We're asking the GPU to do the composition on client-supplied buffers, would we even communicate the damage there?
It would matter for a software renderer, but for a GL one?
There was a problem hiding this comment.
Yes, it matters: a while back Firefox was submitting buffers with damage hints, but only the pixels corresponding to the damage were supplied (which was a Firefox bug). Because we don't pass on the damage info the results were wrong. (This shows that other compositors do pass on the damage hints)
AlanGriffiths
force-pushed
the
MIRENG-505/energy-efficiency
branch
from
a7058cb
to
a3854c7
Compare
AlanGriffiths
changed the base branch from
main
to
MIRENG-504-Draft-the-performance-document
I've stacked this on top of |
AlanGriffiths
force-pushed
the
MIRENG-505/energy-efficiency
branch
from
a3854c7
to
1142fac
Compare
AlanGriffiths
force-pushed
the
MIRENG-505/energy-efficiency
branch
from
1142fac
to
ef50b0f
Compare
No description provided.