Fall back to primary gpu for rendering #1281
Conversation
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
from
dd7e90d to
5e37e19
Compare
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
from
5e37e19 to
f2be61c
Compare
|
doc builds fail caused by ongoing migration of fdo gitlab |
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
2 times, most recently
from
1f85ad6 to
6083bb4
Compare
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
4 times, most recently
from
c4a170b to
6571e85
Compare
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
2 times, most recently
from
43c9e7f to
d0e6120
Compare
cmeissl
force-pushed
the
feature/primary_render_fallback
branch
from
d0e6120 to
2d2767d
Compare
| .map(|surface_render_node| surface_render_node == primary_render_node) | ||
| .unwrap_or(false) |
| ) | ||
| ); | ||
|
|
||
| !is_ccs && (device.render_node.is_some() || format.modifier == Modifier::Linear) |
There was a problem hiding this comment.
nit: would be clearer to write this as an
if device.render_node.is_none() {
return format.modifier == Modifier::Linear;
}at the top of the closure
| .as_mut() | ||
| .add_node(render_node, gbm.clone()) | ||
| .context("error adding render node to GPU manager")?; | ||
| anyhow::ensure!(!egl_device.is_software(), "skipping software device"); |
There was a problem hiding this comment.
Question: will this break running niri on devices where only llvmpipe is available? E.g. the main GPU driver has too old OpenGL.
(I have no idea if this even works right now, maybe it doesn't, then it's fine, but should add a comment here)
src/backend/tty.rs
Outdated
| // Create GBM allocator. | ||
| let gbm_flags = GbmBufferFlags::RENDERING | GbmBufferFlags::SCANOUT; | ||
| let allocator = GbmAllocator::new(device.gbm.clone(), gbm_flags); | ||
|
|
There was a problem hiding this comment.
nit: if you feel like, could split "putting allocator into OutputDevice" into its own commit prior to this commit
| .map(|render_node| render_node == self.primary_render_node) | ||
| .unwrap_or(false); |
| output_state | ||
| .frame_clock | ||
| .refresh_interval() |
There was a problem hiding this comment.
nit: might be cleaner to extract if let Some(refresh_interval) = output_state.frame_clock.refresh_interval() { outside this whole vblank_remaining_time handling section
| }; | ||
| let throttled_metadata = DrmEventMetadata { | ||
| sequence: meta.sequence, | ||
| time: DrmEventTime::Monotonic(throttled_time), |
There was a problem hiding this comment.
Wouldn't it be better to set DrmEventTime::Monotonic(Duration::ZERO) here? So that it doesn't get set as as HwClock presentation.
|
Merged the Smithay bump commit |
|
This can be closed now that #2312 is merged |
|
superseded by #2312 @scottmckendry Thanks for taking over, nice work! |
3 participants
depends on
Smithay/smithay#1680(merged)fixes #910