Noticeable latency on Asahi Linux #13734
Comments
jvoisin
added
admin read
This comment was marked as off-topic.
This comment was marked as off-topic.
|
Just to be clear, #13203 is about Zed being “half the time just as slow as VS Code, and sometimes 1 frame slower”. In other words, it's pretty fast. If you feel that Zed is truly lagging, then it's possibly a different issue. Are you sure that you're able to run Vulkan apps? I see: Doesn't that mean that you're using a software renderer? What happens when you run vkcube or vkgears? @mrnugget I don't know who else to ping, but this is this issue again: #13154. Even though Zed is not wrong in this case, and it's probably a good idea to fall back to a software renderer, I'm now 100% convinced that Zed should show a warning to the user when that happens. Otherwise people start assuming that zed is lagging, while they're simply using the software renderer. |
|
How to detect software rendering: |
|
Asahi doesn’t have a public Vulkan driver so you are almost certainly running llvmpipe. Try using the OpenGL backend: #13305. There’s a chance it will not work because it’s very new and not many people use it. |
|
I'm indeed running llmvpipe: $ vulkaninfo | grep -i llvmpipe
GPU id = 0 (llvmpipe (LLVM 18.1.6, 128 bits))
GPU id : 0 (llvmpipe (LLVM 18.1.6, 128 bits)):
GPU id : 0 (llvmpipe (LLVM 18.1.6, 128 bits)):
deviceName = llvmpipe (LLVM 18.1.6, 128 bits)
driverID = DRIVER_ID_MESA_LLVMPIPE
driverName = llvmpipe
$Having a warning about this would be nice indeed. On the bright side, @alyssarosenzweig has a working vulkan implementation that should land into Asahi Linux at some point. |
notpeter
added
performance
|
Zed works well on Asahi Linux when compiling with |
|
@lukaslihotzki Thanks for the hint! I am seeing the following error on Asahi Linux: It seems that transparency is not working correctly. With the following changes, the error disappears and Zed is fully functional: diff --git a/crates/gpui/src/platform/blade/blade_renderer.rs b/crates/gpui/src/platform/blade/blade_renderer.rs
index 1830d0cd0..06e1d62cb 100644
--- a/crates/gpui/src/platform/blade/blade_renderer.rs
+++ b/crates/gpui/src/platform/blade/blade_renderer.rs
@@ -365,7 +365,7 @@ impl BladeRenderer {
display_sync: gpu::DisplaySync::Recent,
color_space: gpu::ColorSpace::Linear,
allow_exclusive_full_screen: false,
- transparent: config.transparent,
+ transparent: false,
};
let surface_info = gpu.resize(surface_config);
@@ -431,15 +431,7 @@ impl BladeRenderer {
}
}
- pub fn update_transparency(&mut self, transparent: bool) {
- if transparent != self.surface_config.transparent {
- self.wait_for_gpu();
- self.surface_config.transparent = transparent;
- let surface_info = self.gpu.resize(self.surface_config);
- self.pipelines = BladePipelines::new(&self.gpu, surface_info);
- self.alpha_mode = surface_info.alpha;
- }
- }
+ pub fn update_transparency(&mut self, transparent: bool) {}
#[cfg_attr(target_os = "macos", allow(dead_code))]
pub fn viewport_size(&self) -> gpu::Extent { |
|
The Honeykrisp driver (Vulkan driver for Asahi Linux) is now released for general availability, @jvoisin can you test again if you have this issue? For me, it is working perfectly. |
|
I'm on Asahi too and have been playing with Zed for the first time since the Honeykrisp release yesterday. Everything seems smoooth 🥳 |
|
Yup, it's working well now :) |
Check for existing issues
Describe the bug / provide steps to reproduce it
curl https://zed.dev/install.sh | bashI asked a couple of non-tech people "Does the movement seem more sluggish in this window (Zed) or this window (neovim in a terminal)?" and all of them said that the neovim window had instantaneous cursor movement, while the Zed one was a bit lagging.
Environment
Zed: v0.142.3 (Zed Preview)
OS: Linux Wayland fedora-asahi-remix 40
Memory: 23.2 GiB
Architecture: aarch64
If applicable, add mockups / screenshots to help explain present your vision of the feature
No response
If applicable, attach your
~/Library/Logs/Zed/Zed.logfile to this issue.The text was updated successfully, but these errors were encountered: