aboutsummaryrefslogtreecommitdiff
path: root/src/main.cpp
diff options
context:
space:
mode:
authordec05eba <dec05eba@protonmail.com>2024-10-16 01:07:41 +0200
committerdec05eba <dec05eba@protonmail.com>2024-10-16 01:07:58 +0200
commiteb602b9b531d58908e91a3e44a5004c935cc9a46 (patch)
tree225f3d2e739727c824e95a51cd7bc337fdfa6d51 /src/main.cpp
parent19d77b8a9660c31792afbeab229f66dc73e74df0 (diff)
4.2.2: AMD IS SHIT4.2.2
Diffstat (limited to 'src/main.cpp')
-rw-r--r--src/main.cpp2
1 files changed, 2 insertions, 0 deletions
diff --git a/src/main.cpp b/src/main.cpp
index 4706c9b..2073682 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -2611,6 +2611,8 @@ int main(int argc, char **argv) {
// The downside of this is that it always uses maximum power, which is not ideal for replay mode that runs on system startup.
// This option was added in mesa 24.1.4, released on july 17, 2024.
// TODO: Add an option to enable/disable this?
+ // Seems like the performance issue is not in encoding, but rendering the frame.
+ // Some frames end up taking 10 times longer. Seems to be an issue with amd gpu power management when letting the application sleep on the cpu side?
setenv("AMD_DEBUG", "lowlatencyenc", true);
// Some people set this to nvidia (for nvdec) or vdpau (for nvidia vdpau), which breaks gpu screen recorder since
// nvidia doesn't support vaapi and nvidia-vaapi-driver doesn't support encoding yet.