Quick strace after updating to the 570.153.02 Production Branch driver.
Looks like the /proc/stat issue #5238502 may have been addressed?
[pid 5391] futex(0x7a3318000f50, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1747773368, tv_nsec=904899000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid 5396] <... futex resumed>) = -1 ETIMEDOUT (Connection timed out)
[pid 5396] futex(0x5c62a658a220, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 5396] futex(0x7a330c000f50, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1747773368, tv_nsec=815262000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid 5411] <... futex resumed>) = -1 ETIMEDOUT (Connection timed out)
[pid 5411] futex(0x5c62a7337400, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 5411] futex(0x7a32cc000f50, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1747773368, tv_nsec=824794000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid 5396] <... futex resumed>) = -1 ETIMEDOUT (Connection timed out)
[pid 5396] futex(0x5c62a658a220, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 5396] futex(0x7a330c000f50, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1747773368, tv_nsec=825372000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid 5411] <... futex resumed>) = -1 ETIMEDOUT (Connection timed out)
[pid 5411] futex(0x5c62a7337400, FUTEX_WAKE_PRIVATE, 1) = 0
Futexes still spinning, but CPU time consumed is improved slightly.
I’ll analyze and report accurate measurements over a longer time period.
- kernel 6.14.8
- gnome-shell 48.1 (Wayland)
- mutter 48.2 (-Dxwayland=false -Dx11=false)
- gtk4 4.18.5 (-Dx11-backend=false)
- nvidia 570.153.02
EDIT: Fresh boot with 6.14.8 kernel and after 6hrs uptime the CPU usage is very similar to that described in comment #23.
So, not a statistically significant improvement, but after checking the strace of several GTK4+Vulkan apps, the openat.../proc/stat calls seem to be no longer occurring.