Mesa Fixes A Large Performance Regression For Systems Using HPET

[phoronix]

Phoronix: Mesa Fixes A Large Performance Regression For Systems Using HPET

The High Precision Event Timer (HPET) has long been a source of issues for Linux developers and it turns out systems relying on HPET rather than the CPU's TSC have in recent months suffered significant performance degradation with the Mesa OpenGL driver code...

https://www.phoronix.com/news/Mesa-Fixes-HPET-Regression

[darkbasic]

niko@talos2 ~ $ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
timebase​

What the hell is timebase?

[nuetzel]

niko@talos2 ~ $ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
timebase​

What the hell is timebase?

No clue.

/home/dieter> cat /sys/devices/system/clocksource/clocksource0/available_clocksource
tsc hpet acpi_pm
/home/dieter> cat /sys/devices/system/clocksource/clocksource0/current_clocksource
tsc

[hiryu]

niko@talos2 ~ $ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
timebase​

What the hell is timebase?

Ahahaha! I get the same on my Talos II. Glad to see I'm not the only who checked.

[V1tol]

/home/dieter> cat /sys/devices/system/clocksource/clocksource0/available_clocksource
tsc hpet acpi_pm
/home/dieter> cat /sys/devices/system/clocksource/clocksource0/current_clocksource
tsc

Same on mine. That's interesting. I thought my 5950X should definitely use HPET and I don't remember tinkering anything in UEFI related to that.

[nuetzel]

Same on mine. That's interesting. I thought my 5950X should definitely use HPET and I don't remember tinkering anything in UEFI related to that.

I have HPET in for ages - old Nehalem currently.
But...
...so I didn't noticed and haven't pinged Marek...

[ 0.014201] ACPI: HPET 0x000000007F7BBEE4 000038 (v01 PTLTD HPETTBL 00060000 LTP 00000001)
[ 0.014244] ACPI: Reserving HPET table memory at [mem 0x7f7bbee4-0x7f7bbf1b]
[ 0.021434] ACPI: HPET id: 0x8086a701 base: 0xfed00000

[marek]

TSC is fast. HPET is crazy slow and is likely used on old CPUs where TSC is unusable because each core probably increments it at a slightly different pace. I don't know about other clock sources.

[nuetzel]

Thanks Marek!

Even for this one:
mesa: fix 38% decrease in display list performance of Viewperf2020/NX8_StudioAA

https://cgit.freedesktop.org/mesa/mesa/commit/?id=c97961a855ad1ae5c58d47634862b459e682e158

Michael

[jntesteves]

TSC is fast. HPET is crazy slow and is likely used on old CPUs where TSC is unusable because each core probably increments it at a slightly different pace. I don't know about other clock sources.

Should AMD's Ryzen 9 5900HX be included in "old CPUs" here? I currently get:

❯ uname -a
Linux 6.3.12-200.fc38.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Jul  6 04:05:18 UTC 2023 x86_64 GNU/Linux​
❯ cat /sys/devices/system/clocksource/clocksource0/available_clocksource
hpet acpi_pm​
❯ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
hpet​

I noticed recently my performance took a nosedive in F1 22 (on VKD3D-proton/Vulkan) and the game is now unplayable. Wondering if this could be the cause. I'll have to test older Kernels.

[Vistaus]

Same on mine. That's interesting. I thought my 5950X should definitely use HPET and I don't remember tinkering anything in UEFI related to that.

That's odd. My 5900HX uses TSC.