Experimental Schedutil Patches Yield 30% Boost To Web Browser Benchmark On Linux
([Linux Kernel] 2 Hours Ago
Schedutil Improvements)
- Reference: 0001486834
- News link: https://www.phoronix.com/news/Schedutil-30p-Speedometer-Boost
- Source link:
Google engineer Qais Yousef has posted a set of 16 patches for the " [1]Schedutil " scheduler utilization code within the Linux kernel to better manage system response time. Schedutil is often used by default on many Linux distributions and with these patches a popular web browser benchmark can be as much as 30% faster with these kernel patches.
Qais Yousef posted the set of 16 patches on Tuesday as a new take on a patch series he posted last year to remove hardcoded margings within the CPUFreq scheduler code.
The patch cover letter to this new series [2][RFC PATCH 00/16] sched/fair/schedutil: Better manage system response time goes into the technical details on all of the low-level work involved for better managing the system response time within the Schedutil code.
For end-users the exciting aspect is the Speedometer web browser benchmark result shared:
Starting out at 352 runs per minute with the Speedometer scheduler benchmark and ending with these patch series and a ramp-up multiplier of four to yield a score of 456 runs per minute, a 29.5% boost. That was the only benchmark provided at this time.
If these patches move along and are on a trajectory for mainlining in the Linux kernel, it will be very interesting to test these patches with a more diverse set of workloads in looking at the power and performance.
[1] https://www.phoronix.com/search/Schedutil
[2] https://lore.kernel.org/linux-pm/20240820163512.1096301-1-qyousef@layalina.io/
Qais Yousef posted the set of 16 patches on Tuesday as a new take on a patch series he posted last year to remove hardcoded margings within the CPUFreq scheduler code.
The patch cover letter to this new series [2][RFC PATCH 00/16] sched/fair/schedutil: Better manage system response time goes into the technical details on all of the low-level work involved for better managing the system response time within the Schedutil code.
For end-users the exciting aspect is the Speedometer web browser benchmark result shared:
Starting out at 352 runs per minute with the Speedometer scheduler benchmark and ending with these patch series and a ramp-up multiplier of four to yield a score of 456 runs per minute, a 29.5% boost. That was the only benchmark provided at this time.
If these patches move along and are on a trajectory for mainlining in the Linux kernel, it will be very interesting to test these patches with a more diverse set of workloads in looking at the power and performance.
[1] https://www.phoronix.com/search/Schedutil
[2] https://lore.kernel.org/linux-pm/20240820163512.1096301-1-qyousef@layalina.io/
Quackdoc