k315rc7doug2_r1: CPU Frequency Verses load experiment: Powersave mode;
The intel_pstate code includes, sort of, Dirk's (private, not public):
Patch set 4 - patches 1, 3, and 5 of 5.
Patch set 3 - patch 6 of 8.
Patch set 2 - patch 6 of 6, with additional changes. Patch set 1 - patch 7 of 7, with additional changes, per above patch 6 of 6.
Note that for this experiment "perf record" data was acquired also.
Overview:
Detail A - Start of Experiment:
Detail B - End of Experiment:
. The CPU frequency is calm and low at the start of the experiment, and prior, as shown in detail A.
. The CPU frequency settles (fast) to a calm and low value at the end of the experiment as shown in detail B.
Load / sleep frequency sweep test - 85% average load with load / sleep frequency sweeping from 2 hertz to 250 hertz:
Conclusion: Looks good.
Other data:
The time difference between timer interrupt runs. (warning 1.2 megabytes, and my server is slow)
The time difference histogram file. (I haven't actually made the histogram graph yet.)
Another test was done, but the "perf record" data was acquired for all CPU's instead of just the one the test was for. There is some curious data, detailed in an e-mail, which needs to be investigated, understood and explained.
References:
Doug's (3.15RC7) intel_pstate.c 2.
Doug's (3.15RC7) power.h. (Actually Dirk's).
Script used to ramp cpu load.
Script used to ramp cpu sleep / load frequency.
Program used to create load ramp script.
Program used to create sleep / load frequency script.
Consume Program.
CPU frequency monitoring program (CPU 7).