On Tue, Nov 30, 2021 at 1:37 PM Huang Rui [off-list ref] wrote:

Hi all,

We would like to introduce a new AMD CPU frequency control mechanism as the
"amd-pstate" driver for modern AMD Zen based CPU series in Linux Kernel.
The new mechanism is based on Collaborative processor performance control
(CPPC) which is finer grain frequency management than legacy ACPI hardware
P-States. Current AMD CPU platforms are using the ACPI P-states driver to
manage CPU frequency and clocks with switching only in 3 P-states. AMD
P-States is to replace the ACPI P-states controls, allows a flexible,
low-latency interface for the Linux kernel to directly communicate the
performance hints to hardware.

"amd-pstate" leverages the Linux kernel governors such as *schedutil*,
*ondemand*, etc. to manage the performance hints which are provided by CPPC
hardware functionality. The first version for amd-pstate is to support one
of the Zen3 processors, and we will support more in future after we verify
the hardware and SBIOS functionalities.

There are two types of hardware implementations for amd-pstate: one is full
MSR support and another is shared memory support. It can use
X86_FEATURE_CPPC feature flag to distinguish the different types.

Using the new AMD P-States method + kernel governors (*schedutil*,
*ondemand*, ...) to manage the frequency update is the most appropriate
bridge between AMD Zen based hardware processor and Linux kernel, the
processor is able to adjust to the most efficiency frequency according to
the kernel scheduler loading.

Please check the detailed CPU feature and MSR register description in
Processor Programming Reference (PPR) for AMD Family 19h Model 51h,
Revision A1 Processors:

https://www.amd.com/system/files/TechDocs/56569-A1-PUB.zip

Performance Per Watt (PPW) Calculation:

We use the RAPL interface with "perf" tool to get the energy data of the
package power.

The data comparisons between amd-pstate and acpi-freq module are tested on
AMD Cezanne processor (mobile CPU):

1) TBench CPU benchmark:

+----------------------------------------------------------------------------------------------+
|                                                                                              |
|                               TBench4 (Performance Per Watt)                                 |
|                                                                             Higher is better |
+-------------------+------------------------+------------------------+------------------------+
|                   |  Performance Per Watt  |  Performance Per Watt  |  Performance Per Watt  |
|   Kernel Module   |       (Schedutil)      |       (Ondemand)       |       (Performance)    |
|                   |       Unit: MB / J     |       Unit: MB / J     |       Unit: MB / J     |
+-------------------+------------------------+------------------------+------------------------+
|                   |                        |                        |                        |
|    acpi-cpufreq   |         48.56          |        48.89           |        47.81           |
|                   |                        |                        |                        |
+-------------------+------------------------+------------------------+------------------------+
|                   |                        |                        |                        |
|     amd-pstate    |         48.38          |        47.38           |        48.77           |
|                   |                        |                        |                        |
+-------------------+------------------------+------------------------+------------------------+

Note: The previous data was based on TBench2, as align with Suse, we use
TBench4 to re-test it. The PPW is very closed to acpi-cpufreq. And we are
still re-runing other tests.

Steam Game Demo on Ryzen 5900X (12 core 24 threads):

The picture to compare acpi-cpufreq vs amd-pstate:
https://drive.google.com/file/d/1PvSduykJn9U5MMOhzFWycnbmGmznalM3/view?usp=sharing

Two videos:
https://drive.google.com/file/d/1nQQEteL-v_zQxnOJpyW8JqvRW2FFDN2Z/view?usp=sharing
https://drive.google.com/file/d/1heuPgFG71SQHvGb6wfedrQciBfE2rhnu/view?usp=sharing

Actually, the amd-pstate driver doesn't change the physical maximum
frequency capacity in the processor. But it's able to provide the finer
grain performance control range instead of legacy 3 P-States. It has a
better performance and power efficiency than before. We will continue
optimize amd-pstate function on kernel governors to support different types
of processors such as mobile latop, performance desktop, and etc.

See patch series in below git repo:
V1: https://git.kernel.org/pub/scm/linux/kernel/git/rui/linux.git/log/?h=amd-pstate-dev-v1
V2: https://git.kernel.org/pub/scm/linux/kernel/git/rui/linux.git/log/?h=amd-pstate-dev-v2
V3: https://git.kernel.org/pub/scm/linux/kernel/git/rui/linux.git/log/?h=amd-pstate-dev-v3
V4: https://git.kernel.org/pub/scm/linux/kernel/git/rui/linux.git/log/?h=amd-pstate-dev-v4
V5: https://git.kernel.org/pub/scm/linux/kernel/git/rui/linux.git/log/?h=amd-pstate-dev-v5

For details introduction, please see the patch 22.

After going through all of the series I'm still unsure what the
freq_qos requests added in patch [06/22] are for.

The cpupower patches need ACKs from Shuah for me to take them.

The tracing patch needs an ACK from Steve Rostedt.

Thanks!