| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The cppc_cpudata per-cpu storage was inefficient (1) additional to causing
functional issues (2) when CPUs are hotplugged out, due to per-cpu data
being improperly initialised.
(1) The amount of information needed for CPPC performance control in its
cpufreq driver depends on the domain (PSD) coordination type:
ANY: One set of CPPC control and capability data (e.g desired
performance, highest/lowest performance, etc) applies to all
CPUs in the domain.
ALL: Same as ANY. To be noted that this type is not currently
supported. When supported, information about which CPUs
belong to a domain is needed in order for frequency change
requests to be sent to each of them.
HW: It's necessary to store CPPC control and capability
information for all the CPUs. HW will then coordinate the
performance state based on their limitations and requests.
NONE: Same as HW. No HW coordination is expected.
Despite this, the previous initialisation code would indiscriminately
allocate memory for all CPUs (all_cpu_data) and unnecessarily
duplicate performance capabilities and the domain sharing mask and type
for each possible CPU.
(2) With the current per-cpu structure, when having ANY coordination,
the cppc_cpudata cpu information is not initialised (will remain 0)
for all CPUs in a policy, other than policy->cpu. When policy->cpu is
hotplugged out, the driver will incorrectly use the uninitialised (0)
value of the other CPUs when making frequency changes. Additionally,
the previous values stored in the perf_ctrls.desired_perf will be
lost when policy->cpu changes.
Therefore replace the array of per cpu data with a list. The memory for
each structure is allocated at policy init, where a single structure
can be allocated per policy, not per cpu. In order to accommodate the
struct list_head node in the cppc_cpudata structure, the now unused cpu
and cur_policy variables are removed.
For example, on a arm64 Juno platform with 6 CPUs: (0, 1, 2, 3) in PSD1,
(4, 5) in PSD2 - ANY coordination, the memory allocation comparison shows:
Before patch:
- ANY coordination:
total slack req alloc/free caller
0 0 0 0/1 _kernel_size_le_hi32+0x0xffff800008ff7810
0 0 0 0/6 _kernel_size_le_hi32+0x0xffff800008ff7808
128 80 48 1/0 _kernel_size_le_hi32+0x0xffff800008ffc070
768 0 768 6/0 _kernel_size_le_hi32+0x0xffff800008ffc0e4
After patch:
- ANY coordination:
total slack req alloc/free caller
256 0 256 2/0 _kernel_size_le_hi32+0x0xffff800008fed410
0 0 0 0/2 _kernel_size_le_hi32+0x0xffff800008fed274
Additional notes:
- A pointer to the policy's cppc_cpudata is stored in policy->driver_data
- Driver registration is skipped if _CPC entries are not present.
Signed-off-by: Ionela Voinescu <[email protected]>
Tested-by: Mian Yousaf Kaukab <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Based on 1 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation version 2 of the license
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 315 file(s).
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Allison Randal <[email protected]>
Reviewed-by: Armijn Hemel <[email protected]>
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>
|
|
This patch add a helper to get the value of desired performance
register.
Signed-off-by: Xiongfeng Wang <[email protected]>
[ rjw: More white space ]
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
... with the goal of eventually enabling -Wmissing-prototypes by
default. At least on x86.
Make functions static where possible, otherwise add prototypes or make
them visible through includes.
asm/trace/ changes courtesy of Steven Rostedt <[email protected]>.
Signed-off-by: Borislav Petkov <[email protected]>
Reviewed-by: Masami Hiramatsu <[email protected]>
Reviewed-by: Ingo Molnar <[email protected]>
Acked-by: Rafael J. Wysocki <[email protected]> # ACPI + cpufreq bits
Cc: Andrew Banman <[email protected]>
Cc: Dimitri Sivanich <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Mike Travis <[email protected]>
Cc: "Steven Rostedt (VMware)" <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Yi Wang <[email protected]>
Cc: [email protected]
|
|
The Continuous Performance Control package may contain an optional
guaranteed performance field.
Add support to read guaranteed performance from _CPC.
Signed-off-by: Srinivas Pandruvada <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
CPPC V3 introduces two new entries to make it easier to convert between
abstract processor performance and frequency. The two new entries are
lowest frequency and nominal frequency. These are the frequencies
corresponding to lowest and nominal abstract performance.
Add support to read the new entries and populate them as part of the
CPPC performance capabilities which can be used by cpufreq drivers
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Computed delivered performance using CPPC feedback counters are in the
CPPC abstract scale, whereas cppc_cpufreq driver operates in KHz scale.
Exposing the CPPC performance capabilities (highest,lowest, nominal,
lowest non-linear) will allow userspace to figure out the conversion
factor from CPPC abstract scale to KHz.
Also rename ctr_wrap_time to wraparound_time so that show_cppc_data()
macro will work with it.
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Read lowest non linear perf in cppc_get_perf_caps so that it can be exposed
via sysfs to the usespace. Lowest non linear perf is the lowest performance
level at which nonlinear power savings are achieved.
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Since struct cpudata is defined in a header file, add prefix cppc_ to
make it not a generic name. Otherwise it causes compile issue in locally
define structure with the same name.
Signed-off-by: Srinivas Pandruvada <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
PCC status field exposes an error bit(2) to indicate any errors during
the execution of last comamnd. This patch checks the error bit before
notifying success/failure to the cpufreq driver.
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
The CPPC tables contain entries for per CPU feedback counters which
allows us to compute the delivered performance over a given interval
of time.
The math for delivered performance per the CPPCv5.0+ spec is:
reference perf * delta(delivered perf ctr)/delta(ref perf ctr)
Maintaining deltas of the counters in the kernel is messy, as it
depends on when the reads are triggered. (e.g. via the cpufreq
->get() interface). Also the ->get() interace only returns one
value, so cant return raw values. So instead, leave it to userspace
to keep track of raw values and do its math for CPUs it cares about.
delivered and reference perf counters are exposed via the same
sysfs file to avoid the potential "skid", if these values are read
individually from userspace.
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Ashwin Chaugule <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Compute the expected transition latency for frequency transitions
using the values from the PCCT tables when the desired perf
register is in PCC.
Signed-off-by: Prashanth Prakash <[email protected]>
Reviewed-by: Alexey Klimov <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
CPPC defined in section 8.4.7 of ACPI 6.0 specification suggests
"To amortize the cost of PCC transactions, OSPM should read or write
all PCC registers via a single read or write command when possible"
This patch enables opportunistic batching of frequency transition
requests whenever the request happen to overlap in time.
Currently the access to pcc is serialized by a spin lock which does
not scale well as we increase the number of cores in the system. This
patch improves the scalability by allowing the differnt CPU cores to
update PCC subspace in parallel and by batching requests which will
reduce the certain types of operation(checking command completion bit,
ringing doorbell) by a significant margin.
Profiling shows significant improvement in the overall effeciency
to service freq. transition requests. With this patch we observe close
to 30% of the frequency transition requests being batched with other
requests while running apache bench on a ARM platform with 6
independent domains(or sets of related cpus).
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
For cases where sys mapped CPC registers need to be accessed
frequently, it helps immensly to pre-map them rather than map
and unmap for each operation. e.g. case where feedback counters
are sys mem map registers.
Restructure cpc_read/write and the cpc_regs structure to allow
pre-mapping the system addresses and unmap them when the CPU exits.
Signed-off-by: Ashwin Chaugule <[email protected]>
Signed-off-by: Prashanth Prakash <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Exports pcc_mbox_request_channel() and pcc_mbox_free_channel()
declarations into a pcc.h header file.
Looks-good-to: Prashanth Prakash <[email protected]>
Signed-off-by: Hoan Tran <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Remove a redundant function declaration in cppc_acpi.h for
mbox_send_message(). That function is defined in mailbox_client.h,
which is already included.
Signed-off-by: Timur Tabi <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
CPPC stands for Collaborative Processor Performance Controls
and is defined in the ACPI v5.0+ spec. It describes CPU
performance controls on an abstract and continuous scale
allowing the platform (e.g. remote power processor) to flexibly
optimize CPU performance with its knowledge of power budgets
and other architecture specific knowledge.
This patch adds a shim which exports commonly used functions
to get and set CPPC specific controls for each CPU. This enables
CPUFreq drivers to gather per CPU performance data and use
with exisiting governors or even allows for customized governors
which are implemented inside CPUFreq drivers.
Signed-off-by: Ashwin Chaugule <[email protected]>
Reviewed-by: Al Stone <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|