1 .. SPDX-License-Identifier: GPL-2.0 2 3 ======================================================= 4 Legacy Documentation of CPU Performance Scaling Drivers 5 ======================================================= 6 7 Included below are historic documents describing assorted 8 :doc:`CPU performance scaling <cpufreq>` drivers. They are reproduced verbatim, 9 with the original white space formatting and indentation preserved, except for 10 the added leading space character in every line of text. 11 12 13 AMD PowerNow! Drivers 14 ===================== 15 16 :: 17 18 PowerNow! and Cool'n'Quiet are AMD names for frequency 19 management capabilities in AMD processors. As the hardware 20 implementation changes in new generations of the processors, 21 there is a different cpu-freq driver for each generation. 22 23 Note that the driver's will not load on the "wrong" hardware, 24 so it is safe to try each driver in turn when in doubt as to 25 which is the correct driver. 26 27 Note that the functionality to change frequency (and voltage) 28 is not available in all processors. The drivers will refuse 29 to load on processors without this capability. The capability 30 is detected with the cpuid instruction. 31 32 The drivers use BIOS supplied tables to obtain frequency and 33 voltage information appropriate for a particular platform. 34 Frequency transitions will be unavailable if the BIOS does 35 not supply these tables. 36 37 6th Generation: powernow-k6 38 39 7th Generation: powernow-k7: Athlon, Duron, Geode. 40 41 8th Generation: powernow-k8: Athlon, Athlon 64, Opteron, Sempron. 42 Documentation on this functionality in 8th generation processors 43 is available in the "BIOS and Kernel Developer's Guide", publication 44 26094, in chapter 9, available for download from www.amd.com. 45 46 BIOS supplied data, for powernow-k7 and for powernow-k8, may be 47 from either the PSB table or from ACPI objects. The ACPI support 48 is only available if the kernel config sets CONFIG_ACPI_PROCESSOR. 49 The powernow-k8 driver will attempt to use ACPI if so configured, 50 and fall back to PST if that fails. 51 The powernow-k7 driver will try to use the PSB support first, and 52 fall back to ACPI if the PSB support fails. A module parameter, 53 acpi_force, is provided to force ACPI support to be used instead 54 of PSB support. 55 56 57 ``cpufreq-nforce2`` 58 =================== 59 60 :: 61 62 The cpufreq-nforce2 driver changes the FSB on nVidia nForce2 platforms. 63 64 This works better than on other platforms, because the FSB of the CPU 65 can be controlled independently from the PCI/AGP clock. 66 67 The module has two options: 68 69 fid: multiplier * 10 (for example 8.5 = 85) 70 min_fsb: minimum FSB 71 72 If not set, fid is calculated from the current CPU speed and the FSB. 73 min_fsb defaults to FSB at boot time - 50 MHz. 74 75 IMPORTANT: The available range is limited downwards! 76 Also the minimum available FSB can differ, for systems 77 booting with 200 MHz, 150 should always work. 78 79 80 ``pcc-cpufreq`` 81 =============== 82 83 :: 84 85 /* 86 * pcc-cpufreq.txt - PCC interface documentation 87 * 88 * Copyright (C) 2009 Red Hat, Matthew Garrett <mjg@redhat.com> 89 * Copyright (C) 2009 Hewlett-Packard Development Company, L.P. 90 * Nagananda Chumbalkar <nagananda.chumbalkar@hp.com> 91 */ 92 93 94 Processor Clocking Control Driver 95 --------------------------------- 96 97 Contents: 98 --------- 99 1. Introduction 100 1.1 PCC interface 101 1.1.1 Get Average Frequency 102 1.1.2 Set Desired Frequency 103 1.2 Platforms affected 104 2. Driver and /sys details 105 2.1 scaling_available_frequencies 106 2.2 cpuinfo_transition_latency 107 2.3 cpuinfo_cur_freq 108 2.4 related_cpus 109 3. Caveats 110 111 1. Introduction: 112 ---------------- 113 Processor Clocking Control (PCC) is an interface between the platform 114 firmware and OSPM. It is a mechanism for coordinating processor 115 performance (ie: frequency) between the platform firmware and the OS. 116 117 The PCC driver (pcc-cpufreq) allows OSPM to take advantage of the PCC 118 interface. 119 120 OS utilizes the PCC interface to inform platform firmware what frequency the 121 OS wants for a logical processor. The platform firmware attempts to achieve 122 the requested frequency. If the request for the target frequency could not be 123 satisfied by platform firmware, then it usually means that power budget 124 conditions are in place, and "power capping" is taking place. 125 126 1.1 PCC interface: 127 ------------------ 128 The complete PCC specification is available here: 129 https://acpica.org/sites/acpica/files/Processor-Clocking-Control-v1p0.pdf 130 131 PCC relies on a shared memory region that provides a channel for communication 132 between the OS and platform firmware. PCC also implements a "doorbell" that 133 is used by the OS to inform the platform firmware that a command has been 134 sent. 135 136 The ACPI PCCH() method is used to discover the location of the PCC shared 137 memory region. The shared memory region header contains the "command" and 138 "status" interface. PCCH() also contains details on how to access the platform 139 doorbell. 140 141 The following commands are supported by the PCC interface: 142 * Get Average Frequency 143 * Set Desired Frequency 144 145 The ACPI PCCP() method is implemented for each logical processor and is 146 used to discover the offsets for the input and output buffers in the shared 147 memory region. 148 149 When PCC mode is enabled, the platform will not expose processor performance 150 or throttle states (_PSS, _TSS and related ACPI objects) to OSPM. Therefore, 151 the native P-state driver (such as acpi-cpufreq for Intel, powernow-k8 for 152 AMD) will not load. 153 154 However, OSPM remains in control of policy. The governor (eg: "ondemand") 155 computes the required performance for each processor based on server workload. 156 The PCC driver fills in the command interface, and the input buffer and 157 communicates the request to the platform firmware. The platform firmware is 158 responsible for delivering the requested performance. 159 160 Each PCC command is "global" in scope and can affect all the logical CPUs in 161 the system. Therefore, PCC is capable of performing "group" updates. With PCC 162 the OS is capable of getting/setting the frequency of all the logical CPUs in 163 the system with a single call to the BIOS. 164 165 1.1.1 Get Average Frequency: 166 ---------------------------- 167 This command is used by the OSPM to query the running frequency of the 168 processor since the last time this command was completed. The output buffer 169 indicates the average unhalted frequency of the logical processor expressed as 170 a percentage of the nominal (ie: maximum) CPU frequency. The output buffer 171 also signifies if the CPU frequency is limited by a power budget condition. 172 173 1.1.2 Set Desired Frequency: 174 ---------------------------- 175 This command is used by the OSPM to communicate to the platform firmware the 176 desired frequency for a logical processor. The output buffer is currently 177 ignored by OSPM. The next invocation of "Get Average Frequency" will inform 178 OSPM if the desired frequency was achieved or not. 179 180 1.2 Platforms affected: 181 ----------------------- 182 The PCC driver will load on any system where the platform firmware: 183 * supports the PCC interface, and the associated PCCH() and PCCP() methods 184 * assumes responsibility for managing the hardware clocking controls in order 185 to deliver the requested processor performance 186 187 Currently, certain HP ProLiant platforms implement the PCC interface. On those 188 platforms PCC is the "default" choice. 189 190 However, it is possible to disable this interface via a BIOS setting. In 191 such an instance, as is also the case on platforms where the PCC interface 192 is not implemented, the PCC driver will fail to load silently. 193 194 2. Driver and /sys details: 195 --------------------------- 196 When the driver loads, it merely prints the lowest and the highest CPU 197 frequencies supported by the platform firmware. 198 199 The PCC driver loads with a message such as: 200 pcc-cpufreq: (v1.00.00) driver loaded with frequency limits: 1600 MHz, 2933 201 MHz 202 203 This means that the OPSM can request the CPU to run at any frequency in 204 between the limits (1600 MHz, and 2933 MHz) specified in the message. 205 206 Internally, there is no need for the driver to convert the "target" frequency 207 to a corresponding P-state. 208 209 The VERSION number for the driver will be of the format v.xy.ab. 210 eg: 1.00.02 211 ----- -- 212 | | 213 | -- this will increase with bug fixes/enhancements to the driver 214 |-- this is the version of the PCC specification the driver adheres to 215 216 217 The following is a brief discussion on some of the fields exported via the 218 /sys filesystem and how their values are affected by the PCC driver: 219 220 2.1 scaling_available_frequencies: 221 ---------------------------------- 222 scaling_available_frequencies is not created in /sys. No intermediate 223 frequencies need to be listed because the BIOS will try to achieve any 224 frequency, within limits, requested by the governor. A frequency does not have 225 to be strictly associated with a P-state. 226 227 2.2 cpuinfo_transition_latency: 228 ------------------------------- 229 The cpuinfo_transition_latency field is 0. The PCC specification does 230 not include a field to expose this value currently. 231 232 2.3 cpuinfo_cur_freq: 233 --------------------- 234 A) Often cpuinfo_cur_freq will show a value different than what is declared 235 in the scaling_available_frequencies or scaling_cur_freq, or scaling_max_freq. 236 This is due to "turbo boost" available on recent Intel processors. If certain 237 conditions are met the BIOS can achieve a slightly higher speed than requested 238 by OSPM. An example: 239 240 scaling_cur_freq : 2933000 241 cpuinfo_cur_freq : 3196000 242 243 B) There is a round-off error associated with the cpuinfo_cur_freq value. 244 Since the driver obtains the current frequency as a "percentage" (%) of the 245 nominal frequency from the BIOS, sometimes, the values displayed by 246 scaling_cur_freq and cpuinfo_cur_freq may not match. An example: 247 248 scaling_cur_freq : 1600000 249 cpuinfo_cur_freq : 1583000 250 251 In this example, the nominal frequency is 2933 MHz. The driver obtains the 252 current frequency, cpuinfo_cur_freq, as 54% of the nominal frequency: 253 254 54% of 2933 MHz = 1583 MHz 255 256 Nominal frequency is the maximum frequency of the processor, and it usually 257 corresponds to the frequency of the P0 P-state. 258 259 2.4 related_cpus: 260 ----------------- 261 The related_cpus field is identical to affected_cpus. 262 263 affected_cpus : 4 264 related_cpus : 4 265 266 Currently, the PCC driver does not evaluate _PSD. The platforms that support 267 PCC do not implement SW_ALL. So OSPM doesn't need to perform any coordination 268 to ensure that the same frequency is requested of all dependent CPUs. 269 270 3. Caveats: 271 ----------- 272 The "cpufreq_stats" module in its present form cannot be loaded and 273 expected to work with the PCC driver. Since the "cpufreq_stats" module 274 provides information wrt each P-state, it is not applicable to the PCC driver.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.