* [PATCH] 3/3 [retry 1] Add support for AMD MPERF/APERF
@ 2010-03-29 18:28 Mark Langsdorf
2010-03-30 9:16 ` Yu, Ke
0 siblings, 1 reply; 2+ messages in thread
From: Mark Langsdorf @ 2010-03-29 18:28 UTC (permalink / raw)
To: xen-devel, osrc-patches
# HG changeset patch
# User mark.langsdorf@amd.com
# Date 1269570377 18000
# Node ID 4009c47dfed487f0432b7cbe3b5bac2f805ada73
# Parent d7401eeeac32a315aa9228d012dea2e78052120d
Starting with Family 0x10, model 10 processors, some AMD processors
will have support for the APERF/MPERF MSRs. This patch adds the
checks necessary to support those MSRs.
It also makes the get_measured_perf function defined inside cpufreq.c
driver independent. The max_freq is passed to the function by the caller
instead of being taking from the acpi-cpufreq only drv_data structure.
This patch supersedes the patch "Add APERF/MPERF support for AMD processors"
I sent on March 1, 2010.
Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com>
diff -r d7401eeeac32 -r 4009c47dfed4 xen/arch/x86/acpi/cpufreq/cpufreq.c
--- a/xen/arch/x86/acpi/cpufreq/cpufreq.c Thu Mar 25 13:02:43 2010 -0500
+++ b/xen/arch/x86/acpi/cpufreq/cpufreq.c Thu Mar 25 21:26:17 2010 -0500
@@ -269,7 +269,7 @@
* Only IA32_APERF/IA32_MPERF ratio is architecturally defined and
* no meaning should be associated with absolute values of these MSRs.
*/
-static unsigned int get_measured_perf(unsigned int cpu, unsigned int flag)
+unsigned int get_measured_perf(unsigned int cpu, unsigned int flag, unsigned
int max_freq)
{
struct cpufreq_policy *policy;
struct perf_pair readin, cur, *saved;
@@ -353,7 +353,7 @@
#endif
- retval = drv_data[policy->cpu]->max_freq * perf_percent / 100;
+ retval = max_freq * perf_percent / 100;
return retval;
}
diff -r d7401eeeac32 -r 4009c47dfed4 xen/arch/x86/acpi/cpufreq/powernow.c
--- a/xen/arch/x86/acpi/cpufreq/powernow.c Thu Mar 25 13:02:43 2010 -0500
+++ b/xen/arch/x86/acpi/cpufreq/powernow.c Thu Mar 25 21:26:17 2010 -0500
@@ -38,6 +38,7 @@
#include <acpi/acpi.h>
#include <acpi/cpufreq/cpufreq.h>
+#define CPUID_6_ECX_APERFMPERF_CAPABILITY (0x1)
#define CPUID_FREQ_VOLT_CAPABILITIES 0x80000007
#define CPB_CAPABLE 0x00000200
#define USE_HW_PSTATE 0x00000080
@@ -61,6 +62,8 @@
static struct powernow_cpufreq_data *drv_data[NR_CPUS];
+static struct cpufreq_driver powernow_cpufreq_driver;
+
struct drv_cmd {
unsigned int type;
cpumask_t mask;
@@ -249,6 +252,10 @@
if (c->cpuid_level >= 6) {
unsigned int edx;
+ unsigned int ecx;
+ ecx = cpuid_ecx(6);
+ if (ecx & CPUID_6_ECX_APERFMPERF_CAPABILITY)
+ powernow_cpufreq_driver.getavg = get_measured_perf;
edx = cpuid_edx(CPUID_FREQ_VOLT_CAPABILITIES);
if ((edx & CPB_CAPABLE) == CPB_CAPABLE) {
policy->turbo = 1;
diff -r d7401eeeac32 -r 4009c47dfed4 xen/drivers/cpufreq/utility.c
--- a/xen/drivers/cpufreq/utility.c Thu Mar 25 13:02:43 2010 -0500
+++ b/xen/drivers/cpufreq/utility.c Thu Mar 25 21:26:17 2010 -0500
@@ -382,7 +382,8 @@
if (cpufreq_driver->getavg)
{
- freq_avg = cpufreq_driver->getavg(cpu, flag);
+ freq_avg = cpufreq_driver->getavg(cpu, flag,
+ policy->cpuinfo.max_freq);
if (freq_avg > 0)
return freq_avg;
}
diff -r d7401eeeac32 -r 4009c47dfed4 xen/include/acpi/cpufreq/cpufreq.h
--- a/xen/include/acpi/cpufreq/cpufreq.h Thu Mar 25 13:02:43 2010 -0500
+++ b/xen/include/acpi/cpufreq/cpufreq.h Thu Mar 25 21:26:17 2010 -0500
@@ -138,7 +138,8 @@
unsigned int target_freq,
unsigned int relation);
unsigned int (*get)(unsigned int cpu);
- unsigned int (*getavg)(unsigned int cpu, unsigned int flag);
+ unsigned int (*getavg)(unsigned int cpu, unsigned int flag,
+ unsigned int max_freq);
int (*exit)(struct cpufreq_policy *policy);
};
diff -r d7401eeeac32 -r 4009c47dfed4 xen/include/acpi/cpufreq/processor_perf.h
--- a/xen/include/acpi/cpufreq/processor_perf.h Thu Mar 25 13:02:43 2010 -0500
+++ b/xen/include/acpi/cpufreq/processor_perf.h Thu Mar 25 21:26:17 2010 -0500
@@ -9,6 +9,8 @@
int get_cpu_id(u8);
int powernow_cpufreq_init(void);
unsigned int powernow_register_driver(void);
+unsigned int get_measured_perf(unsigned int cpu, unsigned int flag,
+ unsigned int max_freq);
void cpufreq_residency_update(unsigned int, uint8_t);
void cpufreq_statistic_update(unsigned int, uint8_t, uint8_t);
-------------------------------------------------------
^ permalink raw reply [flat|nested] 2+ messages in thread
* RE: [PATCH] 3/3 [retry 1] Add support for AMD MPERF/APERF
2010-03-29 18:28 [PATCH] 3/3 [retry 1] Add support for AMD MPERF/APERF Mark Langsdorf
@ 2010-03-30 9:16 ` Yu, Ke
0 siblings, 0 replies; 2+ messages in thread
From: Yu, Ke @ 2010-03-30 9:16 UTC (permalink / raw)
To: Mark Langsdorf, xen-devel, osrc-patches
> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-
> bounces@lists.xensource.com] On Behalf Of Mark Langsdorf
> Sent: Tuesday, March 30, 2010 2:29 AM
> To: xen-devel@lists.xensource.com; osrc-patches@elbe.amd.com
> Subject: [Xen-devel] [PATCH] 3/3 [retry 1] Add support for AMD
> MPERF/APERF
>
> # HG changeset patch
> # User mark.langsdorf@amd.com
> # Date 1269570377 18000
> # Node ID 4009c47dfed487f0432b7cbe3b5bac2f805ada73
> # Parent d7401eeeac32a315aa9228d012dea2e78052120d
> Starting with Family 0x10, model 10 processors, some AMD processors
> will have support for the APERF/MPERF MSRs. This patch adds the
> checks necessary to support those MSRs.
>
> It also makes the get_measured_perf function defined inside cpufreq.c
> driver independent. The max_freq is passed to the function by the caller
> instead of being taking from the acpi-cpufreq only drv_data structure.
>
> This patch supersedes the patch "Add APERF/MPERF support for AMD
> processors"
> I sent on March 1, 2010.
>
> Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com>
>
> diff -r d7401eeeac32 -r 4009c47dfed4 xen/arch/x86/acpi/cpufreq/cpufreq.c
> --- a/xen/arch/x86/acpi/cpufreq/cpufreq.c Thu Mar 25 13:02:43 2010 -0500
> +++ b/xen/arch/x86/acpi/cpufreq/cpufreq.c Thu Mar 25 21:26:17
> 2010 -0500
> @@ -269,7 +269,7 @@
> * Only IA32_APERF/IA32_MPERF ratio is architecturally defined and
> * no meaning should be associated with absolute values of these MSRs.
> */
> -static unsigned int get_measured_perf(unsigned int cpu, unsigned int flag)
> +unsigned int get_measured_perf(unsigned int cpu, unsigned int flag,
> unsigned
> int max_freq)
> {
> struct cpufreq_policy *policy;
> struct perf_pair readin, cur, *saved;
> @@ -353,7 +353,7 @@
>
> #endif
>
> - retval = drv_data[policy->cpu]->max_freq * perf_percent / 100;
> + retval = max_freq * perf_percent / 100;
>
> return retval;
> }
Changing the get_measured_perf interface definition looks not good idea. Leaving interface unchanged and modifying the implementation is more reasonable to me, e.g.
- retval = drv_data[policy->cpu]->max_freq * perf_percent / 100;
+ retval = policy->cpuinfo.max_freq * perf_percent / 100;
Best Regards
Ke
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2010-03-30 9:16 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-03-29 18:28 [PATCH] 3/3 [retry 1] Add support for AMD MPERF/APERF Mark Langsdorf
2010-03-30 9:16 ` Yu, Ke
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.