From: dino@in.ibm.com
To: Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@elte.hu>,
Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: linux-kernel@vger.kernel.org, linux-rt-users@vger.kernel.org,
John Stultz <johnstul@us.ibm.com>,
Darren Hart <dvhltc@us.ibm.com>, John Kacur <jkacur@redhat.com>
Subject: [patch -rt 10/17] x86: Add generic aperf/mperf code
Date: Thu, 22 Oct 2009 18:07:53 +0530 [thread overview]
Message-ID: <20091022124111.756583994@spinlock.in.ibm.com> (raw)
In-Reply-To: 20091022123743.506956796@spinlock.in.ibm.com
[-- Attachment #1: sched-lb-9-new.patch --]
[-- Type: text/plain, Size: 4532 bytes --]
Move some of the aperf/mperf code out from the cpufreq driver
thingy so that other people can enjoy it too.
Index: linux-2.6.31.4-rt14-lb1/arch/x86/include/asm/processor.h
===================================================================
--- linux-2.6.31.4-rt14-lb1.orig/arch/x86/include/asm/processor.h 2009-10-21 10:47:17.000000000 -0400
+++ linux-2.6.31.4-rt14-lb1/arch/x86/include/asm/processor.h 2009-10-21 10:48:41.000000000 -0400
@@ -27,6 +27,7 @@
#include <linux/cpumask.h>
#include <linux/cache.h>
#include <linux/threads.h>
+#include <linux/math64.h>
#include <linux/init.h>
/*
@@ -1010,4 +1011,33 @@
extern int get_tsc_mode(unsigned long adr);
extern int set_tsc_mode(unsigned int val);
+struct aperfmperf {
+ u64 aperf, mperf;
+};
+
+static inline void get_aperfmperf(struct aperfmperf *am)
+{
+ WARN_ON_ONCE(!boot_cpu_has(X86_FEATURE_APERFMPERF));
+
+ rdmsrl(MSR_IA32_APERF, am->aperf);
+ rdmsrl(MSR_IA32_MPERF, am->mperf);
+}
+
+#define APERFMPERF_SHIFT 10
+
+static inline
+unsigned long calc_aperfmperf_ratio(struct aperfmperf *old,
+ struct aperfmperf *new)
+{
+ u64 aperf = new->aperf - old->aperf;
+ u64 mperf = new->mperf - old->mperf;
+ unsigned long ratio = aperf;
+
+ mperf >>= APERFMPERF_SHIFT;
+ if (mperf)
+ ratio = div64_u64(aperf, mperf);
+
+ return ratio;
+}
+
#endif /* _ASM_X86_PROCESSOR_H */
Index: linux-2.6.31.4-rt14-lb1/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
===================================================================
--- linux-2.6.31.4-rt14-lb1.orig/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c 2009-10-21 10:47:17.000000000 -0400
+++ linux-2.6.31.4-rt14-lb1/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c 2009-10-21 10:48:41.000000000 -0400
@@ -70,11 +70,7 @@
static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data);
-struct acpi_msr_data {
- u64 saved_aperf, saved_mperf;
-};
-
-static DEFINE_PER_CPU(struct acpi_msr_data, msr_data);
+static DEFINE_PER_CPU(struct aperfmperf, old_perf);
DEFINE_TRACE(power_mark);
@@ -243,23 +239,12 @@
return cmd.val;
}
-struct perf_pair {
- union {
- struct {
- u32 lo;
- u32 hi;
- } split;
- u64 whole;
- } aperf, mperf;
-};
-
/* Called via smp_call_function_single(), on the target CPU */
static void read_measured_perf_ctrs(void *_cur)
{
- struct perf_pair *cur = _cur;
+ struct aperfmperf *am = _cur;
- rdmsr(MSR_IA32_APERF, cur->aperf.split.lo, cur->aperf.split.hi);
- rdmsr(MSR_IA32_MPERF, cur->mperf.split.lo, cur->mperf.split.hi);
+ get_aperfmperf(am);
}
/*
@@ -278,63 +263,17 @@
static unsigned int get_measured_perf(struct cpufreq_policy *policy,
unsigned int cpu)
{
- struct perf_pair readin, cur;
- unsigned int perf_percent;
+ struct aperfmperf perf;
+ unsigned long ratio;
unsigned int retval;
- if (smp_call_function_single(cpu, read_measured_perf_ctrs, &readin, 1))
+ if (smp_call_function_single(cpu, read_measured_perf_ctrs, &perf, 1))
return 0;
- cur.aperf.whole = readin.aperf.whole -
- per_cpu(msr_data, cpu).saved_aperf;
- cur.mperf.whole = readin.mperf.whole -
- per_cpu(msr_data, cpu).saved_mperf;
- per_cpu(msr_data, cpu).saved_aperf = readin.aperf.whole;
- per_cpu(msr_data, cpu).saved_mperf = readin.mperf.whole;
-
-#ifdef __i386__
- /*
- * We dont want to do 64 bit divide with 32 bit kernel
- * Get an approximate value. Return failure in case we cannot get
- * an approximate value.
- */
- if (unlikely(cur.aperf.split.hi || cur.mperf.split.hi)) {
- int shift_count;
- u32 h;
-
- h = max_t(u32, cur.aperf.split.hi, cur.mperf.split.hi);
- shift_count = fls(h);
-
- cur.aperf.whole >>= shift_count;
- cur.mperf.whole >>= shift_count;
- }
-
- if (((unsigned long)(-1) / 100) < cur.aperf.split.lo) {
- int shift_count = 7;
- cur.aperf.split.lo >>= shift_count;
- cur.mperf.split.lo >>= shift_count;
- }
-
- if (cur.aperf.split.lo && cur.mperf.split.lo)
- perf_percent = (cur.aperf.split.lo * 100) / cur.mperf.split.lo;
- else
- perf_percent = 0;
-
-#else
- if (unlikely(((unsigned long)(-1) / 100) < cur.aperf.whole)) {
- int shift_count = 7;
- cur.aperf.whole >>= shift_count;
- cur.mperf.whole >>= shift_count;
- }
-
- if (cur.aperf.whole && cur.mperf.whole)
- perf_percent = (cur.aperf.whole * 100) / cur.mperf.whole;
- else
- perf_percent = 0;
-
-#endif
+ ratio = calc_aperfmperf_ratio(&per_cpu(old_perf, cpu), &perf);
+ per_cpu(old_perf, cpu) = perf;
- retval = (policy->cpuinfo.max_freq * perf_percent) / 100;
+ retval = (policy->cpuinfo.max_freq * ratio) >> APERFMPERF_SHIFT;
return retval;
}
--
next prev parent reply other threads:[~2009-10-22 12:41 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-10-22 12:37 [patch -rt 00/17] [patch -rt] Sched load balance backport dino
2009-10-22 12:37 ` [patch -rt 01/17] sched: restore __cpu_power to a straight sum of power dino
2009-10-22 12:37 ` [patch -rt 02/17] sched: SD_PREFER_SIBLING dino
2009-10-22 12:37 ` [patch -rt 03/17] sched: update the cpu_power sum during load-balance dino
2009-10-22 12:37 ` [patch -rt 04/17] sched: add smt_gain dino
2009-10-22 12:37 ` [patch -rt 05/17] sched: dynamic cpu_power dino
2009-10-22 12:37 ` [patch -rt 06/17] sched: scale down cpu_power due to RT tasks dino
2009-10-22 12:37 ` [patch -rt 07/17] sched: try to deal with low capacity dino
2009-10-22 12:37 ` [patch -rt 08/17] sched: remove reciprocal for cpu_power dino
2009-10-22 12:37 ` [patch -rt 09/17] x86: move APERF/MPERF into a X86_FEATURE dino
2009-10-22 12:37 ` dino [this message]
2009-10-22 12:37 ` [patch -rt 11/17] Provide an arch specific hook for cpufreq based scaling of cpu_power dino
2009-10-22 12:37 ` [patch -rt 12/17] x86: sched: provide arch implementations using aperf/mperf dino
2009-10-22 12:37 ` [patch -rt 13/17] sched: cleanup wake_idle power saving dino
2009-10-22 12:37 ` [patch -rt 14/17] sched: cleanup wake_idle dino
2009-10-22 12:37 ` [patch -rt 15/17] sched: Add a missing = dino
2009-10-22 12:37 ` [patch -rt 16/17] sched: Deal with low-load in wake_affine() dino
2009-10-22 12:38 ` [patch -rt 17/17] sched: Fix dynamic power-balancing crash dino
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20091022124111.756583994@spinlock.in.ibm.com \
--to=dino@in.ibm.com \
--cc=a.p.zijlstra@chello.nl \
--cc=dvhltc@us.ibm.com \
--cc=jkacur@redhat.com \
--cc=johnstul@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rt-users@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox