There is clearly wrong output when mperf monitor runs in MAX_FREQ_SYSFS mode:
average frequency shows in kHz unit (despite the intended output to be in MHz),
and percentages for C state information are all wrong (including high/negative
values shown).
The problem is that the max_frequency read on initialization isn't used where it
should have been used on mperf_get_count_percent (to estimate the number of
ticks in the given time period), and the value we read from sysfs is in kHz, so
we must divide it to get the MHz value to use in current calculations.
While at it, also I fixed another small issues in the debug output of
max_frequency value in mperf_get_count_freq.
Signed-off-by: Herton R. Krzesinski <[email protected]>
---
tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index 90a8c4f..369d49ca 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -135,7 +135,7 @@ static int mperf_get_count_percent(unsigned int id, double *percent,
dprint("%s: TSC Ref - mperf_diff: %llu, tsc_diff: %llu\n",
mperf_cstates[id].name, mperf_diff, tsc_diff);
} else if (max_freq_mode == MAX_FREQ_SYSFS) {
- timediff = timespec_diff_us(time_start, time_end);
+ timediff = max_frequency * timespec_diff_us(time_start, time_end);
*percent = 100.0 * mperf_diff / timediff;
dprint("%s: MAXFREQ - mperf_diff: %llu, time_diff: %llu\n",
mperf_cstates[id].name, mperf_diff, timediff);
@@ -176,7 +176,7 @@ static int mperf_get_count_freq(unsigned int id, unsigned long long *count,
dprint("%s: Average freq based on %s maximum frequency:\n",
mperf_cstates[id].name,
(max_freq_mode == MAX_FREQ_TSC_REF) ? "TSC calculated" : "sysfs read");
- dprint("%max_frequency: %lu", max_frequency);
+ dprint("%%max_frequency: %lu\n", max_frequency);
dprint("aperf_diff: %llu\n", aperf_diff);
dprint("mperf_diff: %llu\n", mperf_diff);
dprint("avg freq: %llu\n", *count);
@@ -279,6 +279,7 @@ use_sysfs:
return -1;
}
max_freq_mode = MAX_FREQ_SYSFS;
+ max_frequency /= 1000; /* Default automatically to MHz value */
return 0;
}
--
2.1.0
On Mon, May 11, 2015 at 11:35:35AM -0300, Herton R. Krzesinski wrote:
> There is clearly wrong output when mperf monitor runs in MAX_FREQ_SYSFS mode:
> average frequency shows in kHz unit (despite the intended output to be in MHz),
> and percentages for C state information are all wrong (including high/negative
> values shown).
>
> The problem is that the max_frequency read on initialization isn't used where it
> should have been used on mperf_get_count_percent (to estimate the number of
> ticks in the given time period), and the value we read from sysfs is in kHz, so
> we must divide it to get the MHz value to use in current calculations.
>
> While at it, also I fixed another small issues in the debug output of
> max_frequency value in mperf_get_count_freq.
>
> Signed-off-by: Herton R. Krzesinski <[email protected]>
> ---
> tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
Actually please consider v2 patch below, just a minor change in the debug
output, which isn't a percentage...
From: "Herton R. Krzesinski" <[email protected]>
Date: Mon, 11 May 2015 11:18:14 -0300
Subject: [PATCH v2] cpupower: mperf monitor: fix output in MAX_FREQ_SYSFS mode
There is clearly wrong output when mperf monitor runs in MAX_FREQ_SYSFS mode:
average frequency shows in kHz unit (despite the intended output to be in MHz),
and percentages for C state information are all wrong (including high/negative
values shown).
The problem is that the max_frequency read on initialization isn't used where it
should have been used on mperf_get_count_percent (to estimate the number of
ticks in the given time period), and the value we read from sysfs is in kHz, so
we must divide it to get the MHz value to use in current calculations.
While at it, also I fixed another small issues in the debug output of
max_frequency value in mperf_get_count_freq.
Signed-off-by: Herton R. Krzesinski <[email protected]>
---
tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
v2: remove percent from debug output fix
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index 90a8c4f..c83f160 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -135,7 +135,7 @@ static int mperf_get_count_percent(unsigned int id, double *percent,
dprint("%s: TSC Ref - mperf_diff: %llu, tsc_diff: %llu\n",
mperf_cstates[id].name, mperf_diff, tsc_diff);
} else if (max_freq_mode == MAX_FREQ_SYSFS) {
- timediff = timespec_diff_us(time_start, time_end);
+ timediff = max_frequency * timespec_diff_us(time_start, time_end);
*percent = 100.0 * mperf_diff / timediff;
dprint("%s: MAXFREQ - mperf_diff: %llu, time_diff: %llu\n",
mperf_cstates[id].name, mperf_diff, timediff);
@@ -176,7 +176,7 @@ static int mperf_get_count_freq(unsigned int id, unsigned long long *count,
dprint("%s: Average freq based on %s maximum frequency:\n",
mperf_cstates[id].name,
(max_freq_mode == MAX_FREQ_TSC_REF) ? "TSC calculated" : "sysfs read");
- dprint("%max_frequency: %lu", max_frequency);
+ dprint("max_frequency: %lu\n", max_frequency);
dprint("aperf_diff: %llu\n", aperf_diff);
dprint("mperf_diff: %llu\n", mperf_diff);
dprint("avg freq: %llu\n", *count);
@@ -279,6 +279,7 @@ use_sysfs:
return -1;
}
max_freq_mode = MAX_FREQ_SYSFS;
+ max_frequency /= 1000; /* Default automatically to MHz value */
return 0;
}
--
2.1.0
On Monday, May 11, 2015 12:40:22 PM Herton R. Krzesinski wrote:
> On Mon, May 11, 2015 at 11:35:35AM -0300, Herton R. Krzesinski wrote:
> > There is clearly wrong output when mperf monitor runs in MAX_FREQ_SYSFS mode:
> > average frequency shows in kHz unit (despite the intended output to be in MHz),
> > and percentages for C state information are all wrong (including high/negative
> > values shown).
> >
> > The problem is that the max_frequency read on initialization isn't used where it
> > should have been used on mperf_get_count_percent (to estimate the number of
> > ticks in the given time period), and the value we read from sysfs is in kHz, so
> > we must divide it to get the MHz value to use in current calculations.
> >
> > While at it, also I fixed another small issues in the debug output of
> > max_frequency value in mperf_get_count_freq.
> >
> > Signed-off-by: Herton R. Krzesinski <[email protected]>
> > ---
> > tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 5 +++--
> > 1 file changed, 3 insertions(+), 2 deletions(-)
>
> Actually please consider v2 patch below, just a minor change in the debug
> output, which isn't a percentage...
Thomas, any comments?
> From: "Herton R. Krzesinski" <[email protected]>
> Date: Mon, 11 May 2015 11:18:14 -0300
> Subject: [PATCH v2] cpupower: mperf monitor: fix output in MAX_FREQ_SYSFS mode
>
> There is clearly wrong output when mperf monitor runs in MAX_FREQ_SYSFS mode:
> average frequency shows in kHz unit (despite the intended output to be in MHz),
> and percentages for C state information are all wrong (including high/negative
> values shown).
>
> The problem is that the max_frequency read on initialization isn't used where it
> should have been used on mperf_get_count_percent (to estimate the number of
> ticks in the given time period), and the value we read from sysfs is in kHz, so
> we must divide it to get the MHz value to use in current calculations.
>
> While at it, also I fixed another small issues in the debug output of
> max_frequency value in mperf_get_count_freq.
>
> Signed-off-by: Herton R. Krzesinski <[email protected]>
> ---
> tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> v2: remove percent from debug output fix
>
> diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
> index 90a8c4f..c83f160 100644
> --- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
> +++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
> @@ -135,7 +135,7 @@ static int mperf_get_count_percent(unsigned int id, double *percent,
> dprint("%s: TSC Ref - mperf_diff: %llu, tsc_diff: %llu\n",
> mperf_cstates[id].name, mperf_diff, tsc_diff);
> } else if (max_freq_mode == MAX_FREQ_SYSFS) {
> - timediff = timespec_diff_us(time_start, time_end);
> + timediff = max_frequency * timespec_diff_us(time_start, time_end);
> *percent = 100.0 * mperf_diff / timediff;
> dprint("%s: MAXFREQ - mperf_diff: %llu, time_diff: %llu\n",
> mperf_cstates[id].name, mperf_diff, timediff);
> @@ -176,7 +176,7 @@ static int mperf_get_count_freq(unsigned int id, unsigned long long *count,
> dprint("%s: Average freq based on %s maximum frequency:\n",
> mperf_cstates[id].name,
> (max_freq_mode == MAX_FREQ_TSC_REF) ? "TSC calculated" : "sysfs read");
> - dprint("%max_frequency: %lu", max_frequency);
> + dprint("max_frequency: %lu\n", max_frequency);
> dprint("aperf_diff: %llu\n", aperf_diff);
> dprint("mperf_diff: %llu\n", mperf_diff);
> dprint("avg freq: %llu\n", *count);
> @@ -279,6 +279,7 @@ use_sysfs:
> return -1;
> }
> max_freq_mode = MAX_FREQ_SYSFS;
> + max_frequency /= 1000; /* Default automatically to MHz value */
> return 0;
> }
>
>
--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.