LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] selftests/intel_pstate: Improve test, minor fixes
@ 2018-04-10 22:11 Daniel Díaz
  2018-05-08 20:27 ` Shuah Khan
  0 siblings, 1 reply; 2+ messages in thread
From: Daniel Díaz @ 2018-04-10 22:11 UTC (permalink / raw)
  To: shuahkh, linux-kselftest
  Cc: Daniel Díaz, Shuah Khan, Kate Stewart, Thomas Gleixner,
	Philippe Ombredanne, Thomas Meyer, Greg Kroah-Hartman, open list

A few changes improve the overall usability of the test:
* fix a hard-coded maximum frequency (3300),
* don't adjust the CPU frequency if only evaluating results,
* fix a comparison for multiple frequencies.

A symptom of that last issue looked like this:
  ./run.sh: line 107: [: too many arguments
  ./run.sh: line 110: 3099
  3099
  3100-3100: syntax error in expression (error token is \"3099
  3100-3100\")

Because a check will count how many differente frequencies
there are among the CPUs of the system, and after they are
tallied another read is performed, which might produce
different results.

Signed-off-by: Daniel Díaz <daniel.diaz@linaro.org>
---
 tools/testing/selftests/intel_pstate/run.sh | 24 ++++++++++--------------
 1 file changed, 10 insertions(+), 14 deletions(-)

diff --git a/tools/testing/selftests/intel_pstate/run.sh b/tools/testing/selftests/intel_pstate/run.sh
index c670359..bde31a0 100755
--- a/tools/testing/selftests/intel_pstate/run.sh
+++ b/tools/testing/selftests/intel_pstate/run.sh
@@ -48,11 +48,12 @@ function run_test () {
 
 	echo "sleeping for 5 seconds"
 	sleep 5
-	num_freqs=$(cat /proc/cpuinfo | grep MHz | sort -u | wc -l)
-	if [ $num_freqs -le 2 ]; then
-		cat /proc/cpuinfo | grep MHz | sort -u | tail -1 > /tmp/result.$1
+	grep MHz /proc/cpuinfo | sort -u > /tmp/result.freqs
+	num_freqs=$(wc -l /tmp/result.freqs | awk ' { print $1 } ')
+	if [ $num_freqs -ge 2 ]; then
+		tail -n 1 /tmp/result.freqs > /tmp/result.$1
 	else
-		cat /proc/cpuinfo | grep MHz | sort -u > /tmp/result.$1
+		cp /tmp/result.freqs /tmp/result.$1
 	fi
 	./msr 0 >> /tmp/result.$1
 
@@ -82,21 +83,20 @@ _max_freq=$(cpupower frequency-info -l | tail -1 | awk ' { print $2 } ')
 max_freq=$(($_max_freq / 1000))
 
 
-for freq in `seq $max_freq -100 $min_freq`
+[ $EVALUATE_ONLY -eq 0 ] && for freq in `seq $max_freq -100 $min_freq`
 do
 	echo "Setting maximum frequency to $freq"
 	cpupower frequency-set -g powersave --max=${freq}MHz >& /dev/null
-	[ $EVALUATE_ONLY -eq 0 ] && run_test $freq
+	run_test $freq
 done
 
-echo "=============================================================================="
+[ $EVALUATE_ONLY -eq 0 ] && cpupower frequency-set -g powersave --max=${max_freq}MHz >& /dev/null
 
+echo "=============================================================================="
 echo "The marketing frequency of the cpu is $mkt_freq MHz"
 echo "The maximum frequency of the cpu is $max_freq MHz"
 echo "The minimum frequency of the cpu is $min_freq MHz"
 
-cpupower frequency-set -g powersave --max=${max_freq}MHz >& /dev/null
-
 # make a pretty table
 echo "Target      Actual      Difference     MSR(0x199)     max_perf_pct"
 for freq in `seq $max_freq -100 $min_freq`
@@ -104,10 +104,6 @@ do
 	result_freq=$(cat /tmp/result.${freq} | grep "cpu MHz" | awk ' { print $4 } ' | awk -F "." ' { print $1 } ')
 	msr=$(cat /tmp/result.${freq} | grep "msr" | awk ' { print $3 } ')
 	max_perf_pct=$(cat /tmp/result.${freq} | grep "max_perf_pct" | awk ' { print $2 } ' )
-	if [ $result_freq -eq $freq ]; then
-		echo " $freq        $result_freq             0          $msr         $(($max_perf_pct*3300))"
-	else
-		echo " $freq        $result_freq          $(($result_freq-$freq))          $msr          $(($max_perf_pct*$max_freq))"
-	fi
+	echo " $freq        $result_freq          $(($result_freq-$freq))          $msr          $(($max_perf_pct*$max_freq))"
 done
 exit 0
-- 
2.7.4

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] selftests/intel_pstate: Improve test, minor fixes
  2018-04-10 22:11 [PATCH] selftests/intel_pstate: Improve test, minor fixes Daniel Díaz
@ 2018-05-08 20:27 ` Shuah Khan
  0 siblings, 0 replies; 2+ messages in thread
From: Shuah Khan @ 2018-05-08 20:27 UTC (permalink / raw)
  To: Daniel Díaz, shuahkh, linux-kselftest
  Cc: Kate Stewart, Thomas Gleixner, Philippe Ombredanne, Thomas Meyer,
	Greg Kroah-Hartman, open list, Shuah Khan

On 04/10/2018 04:11 PM, Daniel Díaz wrote:
> A few changes improve the overall usability of the test:
> * fix a hard-coded maximum frequency (3300),
> * don't adjust the CPU frequency if only evaluating results,
> * fix a comparison for multiple frequencies.
> 
> A symptom of that last issue looked like this:
>   ./run.sh: line 107: [: too many arguments
>   ./run.sh: line 110: 3099
>   3099
>   3100-3100: syntax error in expression (error token is \"3099
>   3100-3100\")
> 
> Because a check will count how many differente frequencies
> there are among the CPUs of the system, and after they are
> tallied another read is performed, which might produce
> different results.
> 
> Signed-off-by: Daniel Díaz <daniel.diaz@linaro.org>

Applied to linux-kselftest fixes for 4.17-rc5

thanks,
-- Shuah

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2018-05-08 20:27 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-04-10 22:11 [PATCH] selftests/intel_pstate: Improve test, minor fixes Daniel Díaz
2018-05-08 20:27 ` Shuah Khan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).