From: Frank Rowand <frank.rowand@am.sony.com>
To: Bhavesh Davda <bhavesh@vmware.com>
Cc: "linux-rt-users@vger.kernel.org" <linux-rt-users@vger.kernel.org>,
Garrett Smith <garrett@vmware.com>,
Lenin Singaravelu <lsingara@vmware.com>
Subject: Re: [PATCH] cyclictest histogram overflow instance tracking
Date: Thu, 11 Oct 2012 12:57:38 -0700 [thread overview]
Message-ID: <507724B2.5050103@am.sony.com> (raw)
In-Reply-To: <21A665E0-66F4-4C7E-BDB2-820DDC5398C8@vmware.com>
On 10/11/12 11:42, Bhavesh Davda wrote:
> From: Bhavesh Davda <bhavesh@vmware.com>
>
> Add feature to cyclictest histogram mode to track cycle counts every time a
> sample overflows the histogram limit. This should help identify if there is a
> timing pattern to jitters in cyclictest runs.
Nice idea...
>
> Example output (with -h 10):
> ...
> Histogram Overflows: 00278
> Histogram Overflow instances:
> 09373 09374 09375 09376 09377 09378 09379 09380 09381 09382 # 00268 others
>
> Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
> ---
> src/cyclictest/cyclictest.c | 32 +++++++++++++++++++++++++++-----
> 1 files changed, 27 insertions(+), 5 deletions(-)
>
> diff --git a/src/cyclictest/cyclictest.c b/src/cyclictest/cyclictest.c
> index 731b4bd..31131ad 100644
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -147,6 +147,7 @@ struct thread_stat {
> double avg;
> long *values;
> long *hist_array;
> + unsigned long *outliers;
> pthread_t thread;
> int threadstarted;
> int tid;
> @@ -154,6 +155,8 @@ struct thread_stat {
> long redmax;
> long cycleofmax;
> long hist_overflow;
> + long num_outliers;
> + long outliers_overflow;
> };
>
> static int shutdown;
> @@ -756,8 +759,13 @@ void *timerthread(void *param)
>
> /* Update the histogram */
> if (histogram) {
> - if (diff >= histogram)
> + if (diff >= histogram) {
> stat->hist_overflow++;
> + if (stat->num_outliers < histogram)
> + stat->outliers[stat->num_outliers++] = stat->cycles - 1;
> + else
> + stat->outliers_overflow++;
> + }
> else
> stat->hist_array[diff]++;
> }
> @@ -1226,7 +1234,7 @@ static void print_tids(struct thread_param *par[], int nthreads)
>
> static void print_hist(struct thread_param *par[], int nthreads)
> {
> - int i, j;
> + int i, j, k;
> unsigned long long int log_entries[nthreads+1];
> unsigned long maxmax, alloverflows;
>
> @@ -1280,9 +1288,19 @@ static void print_hist(struct thread_param *par[], int nthreads)
> printf("# Histogram Overflows:");
Suggest:
printf("# Histogram Overflows at cycle number:");
> alloverflows = 0;
> for (j = 0; j < nthreads; j++) {
> - printf(" %05lu", par[j]->stats->hist_overflow);
> + printf(" %05lu", par[j]->stats->hist_overflow);
Am I blind, or is that a whitespace change? (The latest version I have does not seem
to have any white space issues in that line.)
> alloverflows += par[j]->stats->hist_overflow;
> }
> + printf("\n");
> + printf("# Histogram Overflow instances:\n");
> + for (j = 0; j < nthreads; j++) {
Suggest:
+ printf("# thread %d:", j);
Before this patch, a graphing program that treats lines beginning with "#" as a
comment could graph the histogram output.
> + for (k = 0; k < par[j]->stats->num_outliers; k++)
> + printf(" %05lu", par[j]->stats->outliers[k]);
Nit: You don't need stats->outliers_overflow. You can calculate it from
stats->hist_overflow - stats->num_outliers:
> + if (par[j]->stats->outliers_overflow > 0)
> + printf(" # %05lu others", par[j]->stats->outliers_overflow);
> + printf("\n");
> + }
> +
> if (histofall && nthreads > 1)
> printf(" %05lu", alloverflows);
> printf("\n");
> @@ -1434,10 +1452,12 @@ int main(int argc, char **argv)
> int bufsize = histogram * sizeof(long);
>
> stat->hist_array = threadalloc(bufsize, node);
> - if (stat->hist_array == NULL)
> + stat->outliers = threadalloc(bufsize, node);
> + if (stat->hist_array == NULL || stat->outliers == NULL)
> fatal("failed to allocate histogram of size %d on node %d\n",
> histogram, i);
> memset(stat->hist_array, 0, bufsize);
> + memset(stat->outliers, 0, bufsize);
> }
>
> if (verbose) {
> @@ -1553,8 +1573,10 @@ int main(int argc, char **argv)
>
> if (histogram) {
> print_hist(parameters, num_threads);
> - for (i = 0; i < num_threads; i++)
> + for (i = 0; i < num_threads; i++) {
> threadfree(statistics[i]->hist_array, histogram*sizeof(long), parameters[i]->node);
> + threadfree(statistics[i]->outliers, histogram*sizeof(long), parameters[i]->node);
> + }
> }
>
> if (tracelimit) {
next prev parent reply other threads:[~2012-10-11 19:57 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-10-11 18:42 [PATCH] cyclictest histogram overflow instance tracking Bhavesh Davda
2012-10-11 19:57 ` Frank Rowand [this message]
2012-10-11 20:59 ` Bhavesh Davda
2012-10-15 16:38 ` Bhavesh Davda
2012-10-15 18:11 ` Frank Rowand
2012-10-15 18:13 ` John Kacur
2012-10-16 0:33 ` John Kacur
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=507724B2.5050103@am.sony.com \
--to=frank.rowand@am.sony.com \
--cc=bhavesh@vmware.com \
--cc=garrett@vmware.com \
--cc=linux-rt-users@vger.kernel.org \
--cc=lsingara@vmware.com \
/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