Rows with "empty" values

[karan...@chromium.org]

I was running a benchmark (loading.desktop) locally, and noticed that rows corresponding to many stories had "empty" values. Does this mean the metrics for these stories were not recorded due to some failure? 

[Juan Antonio Navarro Pérez]

rows for which metrics? can you share the results.html file you produced?

On Tue, Jan 23, 2018 at 5:46 AM <karan...@chromium.org> wrote:

I was running a benchmark (loading.desktop) locally, and noticed that rows corresponding to many stories had "empty" values. Does this mean the metrics for these stories were not recorded due to some failure? 

--
You received this message because you are subscribed to the Google Groups "telemetry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+...@chromium.org.

[Karan Bhatia]

Attached the file. Several of the metrics have empty rows for "cold" cache temperature. Also see timeToFirstIdleCpu. Example: navigate to: 

results.html?s=avg&g=name.stories.cache_temperatureTag&c=0&t=timeToOnload-(e-1.r-(AirBnB-(e-1).Colorado%2eedu-(e-1).ArsTechnica-(e-1).AllRecipes-(e-1))).timeToFirstContentfulPaint-(e-1.r-(ArsTechnica-(e-1).AirBnB-(e-1).AllRecipes-(e-1))).timeToFirstCpuIdle-(e-1).timeToFirstPaint-(e-1.r-(AllRecipes-(e-1).AirBnB-(e-1)))

[Karan Bhatia]

Also, I found it weird that the onload times for "cold" cache temperature were better than those for "warm". Shouldn't it be the other way round?

[Ned]

+Kouhei Ueno +Kunihiko Sakamoto may be able to help with diagnosing the weirdness of the results

[Karan Bhatia]

The file is attached in one of the above messages. Also, I filed a bug https://bugs.chromium.org/p/chromium/issues/detail?id=805115, it seems that loading.desktop is being tested with mobile sites.

On Tue, Jan 23, 2018 at 3:50 PM, Kouhei Ueno <kou...@google.com> wrote:

Would you share us the file?

To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+unsubscribe@chromium.org.

--

kouhei

[Kouhei Ueno]

Would you share us the file?

On Wed, Jan 24, 2018 at 8:46 AM, Ned <nedn...@google.com> wrote:

To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+unsubscribe@chromium.org.

--

kouhei

[Kunihiko Sakamoto]

loading benchmark doesn't wait for CPU idle after load, so timeToFirstCpuIdle / timeToInteractive can be empty.

Other metrics being empty is probably due to bad recordings. https://chromium-review.googlesource.com/c/chromium/src/+/883004 should fix it.

Kunihiko Sakamoto

2018年1月24日(水) 8:56 Karan Bhatia <karan...@google.com>:

To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+...@chromium.org.

--

kouhei

[Karan Bhatia]

Ah, this didn't get posted on the thread:

The fact that the average computations shown in results.html, merely ignores these values without any indication of the same is a bit misleading. The average values can't then be used to compare two runs of the benchmark.

Also, do you have an idea about this discrepancy?

Also, I found it weird that the onload times for "cold" cache temperature were better than those for "warm". Shouldn't it be the other way round?

Thanks.