I started working on a results view in LAVA for the bootchart results since they are now part of the daily runs. This is using a partial copy of the data I have locally, so please don't concern yourself too much with the actual data in it. A couple of things to point out:
1. legend is not placed well, that's something I'm not yet sure how to fix as my javascript-fu is lacking here (zyga, any ideas on this?) 2. The set of results here is not very large. That's adjustable and when it's using live data, should have a lot more data points to better see trends 3. This is purposefully restricted to a single board. We are not running benchmarks in a controlled enough manner as to make comparison of boards to one another reasonable, nor is it encouraged due to the collaborative nature of Linaro.
In particular I'm looking for opinions on how it would be most useful to display this data. This view shows all 4 image types on a single chart. I did a previous version that had them separate. Is there a preference? It would also be easy to do both on the same page, but perhaps a bit redundant.
Thanks, Paul Larson
This time with the attachment :)
On Mon, Sep 12, 2011 at 2:16 PM, Paul Larson paul.larson@linaro.org wrote:
I started working on a results view in LAVA for the bootchart results since they are now part of the daily runs. This is using a partial copy of the data I have locally, so please don't concern yourself too much with the actual data in it. A couple of things to point out:
- legend is not placed well, that's something I'm not yet sure how to fix
as my javascript-fu is lacking here (zyga, any ideas on this?) 2. The set of results here is not very large. That's adjustable and when it's using live data, should have a lot more data points to better see trends 3. This is purposefully restricted to a single board. We are not running benchmarks in a controlled enough manner as to make comparison of boards to one another reasonable, nor is it encouraged due to the collaborative nature of Linaro.
In particular I'm looking for opinions on how it would be most useful to display this data. This view shows all 4 image types on a single chart. I did a previous version that had them separate. Is there a preference? It would also be easy to do both on the same page, but perhaps a bit redundant.
Thanks, Paul Larson
Is this "time to boot to a prompt"?
cheers, Jesse
On Mon, Sep 12, 2011 at 1:13 PM, Paul Larson paul.larson@linaro.org wrote:
This time with the attachment :)
On Mon, Sep 12, 2011 at 2:16 PM, Paul Larson paul.larson@linaro.org wrote:
I started working on a results view in LAVA for the bootchart results since they are now part of the daily runs. This is using a partial copy of the data I have locally, so please don't concern yourself too much with the actual data in it. A couple of things to point out:
- legend is not placed well, that's something I'm not yet sure how to fix
as my javascript-fu is lacking here (zyga, any ideas on this?) 2. The set of results here is not very large. That's adjustable and when it's using live data, should have a lot more data points to better see trends 3. This is purposefully restricted to a single board. We are not running benchmarks in a controlled enough manner as to make comparison of boards to one another reasonable, nor is it encouraged due to the collaborative nature of Linaro. In particular I'm looking for opinions on how it would be most useful to display this data. This view shows all 4 image types on a single chart. I did a previous version that had them separate. Is there a preference? It would also be easy to do both on the same page, but perhaps a bit redundant. Thanks, Paul Larson
linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
On Mon, Sep 12, 2011 at 3:32 PM, Jesse Barker jesse.barker@linaro.orgwrote:
Is this "time to boot to a prompt"?
Someone can correct me if I'm wrong here, but iirc it looks for getty or the display manager (or greeter) depending on which runlevel it hits. http://www.bootchart.org/ has more information on the benchmark itself.
Thanks, Paul Larson
The machine, configuration would be also good to have as part of the context. Media (class if applicable) in use.
The times noted for developer and nano seem quite long.
Regards, Tom
On Mon, Sep 12, 2011 at 3:13 PM, Paul Larson paul.larson@linaro.org wrote:
This time with the attachment :)
On Mon, Sep 12, 2011 at 2:16 PM, Paul Larson paul.larson@linaro.org wrote:
I started working on a results view in LAVA for the bootchart results since they are now part of the daily runs. This is using a partial copy of the data I have locally, so please don't concern yourself too much with the actual data in it. A couple of things to point out:
- legend is not placed well, that's something I'm not yet sure how to fix
as my javascript-fu is lacking here (zyga, any ideas on this?) 2. The set of results here is not very large. That's adjustable and when it's using live data, should have a lot more data points to better see trends 3. This is purposefully restricted to a single board. We are not running benchmarks in a controlled enough manner as to make comparison of boards to one another reasonable, nor is it encouraged due to the collaborative nature of Linaro. In particular I'm looking for opinions on how it would be most useful to display this data. This view shows all 4 image types on a single chart. I did a previous version that had them separate. Is there a preference? It would also be easy to do both on the same page, but perhaps a bit redundant. Thanks, Paul Larson
linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
On Mon, Sep 12, 2011 at 3:37 PM, Tom Gall tom.gall@linaro.org wrote:
The machine, configuration would be also good to have as part of the context. Media (class if applicable) in use.
What kind of configuration are you looking for? This doesn't sound like something that would fit well on the graph, but better in the detailed results which are already there. In this case, it's a pandaboard, as indicated.
As for the sdcard class, I have no idea. Do we even have a way to determine this at runtime? You may want to talk to Avik who contributed the test wrapper if this is something that can be gathered, and we can make sure it's reported along with the time.
Thanks, Paul Larson
On Mon, Sep 12, 2011 at 03:50:17PM -0500, Paul Larson wrote:
On Mon, Sep 12, 2011 at 3:37 PM, Tom Gall tom.gall@linaro.org wrote:
The machine, configuration would be also good to have as part of the context. Media (class if applicable) in use.
What kind of configuration are you looking for? This doesn't sound like something that would fit well on the graph, but better in the detailed results which are already there. In this case, it's a pandaboard, as indicated.
We'll always have a single device per graph, so you could also put some hardware specs under the graph.
As for the sdcard class, I have no idea. Do we even have a way to determine this at runtime?
Tom, do you actually want the class (which AIUI can be misleading performance-wise) or a timed benchmark? Though as Paul is saying, the data he's using to render is kinda bogus..
On Mon, 2011-09-12 at 15:13 -0500, Paul Larson wrote:
This time with the attachment :)
On Mon, Sep 12, 2011 at 2:16 PM, Paul Larson paul.larson@linaro.org wrote: I started working on a results view in LAVA for the bootchart results since they are now part of the daily runs. This is using a partial copy of the data I have locally, so please don't concern yourself too much with the actual data in it. A couple of things to point out: 1. legend is not placed well, that's something I'm not yet sure how to fix as my javascript-fu is lacking here (zyga, any ideas on this?)
By the looks of it, I suspect you're using the flot plugin for jquery to generate the chart. If that's the case, to the best of my knowledge, all you can do (easily) is place the legend on any of the four corners of the chart, so in your case it might be best to place it on the bottom right (se). You can achieve that by adding
legend: {position: "se"}
to the options argument you pass to $.plot()
But once we make nano boot faster its line will overlap with the legend.
2. The set of results here is not very large. That's adjustable and when it's using live data, should have a lot more data points to better see trends 3. This is purposefully restricted to a single board. We are not running benchmarks in a controlled enough manner as to make comparison of boards to one another reasonable, nor is it encouraged due to the collaborative nature of Linaro. In particular I'm looking for opinions on how it would be most useful to display this data. This view shows all 4 image types on a single chart. I did a previous version that had them separate. Is there a preference? It would also be easy to do both on the same page, but perhaps a bit redundant.
I assume the dates refer to when the test was run? I wonder if it wouldn't be better to use the date the hwpack/rootfs was built instead. That way we can easily see what hwpack/rootfs made it faster/slower without having to dig what hwpack+rootfs combination was used on a given date. Actually, it's probably simpler and much nicer to just show the hwpack+rootfs versions on a tooltip when the mouse hovers over one of the points on a given line.
-
Christian Robottom Reis -
Guilherme Salgado -
Jesse Barker -
Paul Larson -
Tom Gall