On 04/06/2012 08:11 AM, YongQin Liu wrote:

Hi, Andy

In the attached tjbench.tgz file, there are two files: tjbench_job.json tjbench_result.json

tjbench_job.json is the job file I used, I defined 7 tjbench actions in that file.

and tjbench_result.json is the result of the job, I use lava-dashboard-tool to get it from my local lava environment. you can use put to upload it to your lava serve to see the result.

I assume your local environment has the code to install the binaries for hw graphics on your Panda? The presence of this often has big effects on the variance you get from the test results.

I have some scripts that can massage the data and summarize it. Overall it looks good. Here's the wiki output summary:

https://wiki.linaro.org/AndyDoan/Sandbox/tjbench-results

The variance between the runs is about the best I've seen for a test on Android. Once the code is merged, I'll start including in the runs I do.

-andy