Often test bundles that take too long to execute become long pole or bottleneck holding up the Bluepill completion for longer than what is ideally needed. So, splitting the test bundles based on the time estimates of all tests in each split could increase the concurrency, re-gain the lost concurrency opportunities and complete the testing as quickly as possible.
The splitting strategy here is to make sure no single bundle takes longer than the time taken by all bundles divided by the maximum number of concurrent simulators allowed.
Often test bundles that take too long to execute become long pole or bottleneck holding up the Bluepill completion for longer than what is ideally needed. So, splitting the test bundles based on the time estimates of all tests in each split could increase the concurrency, re-gain the lost concurrency opportunities and complete the testing as quickly as possible.
The splitting strategy here is to make sure no single bundle takes longer than
the time taken by all bundles divided by the maximum number of concurrent simulators allowed
.Please review. \cc @ob, @chenxiao0228