Lugano Manchester CERN collaboration
December 3rd, 2019
Giles and Walter in vydio, Omar, Josh, Andreas and Giovanni at CERN
. Walter: the outcome of the RV within CI needs to be actionable, i.e. the user needs to be able to take actions to improve the code base performance, re-do the measurement with RV and prove performance gain
. Walter: the presentation of the project by Omar (slides) is too easily dismissive of profilers, the wording "won’t fly" in a paper. Recommandation to either remove the comparison on to the profilers (and just say that you have a new methodology which works well and yields readable and actionable output) , or make specific claims as to what gets done better w/ RV-Vyper
. Omar: we should concentrate on the readability aspect i.e., With RV, we can get a binary outcome which is more suitable for CI compared to a profiler which generates a huge trace.
It is correct that we should not dismiss a profiler in any way because it provides more program information.. Josh: the added value from RV and measurements via Vyper is the readibility of the specifications and the explanation
. Walter: there’s no conflict of ownership with USI over the research which will be carried out by Omar at CERN, since USI is not paying Omar for his work till August 2020
. Walter: the activity proposed today is new in the research path of Omar, as he’s not done CI before. I am supportive of the direction of research, I think it’s promising.
. Giles: introducing profiling and CI in Vyper is sensible and I am supportive of the project presented today.
. Giles: would Vyper would be integrated or experimented with in CI ?
. Josh: we want CI to be able to access and usethe outcome of the RV execution from Vyper
Omar: CI as a way of formulating new specifications for Vypr
. Giles: I can foresee a publication at ASE (Atomated Software Engineering conference) out of this CI+Vyper work
. Walter: pointed out of another conference on March
. Giles and Walter agree that the project outlined can lead to publishable material
. Josh: can we include explanation in the CI process ?
Once we have got something going it would be good to add to Omar’s proposal