StanfordAHA / CGRAFlow

Integration test for entire CGRA flow
BSD 3-Clause "New" or "Revised" License
12 stars 2 forks source link

Srtest #32

Closed steveri closed 7 years ago

steveri commented 7 years ago

Added gold-standard comparisons for intermediate files, and it seems to be running successfully see e.g. https://travis-ci.org/StanfordAHA/CGRAFlow/builds/275589583.

This new flow includes comparisons for three intermediate files, e.g. for file 'pointwise" it compares

Note the Halide output DAG is sufficiently deterministic that (for now at least) I can do a straight text-compare (diff) to check for correctness.

The other two comparisons are a kind of half-assed topological compare of my own invention, necessitated by the fact that, given the same input, the mapper and PNR tools will create nondeterministically different (but correct) outputs. [Update: Ross says mapper output is now deterministic; so I may change this to an exact-compre in a later revision] The topo compare is not perfect, but good enough to have caught or most recent failed build, which was the goal...see me if you want details...

steveri commented 7 years ago

Oof. Lenny, thanks so much for your thoughtful review and reply. Clearly I forgot to do a cleanup step before filing the git pull, something I usually do but somehow missed this time...I will delete a bunch of unnecessary crap before the final merge...

steveri commented 7 years ago

Okay, finished my code cleanup. It's all comments and whitespace removal ish, so should not have affected functionality at all, we shall see after travis script final checks have completed.

steveri commented 7 years ago

PNR is having intermittent problems, is why the latest check failed (https://travis-ci.org/StanfordAHA/CGRAFlow/builds/277409056). Nothing to do with my changes I think, but I am rerunning just in case.

steveri commented 7 years ago

It's been almost a week with no new complaints, I'm gonna go ahead and merge...