ekalda commented on PR #14003: URL: https://github.com/apache/tvm/pull/14003#issuecomment-1438784694
> This seems pretty outstanding to me, tests appear to be passing which indicates it at least functions - I don't think we have any performance regression checks so I'll just trust @ekalda. > > Given the Arduino tests are passing as well, my guess is @ekalda inadvertently lowered memory usage which is out of scope of this PR - but I'll let that slide. > > Leaving this open for @mehrdadh or @guberti to comment on the Arduino memory but otherwise I'll aim to merge this tomorrow. Yes, the lower memory footprint of the operator is a direct consequence of removing the explicit unrolling and vectorization across long axes, but I think it is a good thing, I can't think of a case where in the absence of performance improvement we would want to have a larger code size. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
