anijain2305 opened a new pull request #5196: [Topi x86] Missing vectorize for depthwise conv2d. URL: https://github.com/apache/incubator-tvm/pull/5196 @yzhliu @kevinthesun Missing vectorize. I checked the TVM IR, this can vectorize the stores of partial outputs registers to memory. I did not any see perf benefit, because it might be negligible. But, we should do this explicitly. We already do this for conv2d_avx_common.
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
