altanh opened a new pull request #7851: URL: https://github.com/apache/tvm/pull/7851
Create a rocblas handle once per thread rather than once per kernel invocation (improves perf). This also fixes a memory leak from missing handle destructor in batched_matmul. cc @tqchen @csullivan -- 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]
