lhutton1 opened a new pull request, #14232: URL: https://github.com/apache/tvm/pull/14232
In the test `test_llvm_div`, some cases can lead to a division by 0. Consider the case when `start=-12, end=-12, dstart=-2 and dend=0`, the range of values input to clipb() will be: [-12, ..., -1, 1] (with 1 being specially selected to prevent division by 0). For the input 1, clipb evaluates to: min(/*dend=*/0, max(/*dstart=*/-12, /*i=*/1)) = 0, thus leading to a division by 0. LLVM11, which runs in CI, doesn't seem to throw any kind of error, while using LLVM15 leads to a "floating point exception". Regardless, since division by 0 is undefined behaviour and the compiler can do anything, the solution here is to fix the test so division by 0 doesn't occur. -- 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]
