larroy opened a new issue #16962: TVM compiling something serially for 16 min URL: https://github.com/apache/incubator-mxnet/issues/16962 Seems during Build / GPU: USE_INT64_TENSOR_SIZE TVM is compiling some kernels in serial mode and they are extremely slow to compile. Any ideas what's going on here? There's just one core used during the job. ``` jenkins+ 25861 100 2.7 7020648 6996992 ? R 06:31 16:24 cicc --gnu_version=50400 --allow_managed -arch compute_70 -m64 -ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 --include_file_name tmpxft_0000468a_00000000-2_my_kernel.fatbin.c -tused -nvvmir-library /usr/local/cuda/bin/../nvvm/libdevice/libdevice.10.bc --gen_module_id_file --module_id_file_name /tmp/tmpxft_0000468a_00000000-3_my_kernel.module_id --orig_src_file_name /tmp/tmpcqw6smr7/my_kernel.cu /tmp/tmpxft_0000468a_00000000-6_my_kernel.cpp1.ii -o /tmp/tmpcqw6smr7/my_kernel.ptx ```
---------------------------------------------------------------- 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
