[BACKEND] Allow nvptx to pass ll ir to CUDAModule#404

tqchen merged 1 commit intoapache:masterapache/tvm:masterfrom

tqchen:mastertqchen/tvm:masterCopy head branch name to clipboard