Improving GPU Application Performance with NVIDIA CUDA 11.2 Device Link Time Optimization

[phw89]

Does the JIT compiler perform device link-time optimization?

Before device link-time optimization (DLTO) was introduced in CUDA 11.2, it was relatively easy to ensure forward compatibility without worrying too much about differences in performance. You would typically just create a fatbinary containing PTX for the lowest possible arch and SASS for the specific architectures you would normally target. For any future GPU architectures, the JIT compiler would then assemble the PTX into SASS optimized for that specific GPU arch. Now, however, with DLTO, it i…