The maximum LMUL is 8. We need 16 vector registers for two LMUL-8 arguments. The modification follows the proposal of psABI in https://github.com/riscv/riscv-elf-psabi-doc/pull/171
There are lots of test files need to be updated. It is too large to contain all of these test files in one commit to upload. I only add vadd test files to demonstrate the modification. If this patch is accepted, I will update all of these test files and commit them to the repository.