[Offload] Allow CUDA Kernels to use arbitrarily large shared memory#145963

shiltian merged 3 commits intollvm:mainllvm/llvm-project:mainfrom

gvalson:maingvalson/llvm-project:mainCopy head branch name to clipboard