This repo tests what flags are needed to force LLVM compilers (flang, amdflang, armflang) to allocate automatic arrays on the heap instead of the stack, to avoid stack overflow for large arrays.

As of LLVM 21.1.8, the flags needed are

-mmlir -fdynamic-heap-array

In contrast, -fno-stack-arrays is not sufficient.

Run the GitHub Action workflow to test when new versions of the compilers are released, to see if the flags needed have changed.