Hi everyone -

I'm looking for some suggestion/ideas around following use case:

We have a board that has single precision FPU support - so we enabled hard vfp compiler flag. However, using AssemblyScript, the compiled WASM instructions has a lot of f64 ops. When we encounter those, following error will be thrown: error: AOT module load failed: resolve symbol __aeabi_dsub failed. Which is expected since hard vfp would not compile ARM fplib into the app.

What's the best way to work around this issue? Right now I unblocked this by replacing f64 instructions in the .wat file with f32 ones. But that would break built in lib functions which uses f64 explicitly - for example, f32.toString() no longer works as expected.

We could also avoid using built in libs involve float point altogether - but that makes devX bad and we had to create bindings in C everywhere.

Any suggestions would be greatly appreciated!