[BUG] Overload resolved with incorrect float type from Python scope

[christophercrouzet]

Bug Description

When calling the following function, from either scope, the resulting z value should be a wp.float64:

@wp.func
def foo(a: wp.float64):
    z = -a * a
    z = wp.exp(z)

Currently, the line z = -a * a fails to evaluate at the Python scope due to:

RuntimeError: Couldn't find a function 'mul' compatible with the arguments 'float, float64'

Calling foo() from inside a kernel works fine.

System Information

No response

[c0d1f1ed]

We ran into test failures after https://github.com/NVIDIA/warp/commit/d48c853930535a99d33e1dbedfda0a85fb1f069b switched MacOS tests from x86-64 to ARM64. It might have the same root cause.