RFR: 8302976: C2 intrinsification of Float.floatToFloat16 and Float.float16ToFloat yields different result than the interpreter

[Joe Darcy]

On Wed, 22 Feb 2023 03:46:42 GMT, Joe Darcy <darcy at openjdk.org> wrote:

> I'd like to see a more informative description of the problem:
> 
> "float16 NaN values handled differently with and without intrinsification"
> 
> If that is issue reported, it may not be a problem as opposed to
> 
> "incorrect value returned under Float.float16ToFloat intrinsification", etc.

PS The detailed NaN handling is specifically done in a separate test file since the invariants that are true for the software implementation need not be true for a intrinsified hardware-based one.

However, as done for the intrinsics of the transcendental methods (sin, cos, tan), if the float16 conversion intrinsics are used, they should be consistently used or not used regardless of compilation approach (interpreter, C1, C2, etc.).

HTH

-------------

PR: https://git.openjdk.org/jdk/pull/12704