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

[David Holmes]

On Wed, 22 Feb 2023 02:08:27 GMT, Sandhya Viswanathan <sviswanathan at openjdk.org> wrote:

> Change the java/lang/float.java and the corresponding shared runtime constant expression evaluation to generate QNaN.
> The HW instructions generate QNaNs and not SNaNs for floating point instructions. This happens across double, float, and float16 data types. The most significant bit of mantissa is set to 1 for QNaNs.

Changing the Java code to match the semantics of a given architecture's HW instruction risks requiring that other architectures have to implement additional code to match those semantics, potentially impacting the benefit of using an intrinsic in the first place.

Consistent output across all execution contexts is certainly a desirable quality, but at what cost?

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

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