On Wed, 27 May 2026 22:02:50 GMT, Paul Sandoz <[email protected]> wrote:
>> Hi @jddarcy , >> As can be seen in the commit >> https://github.com/openjdk/jdk/pull/28002/changes/26b5def93fc479548a8694c43f7fb5cad5ba8549 >> >> Change is mainly in the fallback implementation of VectorOperators.NEG. I am >> not sure if this qualifies as specification change. > > Right, it's fixing a bug to ensure it behaves correctlty. Its covered by: > >> An operation referencing a Java operator is mapped to a method on {@code >> Float16} that specifies that operator's semantics. > > The behavior of Unary Minus Operator {@code -} maps `Float16.negate`. To clarify, I was asking to double-check that the paragraph 65 * The specifications for operations on elements of this class are written as if 66 * {@code Float16} is a primitive floating-point type. An operation referencing a 67 * Java operator is mapped to a method on {@code Float16} that specifies that 68 * operator's semantics. For example, the semantics of the {@code +} operator, 69 * as referenced by {@link Vector#add(Vector)} and {@link VectorOperators#ADD}, 70 * is mapped to the method {@link Float16#add(Float16, Float16)}. 71 * An operation referencing a method on {@link Math} is mapped to a method of the 72 * same name on {@code Float16}, if it exists. For example, {@link Math#fma} is 73 * mapped to {@link Float16#fma}, as referenced by {@link Float16Vector#fma(short, short)} 74 * and {@link VectorOperators#FMA}. in Float16Vector.java doesn't not need to be updated. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/28002#discussion_r3314755916
