On Mon, 9 Jan 2023 13:17:42 GMT, Raffaello Giulietti <rgiulie...@openjdk.org> wrote:
>> I see - thanks for the explanation - the problem is with `floatToIntBits` vs >> `floatToRawIntBits` in the "put" operation. The get operation is ok, but >> then having asymmetry where we use a float VH in one case (get) but not in >> the other (put) is not desirable. > > @mcimadamore According to the spec, `floatToRawIntBits` and > `doubleToRawLongBits` ensure that the original input bits are all preserved, > even for NaN values. > On the other hand, the spec says that `intBitsToFloat` and `longBitsToDouble` > only ensure a best effort in preserving the output bits of NaN values. > So I think `VarHandle` `get` and `put` do their best to preserve the bits, > but no more than the underlying platform can possibly offer. > This behavior for floating-point values is not fully specified in > `VarHandle`, so there's room for improvement in the doc. At least for the purposes of this PR, I think leaving the calls to the bitwise conversion methods in place is fine. ------------- PR: https://git.openjdk.org/jdk/pull/11840
