Array of booleans

[David Lloyd]

Out of curiosity, semantically would these vector operations be more like
splaying the bits of the bytes into booleans, or more like mapping each
byte into a true/false value?

On Thu, Mar 18, 2021 at 11:40 AM Paul Sandoz <paul.sandoz at oracle.com> wrote:

> Hmm… if I recall for VarHandle we originally only supported the full set
> of operations for 32 and 64-bit sized primitive types. We then expanded to
> all integral types, leveraging hardware if available otherwise CAS loops
> etc.
>
> An atomic boolean can be supported via an int (see j.u.c.a,AtomicBoolean).
> I don’t recall thinking much about boolean[] arrays at the time.
>
> In addition it’s likely we left it alone since boolean is not an integral
> type, with an unspecified layout, or at least out of all the primitive
> types could be represented differently on different VMs or potentially have
> more varied non-uniform layouts perhaps (for fields and arrays).
>
> --
>
> There is a connected case, not related to directly VarHandles though.
>
> The Vector API already supports loading/storing a VectorMask<E> from/to a
> boolean[] array.
>
> We have recently considered adding loading/storing a Vector<Byte> from/to
> a boolean[] array.
>
> For these cases though it is more contained than exposing general boolean
> access via a VarHandle and MemorySegment.
>
> Paul.
>
> > On Mar 18, 2021, at 8:04 AM, Remi Forax <forax at univ-mlv.fr> wrote:
> >
> > Hi everybody,
> > It seems it's not possible to create a VarHandle on array of booleans
> >  static final VarHandle BOOLEAN_HANDLE = ofSequence(ofValueBits(8,
> LITTLE_ENDIAN))
> >        .varHandle(boolean.class, sequenceElement());
> >
> > Is there a reason for that ?
> >
> > From the JVM POV, a boolean is stores on 8 bits and in the bytecode
> baload/bastore can load/store from either an array of booleans or an array
> of bytes.
> >
> > regards,
> > Rémi
>
>

-- 
- DML • he/him