On Fri, 21 Feb 2025 00:01:36 GMT, Chen Liang <[email protected]> wrote:
> Simplify the layout access var handles to be direct in some common cases.
> Also made `VarHandle::isAccessModeSupported` report if an access mode is
> supported for a VH.
>
> Reduces the instructions to execute this code in a simple main by 47%:
>
> long[] arr = new long[8];
> var ms = MemorySegment.ofArray(arr);
> ms.setAtIndex(ValueLayout.JAVA_BYTE, 12, (byte) 3);
>
>
> Main overheads in FFM are identified to be:
> 1. Eager initialization of direct MethodHandle; can be CDS archived
> 2. MH combinator forms via LambdaFormEditor, not cached right now and always
> have large overhead
>
> Still need other measures to deal with common user patterns of
> `MethodHandles.insertCoordinates(vh, 1, 0L)` which currently is still very
> slow.
>
> Tests: 2 unrelated failures on tier 1-3
src/java.base/share/classes/java/lang/invoke/VarHandleSegmentViewBase.java line
34:
> 32: * Base class for memory segment var handle view implementations.
> 33: */
> 34: final class VarHandleSegmentViewBase extends VarHandle {
Should the name of this class change so that we drop `Base` ? It's no longer
extended by anything.
src/java.base/share/classes/java/lang/invoke/VarHandles.java line 767:
> 765: // new HandleType(Object.class, long.class),
> 766: //
> 767: // // MS[base][offset]->T
I like how you programmatically generate all guards for all the shapes we care
about.
On `MS[base][offset]` I wonder if generating a guard matters (for now) given
that such a var handle will be indirect?
src/java.base/share/classes/java/lang/invoke/X-VarHandleSegmentView.java.template
line 77:
> 75: }
> 76:
> 77: @ForceInline
Question: can this and the `offset` method go in the `VarHandleSegmentViewBase`
class? They don't seem to depend on anything in these classes - so it seems
mostly duplicated code?
src/java.base/share/classes/java/lang/invoke/X-VarHandleSegmentView.java.template
line 85:
> 83:
> 84: @ForceInline
> 85: static $type$ get(VarHandle ob, Object obb, long base) {
I like this: a fast path for non-strided access, while, at the same time,
reusing the impl.
In the future we can also add another variant for when there's only one index
(which is by far the most common case).
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/23720#discussion_r1965223737
PR Review Comment: https://git.openjdk.org/jdk/pull/23720#discussion_r1965218844
PR Review Comment: https://git.openjdk.org/jdk/pull/23720#discussion_r1965222924
PR Review Comment: https://git.openjdk.org/jdk/pull/23720#discussion_r1965226302
