Why asUnbounded()/isUnbounded() was removed in JDK21?
Maurizio Cimadamore
maurizio.cimadamore at oracle.com
Wed May 3 17:19:34 UTC 2023
On 03/05/2023 18:16, Cheng Jin wrote:
>
> Hi Maurizio,
>
> Many thanks for your quick response with detailed clarification.
>
> There is another question that confused me is: why the pointer
> segment should be granted with Long.MAX_VALUE given users should know
> the actual size of the segment. My understanding is they should
> convert a zero-sized segment (pointer) to a segment with the whatever
> size it is supposed to be for safety as explained in terms of
> zero-length memory segments at
> https://download.java.net/java/early_access/jdk21/docs/api/java.base/java/lang/foreign/MemorySegment.html#wrapping-addresses
> <https://urldefense.com/v3/__https://download.java.net/java/early_access/jdk21/docs/api/java.base/java/lang/foreign/MemorySegment.html*wrapping-addresses__;Iw!!ACWV5N9M2RV99hQ!LqVLAbaqhXIVSJ6q9ag4POc8JTH3yNv8N9Est_-g566HIwIl5w5byhodf_2TQxWdRG_kpQeXMAAB4NhQFW-kuqE0$>
>
There are two cases:
* the address layout has no target layout - in this case the pointer
segment has zero size (as you'd expect). Resizing is done, unsafely,
using MemorySegment::reinterpret
* the address layout has a target layout - in this case the pointer
segment has the size of the target layout
In my example I was using the target layout machinery to create a target
layout whose size is Long.MAX_VALUE - as that is semantically equivalent
to an "unbounded" address layout in the Java 20 API.
Maurizio
> Best Regards
>
> Cheng Jin
>
> *From:*Maurizio Cimadamore <maurizio.cimadamore at oracle.com>
> *Sent:* May 3, 2023 9:47 AM
> *To:* Cheng Jin <jincheng at ca.ibm.com>; panama-dev at openjdk.org
> *Subject:* [EXTERNAL] Re: Why asUnbounded()/isUnbounded() was removed
> in JDK21?
>
> Hi, the functionality is there - it has just been renamed to make it
> more general. In 20, one could only create an address layout that was
> _unbounded_ e. g. that pointed to a region of memory of unbounded
> size. In 21, AddressLayout is a new
>
> ZjQcmQRYFpfptBannerStart
>
> *This Message Is From an External Sender *
>
> This message came from outside your organization.
>
> ZjQcmQRYFpfptBannerEnd
>
> Hi,
> the functionality is there - it has just been renamed to make it more
> general.
>
> In 20, one could only create an address layout that was _unbounded_
> e.g. that pointed to a region of memory of unbounded size.
>
> In 21, AddressLayout is a new toplevel class. An address layout has a
> _target layout_ [1] which is used to describe the layout of the
> pointed region of memory.
>
> This means that you can still get a behavior similar to `asUnbounded`,
> by doing:
>
> ```java
> ADDRESS.withTargetLayout(MemoryLayout.sequenceLayout(JAVA_BYTE))
> ```
>
> The above will create a sequence layout of unbound size (so, the size
> of that layout would be Long.MAX_VALUE), and use that layout as the
> target for the address layout. Which means that when this address is
> passed to MemorySegment::get(AddressLayout), or to the Linker, you
> will get back a MemorySegment whose size is Long.MAX_VALUE (as in 21).
>
> But, if you know the layout of the pointed region of memory
> statically, you can also use a sharper target layout - e.g.
>
> ```java
> ADDRESS.withTargetLayout(JAVA_INT)
> ```
>
> The latter will create an address layout whose target layout is
> _exactly_ one int layout. This means that when you pass this address
> layout to MemorySegment::get(AddressLayout),or to the Linker, you will
> get back a MemorySegment whose size is precisely 4 bytes. This is
> especially useful for upcalls (as the qsort example in JEP 442
> demonstrates).
>
> Hope this helps.
>
> Maurizio
>
> [1] -
> https://cr.openjdk.org/~pminborg/panama/21/v2/javadoc/api/java.base/java/lang/foreign/AddressLayout.html#withTargetLayout(java.lang.foreign.MemoryLayout)
>
> On 03/05/2023 14:34, Cheng Jin wrote:
>
> Hi there,
>
> I am wondering why asUnbounded()/isUnbounded() (existing in JDK20)
> related code was removed from
> java.base/share/classes/java/lang/foreign/ValueLayout.java in JDK21.
>
> Is there any reason for deleting them all (or moving them
> somewhere else which I ignored) for the unbounded segment (pointer)?
>
> Best Regards
>
> Cheng Jin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/panama-dev/attachments/20230503/008454da/attachment.htm>
More information about the panama-dev
mailing list