mirror of
https://github.com/openjdk/jdk.git
synced 2025-08-26 14:24:46 +02:00
340 lines
16 KiB
Java
340 lines
16 KiB
Java
/*
|
|
* Copyright (c) 2022, 2025, Oracle and/or its affiliates. All rights reserved.
|
|
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
|
|
*
|
|
* This code is free software; you can redistribute it and/or modify it
|
|
* under the terms of the GNU General Public License version 2 only, as
|
|
* published by the Free Software Foundation. Oracle designates this
|
|
* particular file as subject to the "Classpath" exception as provided
|
|
* by Oracle in the LICENSE file that accompanied this code.
|
|
*
|
|
* This code is distributed in the hope that it will be useful, but WITHOUT
|
|
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
* version 2 for more details (a copy is included in the LICENSE file that
|
|
* accompanied this code).
|
|
*
|
|
* You should have received a copy of the GNU General Public License version
|
|
* 2 along with this work; if not, write to the Free Software Foundation,
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
*
|
|
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
|
|
* or visit www.oracle.com if you need additional information or have any
|
|
* questions.
|
|
*/
|
|
|
|
package java.lang.foreign;
|
|
|
|
import jdk.internal.foreign.MemorySessionImpl;
|
|
import jdk.internal.ref.CleanerFactory;
|
|
|
|
import java.lang.foreign.MemorySegment.Scope;
|
|
import java.util.function.Consumer;
|
|
|
|
/**
|
|
* An arena controls the lifecycle of native memory segments, providing both flexible
|
|
* allocation and timely deallocation.
|
|
* <p>
|
|
* An arena has a {@linkplain MemorySegment.Scope scope} - the <em>arena scope</em>.
|
|
* All the segments allocated by the arena are associated with the arena scope. As such,
|
|
* the arena determines the temporal bounds of all the memory segments allocated by it.
|
|
* <p>
|
|
* Moreover, an arena also determines whether access to memory segments allocated by it
|
|
* should be {@linkplain MemorySegment#isAccessibleBy(Thread) restricted} to specific
|
|
* threads. An arena is a {@link SegmentAllocator} and features several allocation
|
|
* methods that can be used by clients to obtain native segments.
|
|
* <p>
|
|
* The simplest arena is the {@linkplain Arena#global() global arena}. The global arena
|
|
* features an <em>unbounded lifetime</em>. The scope of the global arena is the global
|
|
* scope. As such, native segments allocated with the global arena are always accessible
|
|
* and their backing regions of memory are never deallocated. Moreover, memory segments
|
|
* allocated with the global arena can be {@linkplain MemorySegment#isAccessibleBy(Thread) accessed}
|
|
* from any thread.
|
|
* {@snippet lang = java:
|
|
* MemorySegment segment = Arena.global().allocate(100, 1); // @highlight regex='global()'
|
|
* ...
|
|
* // segment is never deallocated!
|
|
*}
|
|
* <p>
|
|
* Alternatively, clients can obtain an {@linkplain Arena#ofAuto() automatic arena}, that
|
|
* is an arena which features a <em>bounded lifetime</em> that is managed, automatically,
|
|
* by the garbage collector. The scope of an automatic arena is an automatic scope. As
|
|
* such, the regions of memory backing memory segments allocated with the automatic arena
|
|
* are deallocated at some unspecified time <em>after</em> the automatic arena (and all
|
|
* the segments allocated by it) becomes {@linkplain java.lang.ref##reachability unreachable},
|
|
* as shown below:
|
|
* {@snippet lang = java:
|
|
* MemorySegment segment = Arena.ofAuto().allocate(100, 1); // @highlight regex='ofAuto()'
|
|
* ...
|
|
* segment = null; // the segment region becomes available for deallocation after this point
|
|
*}
|
|
* Memory segments allocated with an automatic arena can also be
|
|
* {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} from any thread.
|
|
* <p>
|
|
* Rather than leaving deallocation in the hands of the Java runtime, clients will often
|
|
* wish to exercise control over the timing of deallocation for regions of memory that
|
|
* back memory segments. Two kinds of arenas support this, namely {@linkplain #ofConfined() confined}
|
|
* and {@linkplain #ofShared() shared} arenas. They both feature bounded lifetimes that
|
|
* are managed manually. For instance, when a confined arena is {@linkplain #close() closed}
|
|
* successfully, its scope is {@linkplain Scope#isAlive() invalidated}. As a result, all
|
|
* the memory segments allocated by the arena can no longer be accessed, and their
|
|
* regions of memory are deallocated:
|
|
*
|
|
* {@snippet lang = java:
|
|
* MemorySegment segment = null;
|
|
* try (Arena arena = Arena.ofConfined()) { // @highlight regex='ofConfined()'
|
|
* segment = arena.allocate(100);
|
|
* ...
|
|
* } // segment region deallocated here
|
|
* segment.get(ValueLayout.JAVA_BYTE, 0); // throws IllegalStateException
|
|
*}
|
|
*
|
|
* Memory segments allocated with a {@linkplain #ofConfined() confined arena} can only be
|
|
* accessed (and closed) by the thread that created the arena. If access to a memory
|
|
* segment from multiple threads is required, clients can allocate segments in a
|
|
* {@linkplain #ofShared() shared arena} instead.
|
|
* <p>
|
|
* The characteristics of the various arenas are summarized in the following table:
|
|
*
|
|
* <blockquote><table class="plain">
|
|
* <caption style="display:none">Arenas characteristics</caption>
|
|
* <thead>
|
|
* <tr>
|
|
* <th scope="col">Kind</th>
|
|
* <th scope="col">Bounded lifetime</th>
|
|
* <th scope="col">Explicitly closeable</th>
|
|
* <th scope="col">Accessible from multiple threads</th>
|
|
* </tr>
|
|
* </thead>
|
|
* <tbody>
|
|
* <tr><th scope="row" style="font-weight:normal">Global</th>
|
|
* <td style="text-align:center;">No</td>
|
|
* <td style="text-align:center;">No</td>
|
|
* <td style="text-align:center;">Yes</td></tr>
|
|
* <tr><th scope="row" style="font-weight:normal">Automatic</th>
|
|
* <td style="text-align:center;">Yes</td>
|
|
* <td style="text-align:center;">No</td>
|
|
* <td style="text-align:center;">Yes</td></tr>
|
|
* <tr><th scope="row" style="font-weight:normal">Confined</th>
|
|
* <td style="text-align:center;">Yes</td>
|
|
* <td style="text-align:center;">Yes</td>
|
|
* <td style="text-align:center;">No</td></tr>
|
|
* <tr><th scope="row" style="font-weight:normal">Shared</th>
|
|
* <td style="text-align:center;">Yes</td>
|
|
* <td style="text-align:center;">Yes</td>
|
|
* <td style="text-align:center;">Yes</td></tr>
|
|
* </tbody>
|
|
* </table></blockquote>
|
|
*
|
|
* <h2 id = "thread-confinement">Safety and thread-confinement</h2>
|
|
*
|
|
* Arenas provide strong temporal safety guarantees: a memory segment allocated by an
|
|
* arena cannot be accessed <em>after</em> the arena has been closed. The cost of
|
|
* providing this guarantee varies based on the number of threads that have access to the
|
|
* memory segments allocated by the arena. For instance, if an arena is always created
|
|
* and closed by one thread, and the memory segments allocated by the arena are always
|
|
* accessed by that same thread, then ensuring correctness is trivial.
|
|
* <p>
|
|
* Conversely, if an arena allocates segments that can be accessed by multiple threads,
|
|
* or if the arena can be closed by a thread other than the accessing thread, then
|
|
* ensuring correctness is much more complex. For example, a segment allocated with the
|
|
* arena might be accessed <em>while</em> another thread attempts, concurrently, to close
|
|
* the arena. To provide the strong temporal safety guarantee without forcing every
|
|
* client, even simple ones, to incur a performance impact, arenas are divided into
|
|
* <em>thread-confined</em> arenas, and <em>shared</em> arenas.
|
|
* <p>
|
|
* Confined arenas, support strong thread-confinement guarantees. Upon creation, they are
|
|
* assigned an <em>owner thread</em>, typically the thread which initiated the creation
|
|
* operation. The segments created by a confined arena can only be
|
|
* {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by the owner thread.
|
|
* Moreover, any attempt to close the confined arena from a thread other than the owner
|
|
* thread will fail with a {@link WrongThreadException}.
|
|
* <p>
|
|
* Shared arenas, on the other hand, have no owner thread. The segments created by a
|
|
* shared arena can be {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by
|
|
* any thread. This might be useful when multiple threads need to access the same memory
|
|
* segment concurrently (e.g. in the case of parallel processing). Moreover, a shared
|
|
* arena can be closed by any thread.
|
|
*
|
|
* <h2 id = "custom-arenas">Custom arenas</h2>
|
|
*
|
|
* Clients can define custom arenas to implement more efficient allocation strategies,
|
|
* or to have better control over when (and by whom) an arena can be closed. As an
|
|
* example, the following code defines a <em>slicing arena</em> that behaves like a
|
|
* confined arena (i.e., single-threaded access), but internally uses a
|
|
* {@linkplain SegmentAllocator#slicingAllocator(MemorySegment) slicing allocator} to
|
|
* respond to allocation requests.
|
|
* When the slicing arena is closed, the underlying confined arena is also closed; this
|
|
* will invalidate all segments allocated with the slicing arena (since the scope of the
|
|
* slicing arena is the same as that of the underlying confined arena):
|
|
*
|
|
* {@snippet lang = java:
|
|
* class SlicingArena implements Arena {
|
|
* final Arena arena = Arena.ofConfined();
|
|
* final SegmentAllocator slicingAllocator;
|
|
*
|
|
* SlicingArena(long size) {
|
|
* slicingAllocator = SegmentAllocator.slicingAllocator(arena.allocate(size));
|
|
* }
|
|
*
|
|
* public MemorySegment allocate(long byteSize, long byteAlignment) {
|
|
* return slicingAllocator.allocate(byteSize, byteAlignment);
|
|
* }
|
|
*
|
|
* public MemorySegment.Scope scope() {
|
|
* return arena.scope();
|
|
* }
|
|
*
|
|
* public void close() {
|
|
* arena.close();
|
|
* }
|
|
*
|
|
* }
|
|
* }
|
|
*
|
|
* In other words, a slicing arena provides a vastly more efficient and scalable
|
|
* allocation strategy, while still retaining the timely deallocation guarantee provided
|
|
* by the underlying confined arena:
|
|
*
|
|
* {@snippet lang = java:
|
|
* try (Arena slicingArena = new SlicingArena(1000)) {
|
|
* for (int i = 0; i < 10; i++) {
|
|
* MemorySegment s = slicingArena.allocateFrom(JAVA_INT, 1, 2, 3, 4, 5);
|
|
* ...
|
|
* }
|
|
* } // all memory allocated is released here
|
|
* }
|
|
*
|
|
* @implSpec
|
|
* Implementations of this interface are thread-safe.
|
|
*
|
|
* @see MemorySegment
|
|
*
|
|
* @since 22
|
|
*/
|
|
public interface Arena extends SegmentAllocator, AutoCloseable {
|
|
|
|
/**
|
|
* Creates a new arena that is managed, automatically, by the garbage collector.
|
|
* Segments allocated with the returned arena can be
|
|
* {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by any thread.
|
|
* Calling {@link #close()} on the returned arena will result in an {@link UnsupportedOperationException}.
|
|
* <p>
|
|
* Memory segments {@linkplain #allocate(long, long) allocated} by the returned arena
|
|
* are zero-initialized.
|
|
*
|
|
* @return a new arena that is managed, automatically, by the garbage collector
|
|
*/
|
|
static Arena ofAuto() {
|
|
return MemorySessionImpl.createImplicit(CleanerFactory.cleaner()).asArena();
|
|
}
|
|
|
|
/**
|
|
* {@return the global arena} Segments allocated with the global arena can be
|
|
* {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by any thread.
|
|
* Calling {@link #close()} on the returned arena will result in
|
|
* an {@link UnsupportedOperationException}.
|
|
* <p>
|
|
* Memory segments {@linkplain #allocate(long, long) allocated} by the returned arena
|
|
* are zero-initialized.
|
|
*/
|
|
static Arena global() {
|
|
class Holder {
|
|
static final Arena GLOBAL = MemorySessionImpl.GLOBAL_SESSION.asArena();
|
|
}
|
|
return Holder.GLOBAL;
|
|
}
|
|
|
|
/**
|
|
* {@return a new confined arena} Segments allocated with the confined arena can
|
|
* only be {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by the
|
|
* thread that created the arena, the arena's <em>owner thread</em>.
|
|
* <p>
|
|
* Memory segments {@linkplain #allocate(long, long) allocated} by the returned arena
|
|
* are zero-initialized.
|
|
*/
|
|
static Arena ofConfined() {
|
|
return MemorySessionImpl.createConfined(Thread.currentThread()).asArena();
|
|
}
|
|
|
|
/**
|
|
* {@return a new shared arena} Segments allocated with the shared arena can be
|
|
* {@linkplain MemorySegment#isAccessibleBy(Thread) accessed} by any thread.
|
|
* <p>
|
|
* Memory segments {@linkplain #allocate(long, long) allocated} by the returned arena
|
|
* are zero-initialized.
|
|
*/
|
|
static Arena ofShared() {
|
|
return MemorySessionImpl.createShared().asArena();
|
|
}
|
|
|
|
/**
|
|
* Returns a native memory segment with the given size (in bytes) and alignment
|
|
* constraint (in bytes).
|
|
* The returned segment is associated with this {@linkplain #scope() arena scope}.
|
|
* The segment's {@link MemorySegment#address() address} is the starting address of
|
|
* the allocated off-heap region of memory backing the segment, and the address is
|
|
* aligned according the provided alignment constraint.
|
|
*
|
|
* @implSpec
|
|
* Implementations of this method must return a native segment featuring the
|
|
* requested size, and that is compatible with the provided alignment constraint.
|
|
* Furthermore, for any two segments {@code S1, S2} returned by this method, the
|
|
* following invariant must hold:
|
|
*
|
|
* {@snippet lang = java:
|
|
* S1.asOverlappingSlice(S2).isEmpty() == true
|
|
* }
|
|
*
|
|
* @param byteSize the size (in bytes) of the off-heap region of memory backing
|
|
* the native memory segment
|
|
* @param byteAlignment the alignment constraint (in bytes) of the off-heap region
|
|
* of memory backing the native memory segment
|
|
* @return a new native memory segment
|
|
* @throws IllegalArgumentException if {@code bytesSize < 0}, {@code byteAlignment <= 0},
|
|
* or if {@code byteAlignment} is not a power of 2
|
|
* @throws IllegalStateException if this arena has already been {@linkplain #close() closed}
|
|
* @throws WrongThreadException if this arena is confined, and this method is called
|
|
* from a thread other than the arena's owner thread
|
|
*/
|
|
@Override
|
|
MemorySegment allocate(long byteSize, long byteAlignment);
|
|
|
|
/**
|
|
* {@return the arena scope}
|
|
*/
|
|
Scope scope();
|
|
|
|
/**
|
|
* Closes this arena. If this method completes normally, the arena scope is no longer
|
|
* {@linkplain Scope#isAlive() alive}, and all the memory segments associated with it
|
|
* can no longer be accessed. Furthermore, any off-heap region of memory backing the
|
|
* segments obtained from this arena are also released.
|
|
*
|
|
* @apiNote This operation is not idempotent; that is, closing an already closed arena
|
|
* <em>always</em> results in an exception being thrown. This reflects a
|
|
* deliberate design choice: failure to close an arena might reveal a bug
|
|
* in the underlying application logic.
|
|
*
|
|
* @implSpec If this method completes normally, then
|
|
* {@code this.scope().isAlive() == false}.
|
|
* Implementations are allowed to throw {@link UnsupportedOperationException}
|
|
* if an explicit close operation is not supported.
|
|
*
|
|
* @see Scope#isAlive()
|
|
*
|
|
* @throws IllegalStateException if the arena has already been closed
|
|
* @throws IllegalStateException if a segment associated with this arena is being
|
|
* accessed concurrently, e.g. by a
|
|
* {@linkplain Linker#downcallHandle(FunctionDescriptor, Linker.Option...) downcall method handle}
|
|
* @throws WrongThreadException if this arena is confined, and this method is called
|
|
* from a thread other than the arena's owner thread
|
|
* @throws UnsupportedOperationException if this arena cannot be closed explicitly
|
|
* @throws RuntimeException if an exception is thrown while executing a custom cleanup action
|
|
* associated with this arena (e.g. as a result of calling
|
|
* {@link MemorySegment#reinterpret(long, Arena, Consumer)} or
|
|
* {@link MemorySegment#reinterpret(Arena, Consumer)}).
|
|
*/
|
|
@Override
|
|
void close();
|
|
}
|