Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

8341127: Extra call to MethodHandle::asType from memory segment var handles fails to inline #21283

Closed
wants to merge 9 commits into from
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@


import jdk.internal.loader.ClassLoaders;
import jdk.internal.vm.annotation.DontInline;
import jdk.internal.vm.annotation.IntrinsicCandidate;

import java.lang.constant.ClassDesc;
Expand Down Expand Up @@ -867,7 +868,7 @@ public final MethodHandle asType(MethodType newType) {
if (at != null) {
return at;
}
return setAsTypeCache(asTypeUncached(newType));
return setAsTypeCache(newType);
}

private MethodHandle asTypeCached(MethodType newType) {
Expand All @@ -885,7 +886,9 @@ private MethodHandle asTypeCached(MethodType newType) {
return null;
}

private MethodHandle setAsTypeCache(MethodHandle at) {
@DontInline
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you leave a comment explaining why we put DontInline here?

private MethodHandle setAsTypeCache(MethodType newType) {
MethodHandle at = asTypeUncached(newType);
// Don't introduce a strong reference in the cache if newType depends on any class loader other than
// current method handle already does to avoid class loader leaks.
if (isSafeToCache(at.type)) {
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,140 @@
/*
* Copyright (c) 2024, 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.
*
* 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 org.openjdk.bench.java.lang.foreign;

import java.lang.foreign.Arena;
import java.lang.foreign.MemorySegment;

import org.openjdk.jmh.annotations.Benchmark;
import org.openjdk.jmh.annotations.BenchmarkMode;
import org.openjdk.jmh.annotations.CompilerControl;
import org.openjdk.jmh.annotations.Fork;
import org.openjdk.jmh.annotations.Measurement;
import org.openjdk.jmh.annotations.Mode;
import org.openjdk.jmh.annotations.OutputTimeUnit;
import org.openjdk.jmh.annotations.Param;
import org.openjdk.jmh.annotations.Setup;
import org.openjdk.jmh.annotations.State;
import org.openjdk.jmh.annotations.TearDown;
import org.openjdk.jmh.annotations.Warmup;
import sun.misc.Unsafe;

import java.lang.invoke.MethodHandle;
import java.lang.invoke.MethodHandles;
import java.lang.reflect.Proxy;
import java.net.URL;
import java.net.URLClassLoader;
import java.util.Arrays;
import java.util.concurrent.TimeUnit;

import static java.lang.foreign.ValueLayout.*;

@BenchmarkMode(Mode.AverageTime)
@Warmup(iterations = 5, time = 500, timeUnit = TimeUnit.MILLISECONDS)
@Measurement(iterations = 10, time = 500, timeUnit = TimeUnit.MILLISECONDS)
@State(org.openjdk.jmh.annotations.Scope.Thread)
@OutputTimeUnit(TimeUnit.MILLISECONDS)
@Fork(value = 3, jvmArgsAppend = { "-XX:-TieredCompilation" })
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The benchmark only reproduces if tiered compilation is disabled. This is due to the fact that some threshold are updated accordingly, and set to a smaller value, which is enough to exhibit the issue.

public class LoopOverNonConstantAsType extends JavaLayouts {

static final Unsafe unsafe = Utils.unsafe;

static final int ELEM_SIZE = 1_000_000;
static final int CARRIER_SIZE = (int)JAVA_LONG.byteSize();
static final int ALLOC_SIZE = ELEM_SIZE * CARRIER_SIZE;

@Param({"false", "true"})
public boolean asTypeCompiled;

Arena arena;
MemorySegment segment;
long unsafe_addr;

@Setup
public void setup() {
unsafe_addr = unsafe.allocateMemory(ALLOC_SIZE);
for (int i = 0; i < ELEM_SIZE; i++) {
unsafe.putInt(unsafe_addr + (i * CARRIER_SIZE) , i);
}
arena = Arena.ofConfined();
segment = arena.allocate(ALLOC_SIZE, 1);
for (int i = 0; i < ELEM_SIZE; i++) {
VH_INT.set(segment, (long) i, i);
}
if (asTypeCompiled) {
compileAsType();
}
}

public interface T { }

static final int TYPE_SIZE = 100;
static final Class<?>[] types;

static {
types = new Class<?>[TYPE_SIZE];
ClassLoader customLoader = new URLClassLoader(new URL[0], LoopOverNonConstantAsType.class.getClassLoader());
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Technically, using a different class loadrer is not required - but using classes with different class loaders can end up taking even longer code paths when calling MethodHandle::asType, so I've added that here.

for (int i = 0 ; i < TYPE_SIZE ; i++) {
types[i] = Proxy.newProxyInstance(customLoader,
new Class<?>[] { T.class }, (_, _, _) -> null).getClass();
}
}

@CompilerControl(CompilerControl.Mode.DONT_INLINE)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the intent was to block inlining of asType, so it gets compiled in isolation? That should be done with a CompileCommand though. This annotation just blocks inlining of compileAsType AFAIK.

Copy link
Contributor Author

@mcimadamore mcimadamore Oct 1, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see what you mean... I'll do some experiments and see how the current annotation affects the benchmark - if at all.

Copy link
Contributor Author

@mcimadamore mcimadamore Oct 1, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The issue with using a compile command is that it will then work globally. So the benchmark would not really test much - besides checking that var handle access is slow when asType cannot inline. What I was trying to do here was to avoid asType being inlined into that compileAsType method - but I agree that the annotation I added isn't it.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I believe EXCLUDE is what I was reaching out for

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think EXCLUDE would work, yes. True about CompileCommand having a global effect, so that won't work. (With a CompilerDirective file we could be more precise, not sure if it's worth it though).

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In the latest iteration, I removed the annotation. Doesn't seem to be making any difference.

void compileAsType() {
for (Class<?> type : types) {
MethodHandle handle = MethodHandles.zero(Object.class);
Class<?>[] args = new Class<?>[254];
Arrays.fill(args, Object.class);
handle = MethodHandles.dropArguments(handle, 0, args);
for (int j = 0; j < args.length ; j++) {
handle = handle.asType(handle.type().changeParameterType(j, type));
}
}
}

@TearDown
public void tearDown() {
arena.close();
unsafe.freeMemory(unsafe_addr);
}

@Benchmark
public long unsafe_loop() {
long res = 0;
for (int i = 0; i < ELEM_SIZE; i ++) {
res += unsafe.getLong(unsafe_addr + (i * CARRIER_SIZE));
}
return res;
}

@Benchmark
public long segment_loop() {
long sum = 0;
for (int i = 0; i < ELEM_SIZE; i++) {
sum += segment.get(JAVA_LONG, i * CARRIER_SIZE);

mcimadamore marked this conversation as resolved.
Show resolved Hide resolved
}
return sum;
}
}