115 lines
4.4 KiB
Java
115 lines
4.4 KiB
Java
/*
|
|
* 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.
|
|
*/
|
|
|
|
/*
|
|
* @test
|
|
* @bug 8330819
|
|
* @summary Case where VPointer finds an "adr" CastX2P, which contains a CastLL,
|
|
* that has a ctrl after the pre-loop. This value cannot be used in the
|
|
* pre-loop limit for main-loop adjustment.
|
|
* @modules java.base/jdk.internal.misc
|
|
* @modules java.base/jdk.internal.util
|
|
* @run main/othervm -Xbatch compiler.loopopts.superword.TestMemorySegmentMainLoopAlignment
|
|
*/
|
|
|
|
package compiler.loopopts.superword;
|
|
|
|
import java.lang.foreign.*;
|
|
import jdk.internal.misc.Unsafe;
|
|
import jdk.internal.util.Preconditions;
|
|
|
|
public class TestMemorySegmentMainLoopAlignment {
|
|
static final ValueLayout.OfInt ELEMENT_LAYOUT = ValueLayout.JAVA_INT.withByteAlignment(1);
|
|
static final Unsafe UNSAFE = Unsafe.getUnsafe();
|
|
static long RANGE = 6400;
|
|
|
|
// Type definition for the lambda
|
|
interface MSOp {
|
|
int apply(MemorySegment memory, long offset, int i);
|
|
}
|
|
|
|
// Type definition for the lambda
|
|
interface MemoryUnsafeOp {
|
|
int apply(long base, long offset, int i);
|
|
}
|
|
|
|
public static void main(String[] args) {
|
|
// Allocate some raw memory:
|
|
MemorySegment ms = Arena.ofAuto().allocate(6400, Integer.SIZE);
|
|
for (int i = 0; i < 10_000; i++) {
|
|
test1(ms, 0, TestMemorySegmentMainLoopAlignment::memorySegmentGet);
|
|
}
|
|
// Allocate some raw memory:
|
|
long base = UNSAFE.allocateMemory(6400);
|
|
for (int i = 0; i < 10_000; i++) {
|
|
test2(base, 0, TestMemorySegmentMainLoopAlignment::memoryUnsafeGet);
|
|
}
|
|
}
|
|
|
|
// Somehow, it is necessary to pass this as a lambda
|
|
// the checkIndex inside the "get" method produces the CastLL, which eventually pins the index
|
|
// between the pre and main loop.
|
|
static int memorySegmentGet(MemorySegment ms, long o, int i) {
|
|
return ms.get(ELEMENT_LAYOUT, o + i * 4L);
|
|
}
|
|
|
|
static int test1(MemorySegment a, long offset, MSOp f) {
|
|
// Constant size array size allows a known range for the array access/loop iv i.
|
|
int size = 16;
|
|
int[] res = new int[size];
|
|
int sum = 0;
|
|
for (int i = 0; i < size; i++) {
|
|
// With inlining, this eventually becomes:
|
|
// sum += LoadI(MemorySegment / unsafe) + LoadI(array)
|
|
// and we attempt vectorization.
|
|
sum += f.apply(a, offset, i) + res[i];
|
|
}
|
|
return sum;
|
|
}
|
|
|
|
// Somehow, it is necessary to pass this as a lambda
|
|
static int memoryUnsafeGet(long base, long o, int i) {
|
|
long index = o + i * 4L;
|
|
// checkIndex -> CastLL: index >= 0.
|
|
// Together with the info about i (known range for phi), this CastLL floats up to
|
|
// the offset. Then we get adr = CastX2P(base + CastLL(offset)), where the CastLL
|
|
// is pinned between the pre and main loop.
|
|
Preconditions.checkIndex(index, RANGE, null);
|
|
return UNSAFE.getInt(base + index);
|
|
}
|
|
|
|
static int test2(long base, long offset, MemoryUnsafeOp f) {
|
|
// Constant size array size allows a known range for the array access/loop iv i.
|
|
int size = 16;
|
|
int[] res = new int[size];
|
|
int sum = 0;
|
|
for (int i = 0; i < size; i++) {
|
|
// With inlining, this eventually becomes:
|
|
// sum += LoadI(unsafe) + LoadI(array)
|
|
// and we attempt vectorization.
|
|
sum += f.apply(base, offset, i) + res[i];
|
|
}
|
|
return sum;
|
|
}
|
|
}
|