PaulSandoz/TestLoadStoreBytes.java

## TestLoadStoreBytes.java
package jmh;

import jdk.incubator.foreign.CLinker;
import jdk.incubator.foreign.MemoryAccess;
import jdk.incubator.foreign.MemoryAddress;
import jdk.incubator.foreign.MemorySegment;
import jdk.incubator.foreign.ResourceScope;
import jdk.incubator.vector.ByteVector;
import jdk.incubator.vector.IntVector;
import jdk.incubator.vector.VectorSpecies;
import org.openjdk.jmh.annotations.Benchmark;
import org.openjdk.jmh.annotations.BenchmarkMode;
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.Warmup;

import java.nio.ByteBuffer;
import java.nio.ByteOrder;
import java.util.concurrent.TimeUnit;

@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.NANOSECONDS)
@Fork(value = 1, jvmArgsAppend = {
        "--add-modules=jdk.incubator.foreign,jdk.incubator.vector",
        "-Dforeign.restricted=permit",
        "--enable-native-access", "ALL-UNNAMED"})
public class TestLoadStoreBytes {
    private static final VectorSpecies<Byte> SPECIES = VectorSpecies.ofPreferred(byte.class);

    @Param("1024")
    private int size;

    private byte[] srcArray;

    private byte[] dstArray;


    private ByteBuffer srcBufferHeap;

    private ByteBuffer dstBufferHeap;

    private ByteBuffer srcBufferNative;

    private ByteBuffer dstBufferNative;


    private ResourceScope implicitScope;

    private MemorySegment srcSegmentImplicit;

    private MemorySegment dstSegmentImplicit;

    private ByteBuffer srcBufferSegmentImplicit;

    private ByteBuffer dstBufferSegmentImplicit;


    private MemoryAddress srcAddress;

    private MemoryAddress dstAddress;

    @Setup
    public void setup() {
        srcArray = new byte[size];
        dstArray = srcArray.clone();
        for (int i = 0; i < srcArray.length; i++) {
            srcArray[i] = (byte) i;
        }


        srcBufferHeap = ByteBuffer.allocate(size);
        dstBufferHeap = ByteBuffer.allocate(size);

        srcBufferNative = ByteBuffer.allocateDirect(size);
        dstBufferNative = ByteBuffer.allocateDirect(size);


        implicitScope = ResourceScope.newImplicitScope();
        srcSegmentImplicit = MemorySegment.allocateNative(size, SPECIES.vectorByteSize(), implicitScope);
        srcBufferSegmentImplicit = srcSegmentImplicit.asByteBuffer();
        dstSegmentImplicit = MemorySegment.allocateNative(size, SPECIES.vectorByteSize(), implicitScope);
        dstBufferSegmentImplicit = dstSegmentImplicit.asByteBuffer();


        srcAddress = CLinker.allocateMemory(size);
        dstAddress = CLinker.allocateMemory(size);
    }


    @Benchmark
    public void array() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
            var v = ByteVector.fromArray(SPECIES, srcArray, i);
            v.intoArray(dstArray, i);
        }
    }

    @Benchmark
    public void arrayScalar() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i ++) {
            var v = srcArray[i];
            dstArray[i] = v;
        }
    }


    @Benchmark
    public void bufferHeap() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
            var v = ByteVector.fromByteBuffer(SPECIES, srcBufferHeap, i, ByteOrder.nativeOrder());
            v.intoByteBuffer(dstBufferHeap, i, ByteOrder.nativeOrder());
        }
    }

    @Benchmark
    public void bufferHeapScalar() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
            var v = srcBufferHeap.get(i);
            dstBufferHeap.put(i, v);
        }
    }

    @Benchmark
    public void bufferNative() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
            var v = ByteVector.fromByteBuffer(SPECIES, srcBufferNative, i, ByteOrder.nativeOrder());
            v.intoByteBuffer(dstBufferNative, i, ByteOrder.nativeOrder());
        }
    }

    @Benchmark
    public void bufferNativeScalar() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
            var v = srcBufferNative.get(i);
            dstBufferNative.put(i, v);
        }
    }


    @Benchmark
    public void bufferSegmentImplicit() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
            var v = ByteVector.fromByteBuffer(SPECIES, srcBufferSegmentImplicit, i, ByteOrder.nativeOrder());
            v.intoByteBuffer(dstBufferSegmentImplicit, i, ByteOrder.nativeOrder());
        }
    }

    @Benchmark
    public void segmentImplicitScalar() {
        for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
            var v = MemoryAccess.getByteAtOffset(srcSegmentImplicit, i);
            MemoryAccess.setByteAtOffset(dstSegmentImplicit, i, v);
        }
    }

    @Benchmark
    public void bufferSegmentConfined() {
        try (final var scope = ResourceScope.newConfinedScope()) {
            final var srcBufferSegmentConfined = srcAddress.asSegment(size, scope).asByteBuffer();
            final var dstBufferSegmentConfined = dstAddress.asSegment(size, scope).asByteBuffer();

            for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
                var v = ByteVector.fromByteBuffer(SPECIES, srcBufferSegmentConfined, i, ByteOrder.nativeOrder());
                v.intoByteBuffer(dstBufferSegmentConfined, i, ByteOrder.nativeOrder());
            }
        }
    }
}

/*
git@github.com:openjdk/jdk.git
cd20c01942dd8559a31e51ef2a595c6eba44b8ad refs/remotes/origin/HEAD

# VM options: -XX:-TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
Benchmark                                 (size)  Mode  Cnt      Score     Error  Units
TestLoadStoreBytes.array                    1024  avgt   10     13.767 ±   0.215  ns/op
TestLoadStoreBytes.array                   16384  avgt   10    279.825 ±   5.764  ns/op
TestLoadStoreBytes.arrayScalar              1024  avgt   10     23.255 ±   0.684  ns/op
TestLoadStoreBytes.arrayScalar             16384  avgt   10    195.704 ±   6.330  ns/op
TestLoadStoreBytes.bufferHeap               1024  avgt   10     78.539 ±   0.431  ns/op
TestLoadStoreBytes.bufferHeap              16384  avgt   10   1303.619 ±  31.830  ns/op
TestLoadStoreBytes.bufferHeapScalar         1024  avgt   10    268.601 ±   8.998  ns/op
TestLoadStoreBytes.bufferHeapScalar        16384  avgt   10   4217.291 ±  28.522  ns/op
TestLoadStoreBytes.bufferNative             1024  avgt   10     73.032 ±   0.611  ns/op
TestLoadStoreBytes.bufferNative            16384  avgt   10   1282.294 ±  13.322  ns/op
TestLoadStoreBytes.bufferNativeScalar       1024  avgt   10    257.354 ±  13.796  ns/op
TestLoadStoreBytes.bufferNativeScalar      16384  avgt   10   4064.541 ±  53.939  ns/op
TestLoadStoreBytes.bufferSegmentConfined    1024  avgt   10    439.680 ±   8.065  ns/op
TestLoadStoreBytes.bufferSegmentConfined   16384  avgt   10   1388.378 ±  16.244  ns/op
TestLoadStoreBytes.bufferSegmentImplicit    1024  avgt   10     77.520 ±   0.753  ns/op
TestLoadStoreBytes.bufferSegmentImplicit   16384  avgt   10   1266.489 ±  11.080  ns/op
TestLoadStoreBytes.segmentImplicitScalar    1024  avgt   10    749.856 ±   8.769  ns/op
TestLoadStoreBytes.segmentImplicitScalar   16384  avgt   10  11846.255 ± 130.863  ns/op


# VM options: -XX:+TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
Benchmark                                 (size)  Mode  Cnt      Score     Error  Units
TestLoadStoreBytes.array                    1024  avgt   10     13.728 ±   0.818  ns/op
TestLoadStoreBytes.array                   16384  avgt   10    260.386 ±  11.354  ns/op
TestLoadStoreBytes.arrayScalar              1024  avgt   10     23.388 ±   0.703  ns/op
TestLoadStoreBytes.arrayScalar             16384  avgt   10    190.860 ±  11.174  ns/op
TestLoadStoreBytes.bufferHeap               1024  avgt   10     71.069 ±   0.883  ns/op
TestLoadStoreBytes.bufferHeap              16384  avgt   10   1282.913 ±  11.387  ns/op
TestLoadStoreBytes.bufferHeapScalar         1024  avgt   10    266.658 ±   6.184  ns/op
TestLoadStoreBytes.bufferHeapScalar        16384  avgt   10   4165.673 ±  68.071  ns/op
TestLoadStoreBytes.bufferNative             1024  avgt   10     70.646 ±   1.210  ns/op
TestLoadStoreBytes.bufferNative            16384  avgt   10   1239.830 ±  12.737  ns/op
TestLoadStoreBytes.bufferNativeScalar       1024  avgt   10    251.330 ±   2.522  ns/op
TestLoadStoreBytes.bufferNativeScalar      16384  avgt   10   4042.816 ±  50.017  ns/op
TestLoadStoreBytes.bufferSegmentConfined    1024  avgt   10     50.773 ±   0.413  ns/op
TestLoadStoreBytes.bufferSegmentConfined   16384  avgt   10    535.652 ±   3.546  ns/op
TestLoadStoreBytes.bufferSegmentImplicit    1024  avgt   10     70.323 ±   0.748  ns/op
TestLoadStoreBytes.bufferSegmentImplicit   16384  avgt   10   1265.497 ±   9.701  ns/op
TestLoadStoreBytes.segmentImplicitScalar    1024  avgt   10    747.202 ±   9.150  ns/op
TestLoadStoreBytes.segmentImplicitScalar   16384  avgt   10  11903.696 ± 353.461  ns/op


----
With patch applied:

diff --git a/src/hotspot/share/opto/vectorIntrinsics.cpp b/src/hotspot/share/opto/vectorIntrinsics.cpp
index c36b2a22caf..bb1cc6f723e 100644
--- a/src/hotspot/share/opto/vectorIntrinsics.cpp
+++ b/src/hotspot/share/opto/vectorIntrinsics.cpp
@@ -827,7 +827,7 @@ bool LibraryCallKit::inline_vector_mem_operation(bool is_store) {
   const TypeInstPtr* vbox_type = TypeInstPtr::make_exact(TypePtr::NotNull, vbox_klass);

   if (can_access_non_heap) {
-    insert_mem_bar(Op_MemBarCPUOrder);
+//    insert_mem_bar(Op_MemBarCPUOrder);
   }

   if (is_store) {
@@ -874,7 +874,7 @@ bool LibraryCallKit::inline_vector_mem_operation(bool is_store) {
   old_map->destruct(&_gvn);

   if (can_access_non_heap) {
-    insert_mem_bar(Op_MemBarCPUOrder);
+//    insert_mem_bar(Op_MemBarCPUOrder);
   }

   C->set_max_vector_size(MAX2(C->max_vector_size(), (uint)(num_elem * type2aelembytes(elem_bt))));


# VM options: -XX:-TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
Benchmark                                 (size)  Mode  Cnt      Score     Error  Units
TestLoadStoreBytes.array                    1024  avgt   10     13.713 ±   0.135  ns/op
TestLoadStoreBytes.array                   16384  avgt   10    283.524 ±   4.238  ns/op
TestLoadStoreBytes.arrayScalar              1024  avgt   10     23.123 ±   0.283  ns/op
TestLoadStoreBytes.arrayScalar             16384  avgt   10    200.712 ±   8.645  ns/op
TestLoadStoreBytes.bufferHeap               1024  avgt   10     25.664 ±   0.316  ns/op
TestLoadStoreBytes.bufferHeap              16384  avgt   10    385.964 ±  12.450  ns/op
TestLoadStoreBytes.bufferHeapScalar         1024  avgt   10    265.677 ±   3.302  ns/op
TestLoadStoreBytes.bufferHeapScalar        16384  avgt   10   4163.773 ±  72.358  ns/op
TestLoadStoreBytes.bufferNative             1024  avgt   10     25.556 ±   0.390  ns/op
TestLoadStoreBytes.bufferNative            16384  avgt   10    349.823 ±   4.298  ns/op
TestLoadStoreBytes.bufferNativeScalar       1024  avgt   10    252.061 ±   3.130  ns/op
TestLoadStoreBytes.bufferNativeScalar      16384  avgt   10   4062.956 ±  70.301  ns/op
TestLoadStoreBytes.bufferSegmentConfined    1024  avgt   10     63.998 ±   0.425  ns/op
TestLoadStoreBytes.bufferSegmentConfined   16384  avgt   10    806.393 ±  10.080  ns/op
TestLoadStoreBytes.bufferSegmentImplicit    1024  avgt   10     25.539 ±   0.268  ns/op
TestLoadStoreBytes.bufferSegmentImplicit   16384  avgt   10    349.869 ±   4.381  ns/op
TestLoadStoreBytes.segmentImplicitScalar    1024  avgt   10    754.664 ±  21.680  ns/op
TestLoadStoreBytes.segmentImplicitScalar   16384  avgt   10  11887.237 ± 189.345  ns/op


# VM options: -XX:+TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
Benchmark                                 (size)  Mode  Cnt      Score     Error  Units
TestLoadStoreBytes.array                    1024  avgt   10     14.034 ±   0.176  ns/op
TestLoadStoreBytes.array                   16384  avgt   10    252.624 ±  10.333  ns/op
TestLoadStoreBytes.arrayAdd                 1024  avgt   10     15.911 ±   0.523  ns/op
TestLoadStoreBytes.arrayAdd                16384  avgt   10    271.969 ±   5.597  ns/op
TestLoadStoreBytes.arrayScalar              1024  avgt   10     23.344 ±   0.535  ns/op
TestLoadStoreBytes.arrayScalar             16384  avgt   10    195.957 ±  10.636  ns/op
TestLoadStoreBytes.bufferHeap               1024  avgt   10     25.644 ±   0.281  ns/op
TestLoadStoreBytes.bufferHeap              16384  avgt   10    388.989 ±   4.021  ns/op
TestLoadStoreBytes.bufferHeapScalar         1024  avgt   10    266.135 ±   3.611  ns/op
TestLoadStoreBytes.bufferHeapScalar        16384  avgt   10   4158.670 ±  63.424  ns/op
TestLoadStoreBytes.bufferNative             1024  avgt   10     25.678 ±   0.332  ns/op
TestLoadStoreBytes.bufferNative            16384  avgt   10    349.329 ±   3.850  ns/op
TestLoadStoreBytes.bufferNativeScalar       1024  avgt   10    251.668 ±   3.311  ns/op
TestLoadStoreBytes.bufferNativeScalar      16384  avgt   10   4045.838 ± 129.280  ns/op
TestLoadStoreBytes.bufferSegmentConfined    1024  avgt   10     34.934 ±   0.257  ns/op
TestLoadStoreBytes.bufferSegmentConfined   16384  avgt   10    286.946 ±   6.120  ns/op
TestLoadStoreBytes.bufferSegmentImplicit    1024  avgt   10     25.681 ±   0.501  ns/op
TestLoadStoreBytes.bufferSegmentImplicit   16384  avgt   10    351.099 ±  10.846  ns/op
TestLoadStoreBytes.segmentImplicitScalar    1024  avgt   10    750.972 ±   9.326  ns/op
TestLoadStoreBytes.segmentImplicitScalar   16384  avgt   10  11864.784 ± 135.626  ns/op
 */
	package jmh;

	import jdk.incubator.foreign.CLinker;
	import jdk.incubator.foreign.MemoryAccess;
	import jdk.incubator.foreign.MemoryAddress;
	import jdk.incubator.foreign.MemorySegment;
	import jdk.incubator.foreign.ResourceScope;
	import jdk.incubator.vector.ByteVector;
	import jdk.incubator.vector.IntVector;
	import jdk.incubator.vector.VectorSpecies;
	import org.openjdk.jmh.annotations.Benchmark;
	import org.openjdk.jmh.annotations.BenchmarkMode;
	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.Warmup;

	import java.nio.ByteBuffer;
	import java.nio.ByteOrder;
	import java.util.concurrent.TimeUnit;

	@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.NANOSECONDS)
	@Fork(value = 1, jvmArgsAppend = {
	"--add-modules=jdk.incubator.foreign,jdk.incubator.vector",
	"-Dforeign.restricted=permit",
	"--enable-native-access", "ALL-UNNAMED"})
	public class TestLoadStoreBytes {
	private static final VectorSpecies<Byte> SPECIES = VectorSpecies.ofPreferred(byte.class);

	@Param("1024")
	private int size;

	private byte[] srcArray;

	private byte[] dstArray;


	private ByteBuffer srcBufferHeap;

	private ByteBuffer dstBufferHeap;

	private ByteBuffer srcBufferNative;

	private ByteBuffer dstBufferNative;


	private ResourceScope implicitScope;

	private MemorySegment srcSegmentImplicit;

	private MemorySegment dstSegmentImplicit;

	private ByteBuffer srcBufferSegmentImplicit;

	private ByteBuffer dstBufferSegmentImplicit;


	private MemoryAddress srcAddress;

	private MemoryAddress dstAddress;

	@Setup
	public void setup() {
	srcArray = new byte[size];
	dstArray = srcArray.clone();
	for (int i = 0; i < srcArray.length; i++) {
	srcArray[i] = (byte) i;
	}


	srcBufferHeap = ByteBuffer.allocate(size);
	dstBufferHeap = ByteBuffer.allocate(size);

	srcBufferNative = ByteBuffer.allocateDirect(size);
	dstBufferNative = ByteBuffer.allocateDirect(size);


	implicitScope = ResourceScope.newImplicitScope();
	srcSegmentImplicit = MemorySegment.allocateNative(size, SPECIES.vectorByteSize(), implicitScope);
	srcBufferSegmentImplicit = srcSegmentImplicit.asByteBuffer();
	dstSegmentImplicit = MemorySegment.allocateNative(size, SPECIES.vectorByteSize(), implicitScope);
	dstBufferSegmentImplicit = dstSegmentImplicit.asByteBuffer();


	srcAddress = CLinker.allocateMemory(size);
	dstAddress = CLinker.allocateMemory(size);
	}


	@Benchmark
	public void array() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
	var v = ByteVector.fromArray(SPECIES, srcArray, i);
	v.intoArray(dstArray, i);
	}
	}

	@Benchmark
	public void arrayScalar() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i ++) {
	var v = srcArray[i];
	dstArray[i] = v;
	}
	}


	@Benchmark
	public void bufferHeap() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
	var v = ByteVector.fromByteBuffer(SPECIES, srcBufferHeap, i, ByteOrder.nativeOrder());
	v.intoByteBuffer(dstBufferHeap, i, ByteOrder.nativeOrder());
	}
	}

	@Benchmark
	public void bufferHeapScalar() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
	var v = srcBufferHeap.get(i);
	dstBufferHeap.put(i, v);
	}
	}

	@Benchmark
	public void bufferNative() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
	var v = ByteVector.fromByteBuffer(SPECIES, srcBufferNative, i, ByteOrder.nativeOrder());
	v.intoByteBuffer(dstBufferNative, i, ByteOrder.nativeOrder());
	}
	}

	@Benchmark
	public void bufferNativeScalar() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
	var v = srcBufferNative.get(i);
	dstBufferNative.put(i, v);
	}
	}


	@Benchmark
	public void bufferSegmentImplicit() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
	var v = ByteVector.fromByteBuffer(SPECIES, srcBufferSegmentImplicit, i, ByteOrder.nativeOrder());
	v.intoByteBuffer(dstBufferSegmentImplicit, i, ByteOrder.nativeOrder());
	}
	}

	@Benchmark
	public void segmentImplicitScalar() {
	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i++) {
	var v = MemoryAccess.getByteAtOffset(srcSegmentImplicit, i);
	MemoryAccess.setByteAtOffset(dstSegmentImplicit, i, v);
	}
	}

	@Benchmark
	public void bufferSegmentConfined() {
	try (final var scope = ResourceScope.newConfinedScope()) {
	final var srcBufferSegmentConfined = srcAddress.asSegment(size, scope).asByteBuffer();
	final var dstBufferSegmentConfined = dstAddress.asSegment(size, scope).asByteBuffer();

	for (int i = 0; i < SPECIES.loopBound(srcArray.length); i += SPECIES.length()) {
	var v = ByteVector.fromByteBuffer(SPECIES, srcBufferSegmentConfined, i, ByteOrder.nativeOrder());
	v.intoByteBuffer(dstBufferSegmentConfined, i, ByteOrder.nativeOrder());
	}
	}
	}
	}

	/*
	git@github.com:openjdk/jdk.git
	cd20c01942dd8559a31e51ef2a595c6eba44b8ad refs/remotes/origin/HEAD

	# VM options: -XX:-TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
	Benchmark (size) Mode Cnt Score Error Units
	TestLoadStoreBytes.array 1024 avgt 10 13.767 ± 0.215 ns/op
	TestLoadStoreBytes.array 16384 avgt 10 279.825 ± 5.764 ns/op
	TestLoadStoreBytes.arrayScalar 1024 avgt 10 23.255 ± 0.684 ns/op
	TestLoadStoreBytes.arrayScalar 16384 avgt 10 195.704 ± 6.330 ns/op
	TestLoadStoreBytes.bufferHeap 1024 avgt 10 78.539 ± 0.431 ns/op
	TestLoadStoreBytes.bufferHeap 16384 avgt 10 1303.619 ± 31.830 ns/op
	TestLoadStoreBytes.bufferHeapScalar 1024 avgt 10 268.601 ± 8.998 ns/op
	TestLoadStoreBytes.bufferHeapScalar 16384 avgt 10 4217.291 ± 28.522 ns/op
	TestLoadStoreBytes.bufferNative 1024 avgt 10 73.032 ± 0.611 ns/op
	TestLoadStoreBytes.bufferNative 16384 avgt 10 1282.294 ± 13.322 ns/op
	TestLoadStoreBytes.bufferNativeScalar 1024 avgt 10 257.354 ± 13.796 ns/op
	TestLoadStoreBytes.bufferNativeScalar 16384 avgt 10 4064.541 ± 53.939 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 1024 avgt 10 439.680 ± 8.065 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 16384 avgt 10 1388.378 ± 16.244 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 1024 avgt 10 77.520 ± 0.753 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 16384 avgt 10 1266.489 ± 11.080 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 1024 avgt 10 749.856 ± 8.769 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 16384 avgt 10 11846.255 ± 130.863 ns/op


	# VM options: -XX:+TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
	Benchmark (size) Mode Cnt Score Error Units
	TestLoadStoreBytes.array 1024 avgt 10 13.728 ± 0.818 ns/op
	TestLoadStoreBytes.array 16384 avgt 10 260.386 ± 11.354 ns/op
	TestLoadStoreBytes.arrayScalar 1024 avgt 10 23.388 ± 0.703 ns/op
	TestLoadStoreBytes.arrayScalar 16384 avgt 10 190.860 ± 11.174 ns/op
	TestLoadStoreBytes.bufferHeap 1024 avgt 10 71.069 ± 0.883 ns/op
	TestLoadStoreBytes.bufferHeap 16384 avgt 10 1282.913 ± 11.387 ns/op
	TestLoadStoreBytes.bufferHeapScalar 1024 avgt 10 266.658 ± 6.184 ns/op
	TestLoadStoreBytes.bufferHeapScalar 16384 avgt 10 4165.673 ± 68.071 ns/op
	TestLoadStoreBytes.bufferNative 1024 avgt 10 70.646 ± 1.210 ns/op
	TestLoadStoreBytes.bufferNative 16384 avgt 10 1239.830 ± 12.737 ns/op
	TestLoadStoreBytes.bufferNativeScalar 1024 avgt 10 251.330 ± 2.522 ns/op
	TestLoadStoreBytes.bufferNativeScalar 16384 avgt 10 4042.816 ± 50.017 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 1024 avgt 10 50.773 ± 0.413 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 16384 avgt 10 535.652 ± 3.546 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 1024 avgt 10 70.323 ± 0.748 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 16384 avgt 10 1265.497 ± 9.701 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 1024 avgt 10 747.202 ± 9.150 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 16384 avgt 10 11903.696 ± 353.461 ns/op


	----
	With patch applied:

	diff --git a/src/hotspot/share/opto/vectorIntrinsics.cpp b/src/hotspot/share/opto/vectorIntrinsics.cpp
	index c36b2a22caf..bb1cc6f723e 100644
	--- a/src/hotspot/share/opto/vectorIntrinsics.cpp
	+++ b/src/hotspot/share/opto/vectorIntrinsics.cpp
	@@ -827,7 +827,7 @@ bool LibraryCallKit::inline_vector_mem_operation(bool is_store) {
	const TypeInstPtr* vbox_type = TypeInstPtr::make_exact(TypePtr::NotNull, vbox_klass);

	if (can_access_non_heap) {
	- insert_mem_bar(Op_MemBarCPUOrder);
	+// insert_mem_bar(Op_MemBarCPUOrder);
	}

	if (is_store) {
	@@ -874,7 +874,7 @@ bool LibraryCallKit::inline_vector_mem_operation(bool is_store) {
	old_map->destruct(&_gvn);

	if (can_access_non_heap) {
	- insert_mem_bar(Op_MemBarCPUOrder);
	+// insert_mem_bar(Op_MemBarCPUOrder);
	}

	C->set_max_vector_size(MAX2(C->max_vector_size(), (uint)(num_elem * type2aelembytes(elem_bt))));


	# VM options: -XX:-TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
	Benchmark (size) Mode Cnt Score Error Units
	TestLoadStoreBytes.array 1024 avgt 10 13.713 ± 0.135 ns/op
	TestLoadStoreBytes.array 16384 avgt 10 283.524 ± 4.238 ns/op
	TestLoadStoreBytes.arrayScalar 1024 avgt 10 23.123 ± 0.283 ns/op
	TestLoadStoreBytes.arrayScalar 16384 avgt 10 200.712 ± 8.645 ns/op
	TestLoadStoreBytes.bufferHeap 1024 avgt 10 25.664 ± 0.316 ns/op
	TestLoadStoreBytes.bufferHeap 16384 avgt 10 385.964 ± 12.450 ns/op
	TestLoadStoreBytes.bufferHeapScalar 1024 avgt 10 265.677 ± 3.302 ns/op
	TestLoadStoreBytes.bufferHeapScalar 16384 avgt 10 4163.773 ± 72.358 ns/op
	TestLoadStoreBytes.bufferNative 1024 avgt 10 25.556 ± 0.390 ns/op
	TestLoadStoreBytes.bufferNative 16384 avgt 10 349.823 ± 4.298 ns/op
	TestLoadStoreBytes.bufferNativeScalar 1024 avgt 10 252.061 ± 3.130 ns/op
	TestLoadStoreBytes.bufferNativeScalar 16384 avgt 10 4062.956 ± 70.301 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 1024 avgt 10 63.998 ± 0.425 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 16384 avgt 10 806.393 ± 10.080 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 1024 avgt 10 25.539 ± 0.268 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 16384 avgt 10 349.869 ± 4.381 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 1024 avgt 10 754.664 ± 21.680 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 16384 avgt 10 11887.237 ± 189.345 ns/op


	# VM options: -XX:+TieredCompilation -XX:ObjectAlignmentInBytes=32 --add-modules=jdk.incubator.foreign,jdk.incubator.vector -Dforeign.restricted=permit --enable-native-access ALL-UNNAMED
	Benchmark (size) Mode Cnt Score Error Units
	TestLoadStoreBytes.array 1024 avgt 10 14.034 ± 0.176 ns/op
	TestLoadStoreBytes.array 16384 avgt 10 252.624 ± 10.333 ns/op
	TestLoadStoreBytes.arrayAdd 1024 avgt 10 15.911 ± 0.523 ns/op
	TestLoadStoreBytes.arrayAdd 16384 avgt 10 271.969 ± 5.597 ns/op
	TestLoadStoreBytes.arrayScalar 1024 avgt 10 23.344 ± 0.535 ns/op
	TestLoadStoreBytes.arrayScalar 16384 avgt 10 195.957 ± 10.636 ns/op
	TestLoadStoreBytes.bufferHeap 1024 avgt 10 25.644 ± 0.281 ns/op
	TestLoadStoreBytes.bufferHeap 16384 avgt 10 388.989 ± 4.021 ns/op
	TestLoadStoreBytes.bufferHeapScalar 1024 avgt 10 266.135 ± 3.611 ns/op
	TestLoadStoreBytes.bufferHeapScalar 16384 avgt 10 4158.670 ± 63.424 ns/op
	TestLoadStoreBytes.bufferNative 1024 avgt 10 25.678 ± 0.332 ns/op
	TestLoadStoreBytes.bufferNative 16384 avgt 10 349.329 ± 3.850 ns/op
	TestLoadStoreBytes.bufferNativeScalar 1024 avgt 10 251.668 ± 3.311 ns/op
	TestLoadStoreBytes.bufferNativeScalar 16384 avgt 10 4045.838 ± 129.280 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 1024 avgt 10 34.934 ± 0.257 ns/op
	TestLoadStoreBytes.bufferSegmentConfined 16384 avgt 10 286.946 ± 6.120 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 1024 avgt 10 25.681 ± 0.501 ns/op
	TestLoadStoreBytes.bufferSegmentImplicit 16384 avgt 10 351.099 ± 10.846 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 1024 avgt 10 750.972 ± 9.326 ns/op
	TestLoadStoreBytes.segmentImplicitScalar 16384 avgt 10 11864.784 ± 135.626 ns/op
	*/