Mercurial > hg > truffle
view test/gc/7168848/HumongousAlloc.java @ 11825:463f51256c86
AMD64HotSpot: emit jump to IC miss handler directly instead of emitting it inside verfied_entry
for empty methods we emitted something like this if a inline cache is needed:
prefix:
00: < IC check >
...
0b: jne <ic_miss_call>
11: nop
...
verified_entry:
20: ret
ic_miss_call:
21: jmp <entry of runtime function>
when a method is deoptimized, HotSpot patches the verified_entry (0x20) with a
jump to a stub that handles call-sites that has been made non-entrant. since this jump
is 5 bytes long, it will overwrite ic_miss_call and blow up every caller that
calls this method via the unverified entry (prefix).
the fix is to emit the jump to the runtime function inside the unverfied entry:
prefix:
00: < IC check >
...
0b: je <verified_entry>
11: jeq <entry of runtime function>
16: nop
...
verified_entry:
20: ret
author | Bernhard Urban <bernhard.urban@jku.at> |
---|---|
date | Mon, 30 Sep 2013 09:32:18 +0200 |
parents | 9a344d88dc22 |
children |
line wrap: on
line source
/* * Copyright (c) 2012, 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 Humongous.java * @bug 7168848 * @summary G1: humongous object allocations should initiate marking cycles when necessary * @run main/othervm -Xms100m -Xmx100m -XX:+PrintGC -XX:G1HeapRegionSize=1m -XX:+UseG1GC HumongousAlloc * */ import java.lang.management.GarbageCollectorMXBean; import java.lang.management.ManagementFactory; import java.util.List; public class HumongousAlloc { public static byte[] dummy; private static int sleepFreq = 40; private static int sleepTime = 1000; private static double size = 0.75; private static int iterations = 50; private static int MB = 1024 * 1024; public static void allocate(int size, int sleepTime, int sleepFreq) throws InterruptedException { System.out.println("Will allocate objects of size: " + size + " bytes and sleep for " + sleepTime + " ms after every " + sleepFreq + "th allocation."); int count = 0; while (count < iterations) { for (int i = 0; i < sleepFreq; i++) { dummy = new byte[size - 16]; } Thread.sleep(sleepTime); count++; } } public static void main(String[] args) throws InterruptedException { allocate((int) (size * MB), sleepTime, sleepFreq); List<GarbageCollectorMXBean> collectors = ManagementFactory.getGarbageCollectorMXBeans(); for (GarbageCollectorMXBean collector : collectors) { if (collector.getName().contains("G1 Old")) { long count = collector.getCollectionCount(); if (count > 0) { throw new RuntimeException("Failed: FullGCs should not have happened. The number of FullGC run is " + count); } else { System.out.println("Passed."); } } } } }