/*
* Copyright (c) 2018, 2019, Red Hat, Inc. 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.
*
*/
#include "precompiled.hpp"
#include "gc/shared/gcArguments.hpp"
#include "gc/shared/workerPolicy.hpp"
#include "gc/shenandoah/shenandoahArguments.hpp"
#include "gc/shenandoah/shenandoahCollectorPolicy.hpp"
#include "gc/shenandoah/shenandoahHeap.inline.hpp"
#include "gc/shenandoah/shenandoahHeapRegion.hpp"
#include "utilities/defaultStream.hpp"
void ShenandoahArguments::initialize() {
#if !(defined AARCH64 || defined AMD64 || defined IA32)
vm_exit_during_initialization("Shenandoah GC is not supported on this platform.");
#endif
#if 0 // leave this block as stepping stone for future platforms
log_warning(gc)("Shenandoah GC is not fully supported on this platform:");
log_warning(gc)(" concurrent modes are not supported, only STW cycles are enabled;");
log_warning(gc)(" arch-specific barrier code is not implemented, disabling barriers;");
FLAG_SET_DEFAULT(ShenandoahGCHeuristics, "passive");
FLAG_SET_DEFAULT(ShenandoahSATBBarrier, false);
FLAG_SET_DEFAULT(ShenandoahLoadRefBarrier, false);
FLAG_SET_DEFAULT(ShenandoahKeepAliveBarrier, false);
FLAG_SET_DEFAULT(ShenandoahStoreValEnqueueBarrier, false);
FLAG_SET_DEFAULT(ShenandoahCASBarrier, false);
FLAG_SET_DEFAULT(ShenandoahCloneBarrier, false);
FLAG_SET_DEFAULT(ShenandoahVerifyOptoBarriers, false);
#endif
if (UseLargePages && (MaxHeapSize / os::large_page_size()) < ShenandoahHeapRegion::MIN_NUM_REGIONS) {
warning("Large pages size (" SIZE_FORMAT "K) is too large to afford page-sized regions, disabling uncommit",
os::large_page_size() / K);
FLAG_SET_DEFAULT(ShenandoahUncommit, false);
}
// Enable NUMA by default. While Shenandoah is not NUMA-aware, enabling NUMA makes
// storage allocation code NUMA-aware.
if (FLAG_IS_DEFAULT(UseNUMA)) {
FLAG_SET_DEFAULT(UseNUMA, true);
}
// Set up default number of concurrent threads. We want to have cycles complete fast
// enough, but we also do not want to steal too much CPU from the concurrently running
// application. Using 1/4 of available threads for concurrent GC seems a good
// compromise here.
bool ergo_conc = FLAG_IS_DEFAULT(ConcGCThreads);
if (ergo_conc) {
FLAG_SET_DEFAULT(ConcGCThreads, MAX2(1, os::initial_active_processor_count() / 4));
}
if (ConcGCThreads == 0) {
vm_exit_during_initialization("Shenandoah expects ConcGCThreads > 0, check -XX:ConcGCThreads=#");
}
// Set up default number of parallel threads. We want to have decent pauses performance
// which would use parallel threads, but we also do not want to do too many threads
// that will overwhelm the OS scheduler. Using 1/2 of available threads seems to be a fair
// compromise here. Due to implementation constraints, it should not be lower than
// the number of concurrent threads.
bool ergo_parallel = FLAG_IS_DEFAULT(ParallelGCThreads);
if (ergo_parallel) {
FLAG_SET_DEFAULT(ParallelGCThreads, MAX2(1, os::initial_active_processor_count() / 2));
}
if (ParallelGCThreads == 0) {
vm_exit_during_initialization("Shenandoah expects ParallelGCThreads > 0, check -XX:ParallelGCThreads=#");
}
// Make sure ergonomic decisions do not break the thread count invariants.
// This may happen when user overrides one of the flags, but not the other.
// When that happens, we want to adjust the setting that was set ergonomically.
if (ParallelGCThreads < ConcGCThreads) {
if (ergo_conc && !ergo_parallel) {
FLAG_SET_DEFAULT(ConcGCThreads, ParallelGCThreads);
} else if (!ergo_conc && ergo_parallel) {
FLAG_SET_DEFAULT(ParallelGCThreads, ConcGCThreads);
} else if (ergo_conc && ergo_parallel) {
// Should not happen, check the ergonomic computation above. Fail with relevant error.
vm_exit_during_initialization("Shenandoah thread count ergonomic error");
} else {
// User settings error, report and ask user to rectify.
vm_exit_during_initialization("Shenandoah expects ConcGCThreads <= ParallelGCThreads, check -XX:ParallelGCThreads, -XX:ConcGCThreads");
}
}
if (FLAG_IS_DEFAULT(ParallelRefProcEnabled)) {
FLAG_SET_DEFAULT(ParallelRefProcEnabled, true);
}
if (ShenandoahRegionSampling && FLAG_IS_DEFAULT(PerfDataMemorySize)) {
// When sampling is enabled, max out the PerfData memory to get more
// Shenandoah data in, including Matrix.
FLAG_SET_DEFAULT(PerfDataMemorySize, 2048*K);
}
#ifdef COMPILER2
// Shenandoah cares more about pause times, rather than raw throughput.
if (FLAG_IS_DEFAULT(UseCountedLoopSafepoints)) {
FLAG_SET_DEFAULT(UseCountedLoopSafepoints, true);
if (FLAG_IS_DEFAULT(LoopStripMiningIter)) {
FLAG_SET_DEFAULT(LoopStripMiningIter, 1000);
}
}
#ifdef ASSERT
// C2 barrier verification is only reliable when all default barriers are enabled
if (ShenandoahVerifyOptoBarriers &&
(!FLAG_IS_DEFAULT(ShenandoahSATBBarrier) ||
!FLAG_IS_DEFAULT(ShenandoahLoadRefBarrier) ||
!FLAG_IS_DEFAULT(ShenandoahKeepAliveBarrier) ||
!FLAG_IS_DEFAULT(ShenandoahStoreValEnqueueBarrier) ||
!FLAG_IS_DEFAULT(ShenandoahCASBarrier) ||
!FLAG_IS_DEFAULT(ShenandoahCloneBarrier)
)) {
warning("Unusual barrier configuration, disabling C2 barrier verification");
FLAG_SET_DEFAULT(ShenandoahVerifyOptoBarriers, false);
}
#else
guarantee(!ShenandoahVerifyOptoBarriers, "Should be disabled");
#endif // ASSERT
#endif // COMPILER2
if (AlwaysPreTouch) {
// Shenandoah handles pre-touch on its own. It does not let the
// generic storage code to do the pre-touch before Shenandoah has
// a chance to do it on its own.
FLAG_SET_DEFAULT(AlwaysPreTouch, false);
FLAG_SET_DEFAULT(ShenandoahAlwaysPreTouch, true);
}
// Record more information about previous cycles for improved debugging pleasure
if (FLAG_IS_DEFAULT(LogEventsBufferEntries)) {
FLAG_SET_DEFAULT(LogEventsBufferEntries, 250);
}
if (ShenandoahAlwaysPreTouch) {
if (!FLAG_IS_DEFAULT(ShenandoahUncommit)) {
warning("AlwaysPreTouch is enabled, disabling ShenandoahUncommit");
}
FLAG_SET_DEFAULT(ShenandoahUncommit, false);
}
if ((InitialHeapSize == MaxHeapSize) && ShenandoahUncommit) {
log_info(gc)("Min heap equals to max heap, disabling ShenandoahUncommit");
FLAG_SET_DEFAULT(ShenandoahUncommit, false);
}
// If class unloading is disabled, no unloading for concurrent cycles as well.
// If class unloading is enabled, users should opt-in for unloading during
// concurrent cycles.
if (!ClassUnloading || !FLAG_IS_CMDLINE(ClassUnloadingWithConcurrentMark)) {
log_info(gc)("Consider -XX:+ClassUnloadingWithConcurrentMark if large pause times "
"are observed on class-unloading sensitive workloads");
FLAG_SET_DEFAULT(ClassUnloadingWithConcurrentMark, false);
}
// AOT is not supported yet
if (UseAOT) {
if (!FLAG_IS_DEFAULT(UseAOT)) {
warning("Shenandoah does not support AOT at this moment, disabling UseAOT");
}
FLAG_SET_DEFAULT(UseAOT, false);
}
// TLAB sizing policy makes resizing decisions before each GC cycle. It averages
// historical data, assigning more recent data the weight according to TLABAllocationWeight.
// Current default is good for generational collectors that run frequent young GCs.
// With Shenandoah, GC cycles are much less frequent, so we need we need sizing policy
// to converge faster over smaller number of resizing decisions.
if (FLAG_IS_DEFAULT(TLABAllocationWeight)) {
FLAG_SET_DEFAULT(TLABAllocationWeight, 90);
}
// Shenandoah needs more C2 nodes to compile some methods with lots of barriers.
// NodeLimitFudgeFactor needs to stay the same relative to MaxNodeLimit.
#ifdef COMPILER2
if (FLAG_IS_DEFAULT(MaxNodeLimit)) {
FLAG_SET_DEFAULT(MaxNodeLimit, MaxNodeLimit * 3);
FLAG_SET_DEFAULT(NodeLimitFudgeFactor, NodeLimitFudgeFactor * 3);
}
#endif
// Make sure safepoint deadlocks are failing predictably. This sets up VM to report
// fatal error after 10 seconds of wait for safepoint syncronization (not the VM
// operation itself). There is no good reason why Shenandoah would spend that
// much time synchronizing.
#ifdef ASSERT
FLAG_SET_DEFAULT(SafepointTimeout, true);
FLAG_SET_DEFAULT(SafepointTimeoutDelay, 10000);
FLAG_SET_DEFAULT(AbortVMOnSafepointTimeout, true);
#endif
}
size_t ShenandoahArguments::conservative_max_heap_alignment() {
size_t align = ShenandoahMaxRegionSize;
if (UseLargePages) {
align = MAX2(align, os::large_page_size());
}
return align;
}
void ShenandoahArguments::initialize_alignments() {
// Need to setup sizes early to get correct alignments.
ShenandoahHeapRegion::setup_sizes(MaxHeapSize);
// This is expected by our algorithm for ShenandoahHeap::heap_region_containing().
size_t align = ShenandoahHeapRegion::region_size_bytes();
if (UseLargePages) {
align = MAX2(align, os::large_page_size());
}
SpaceAlignment = align;
HeapAlignment = align;
}
CollectedHeap* ShenandoahArguments::create_heap() {
return new ShenandoahHeap(new ShenandoahCollectorPolicy());
}