Mercurial > hg > graal-compiler
view src/share/vm/gc_implementation/shared/mutableNUMASpace.hpp @ 266:2214b226b7f0
6724367: par compact could clear less young gen summary data
Reviewed-by: jmasa, apetrusenko
author | jcoomes |
---|---|
date | Fri, 11 Jul 2008 16:11:50 -0700 |
parents | 12eea04c8b06 |
children | d6340ab4105b |
line wrap: on
line source
/* * Copyright 2006-2007 Sun Microsystems, 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, * CA 95054 USA or visit www.sun.com if you need additional information or * have any questions. * */ /* * The NUMA-aware allocator (MutableNUMASpace) is basically a modification * of MutableSpace which preserves interfaces but implements different * functionality. The space is split into chunks for each locality group * (resizing for adaptive size policy is also supported). For each thread * allocations are performed in the chunk corresponding to the home locality * group of the thread. Whenever any chunk fills-in the young generation * collection occurs. * The chunks can be also be adaptively resized. The idea behind the adaptive * sizing is to reduce the loss of the space in the eden due to fragmentation. * The main cause of fragmentation is uneven allocation rates of threads. * The allocation rate difference between locality groups may be caused either by * application specifics or by uneven LWP distribution by the OS. Besides, * application can have less threads then the number of locality groups. * In order to resize the chunk we measure the allocation rate of the * application between collections. After that we reshape the chunks to reflect * the allocation rate pattern. The AdaptiveWeightedAverage exponentially * decaying average is used to smooth the measurements. The NUMASpaceResizeRate * parameter is used to control the adaptation speed by restricting the number of * bytes that can be moved during the adaptation phase. * Chunks may contain pages from a wrong locality group. The page-scanner has * been introduced to address the problem. Remote pages typically appear due to * the memory shortage in the target locality group. Besides Solaris would * allocate a large page from the remote locality group even if there are small * local pages available. The page-scanner scans the pages right after the * collection and frees remote pages in hope that subsequent reallocation would * be more successful. This approach proved to be useful on systems with high * load where multiple processes are competing for the memory. */ class MutableNUMASpace : public MutableSpace { friend class VMStructs; class LGRPSpace : public CHeapObj { int _lgrp_id; MutableSpace* _space; MemRegion _invalid_region; AdaptiveWeightedAverage *_alloc_rate; struct SpaceStats { size_t _local_space, _remote_space, _unbiased_space, _uncommited_space; size_t _large_pages, _small_pages; SpaceStats() { _local_space = 0; _remote_space = 0; _unbiased_space = 0; _uncommited_space = 0; _large_pages = 0; _small_pages = 0; } }; SpaceStats _space_stats; char* _last_page_scanned; char* last_page_scanned() { return _last_page_scanned; } void set_last_page_scanned(char* p) { _last_page_scanned = p; } public: LGRPSpace(int l) : _lgrp_id(l), _last_page_scanned(NULL) { _space = new MutableSpace(); _alloc_rate = new AdaptiveWeightedAverage(NUMAChunkResizeWeight); } ~LGRPSpace() { delete _space; delete _alloc_rate; } void add_invalid_region(MemRegion r) { if (!_invalid_region.is_empty()) { _invalid_region.set_start(MIN2(_invalid_region.start(), r.start())); _invalid_region.set_end(MAX2(_invalid_region.end(), r.end())); } else { _invalid_region = r; } } static bool equals(void* lgrp_id_value, LGRPSpace* p) { return *(int*)lgrp_id_value == p->lgrp_id(); } void sample() { alloc_rate()->sample(space()->used_in_bytes()); } MemRegion invalid_region() const { return _invalid_region; } void set_invalid_region(MemRegion r) { _invalid_region = r; } int lgrp_id() const { return _lgrp_id; } MutableSpace* space() const { return _space; } AdaptiveWeightedAverage* alloc_rate() const { return _alloc_rate; } SpaceStats* space_stats() { return &_space_stats; } void clear_space_stats() { _space_stats = SpaceStats(); } void accumulate_statistics(size_t page_size); void scan_pages(size_t page_size, size_t page_count); }; GrowableArray<LGRPSpace*>* _lgrp_spaces; size_t _page_size; unsigned _adaptation_cycles, _samples_count; void set_page_size(size_t psz) { _page_size = psz; } size_t page_size() const { return _page_size; } unsigned adaptation_cycles() { return _adaptation_cycles; } void set_adaptation_cycles(int v) { _adaptation_cycles = v; } unsigned samples_count() { return _samples_count; } void increment_samples_count() { ++_samples_count; } size_t _base_space_size; void set_base_space_size(size_t v) { _base_space_size = v; } size_t base_space_size() const { return _base_space_size; } // Check if the NUMA topology has changed. Add and remove spaces if needed. // The update can be forced by setting the force parameter equal to true. bool update_layout(bool force); // Bias region towards the lgrp. void bias_region(MemRegion mr, int lgrp_id); // Free pages in a given region. void free_region(MemRegion mr); // Get current chunk size. size_t current_chunk_size(int i); // Get default chunk size (equally divide the space). size_t default_chunk_size(); // Adapt the chunk size to follow the allocation rate. size_t adaptive_chunk_size(int i, size_t limit); // Scan and free invalid pages. void scan_pages(size_t page_count); // Return the bottom_region and the top_region. Align them to page_size() boundary. // |------------------new_region---------------------------------| // |----bottom_region--|---intersection---|------top_region------| void select_tails(MemRegion new_region, MemRegion intersection, MemRegion* bottom_region, MemRegion *top_region); // Try to merge the invalid region with the bottom or top region by decreasing // the intersection area. Return the invalid_region aligned to the page_size() // boundary if it's inside the intersection. Return non-empty invalid_region // if it lies inside the intersection (also page-aligned). // |------------------new_region---------------------------------| // |----------------|-------invalid---|--------------------------| // |----bottom_region--|---intersection---|------top_region------| void merge_regions(MemRegion new_region, MemRegion* intersection, MemRegion *invalid_region); public: GrowableArray<LGRPSpace*>* lgrp_spaces() const { return _lgrp_spaces; } MutableNUMASpace(); virtual ~MutableNUMASpace(); // Space initialization. virtual void initialize(MemRegion mr, bool clear_space, bool mangle_space); // Update space layout if necessary. Do all adaptive resizing job. virtual void update(); // Update allocation rate averages. virtual void accumulate_statistics(); virtual void clear(bool mangle_space); virtual void mangle_unused_area() PRODUCT_RETURN; virtual void mangle_unused_area_complete() PRODUCT_RETURN; virtual void mangle_region(MemRegion mr) PRODUCT_RETURN; virtual void check_mangled_unused_area(HeapWord* limit) PRODUCT_RETURN; virtual void check_mangled_unused_area_complete() PRODUCT_RETURN; virtual void set_top_for_allocations(HeapWord* v) PRODUCT_RETURN; virtual void set_top_for_allocations() PRODUCT_RETURN; virtual void ensure_parsability(); virtual size_t used_in_words() const; virtual size_t free_in_words() const; virtual size_t tlab_capacity(Thread* thr) const; virtual size_t unsafe_max_tlab_alloc(Thread* thr) const; // Allocation (return NULL if full) virtual HeapWord* allocate(size_t word_size); virtual HeapWord* cas_allocate(size_t word_size); // Debugging virtual void print_on(outputStream* st) const; virtual void print_short_on(outputStream* st) const; virtual void verify(bool allow_dirty); virtual void set_top(HeapWord* value); };