2008-06-05 15:57:56 -07:00
|
|
|
/*
|
2014-05-09 16:50:54 -04:00
|
|
|
* Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
|
2008-06-05 15:57:56 -07:00
|
|
|
* 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.
|
|
|
|
*
|
2010-05-27 19:08:38 -07:00
|
|
|
* 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.
|
2008-06-05 15:57:56 -07:00
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2010-11-23 13:22:55 -08:00
|
|
|
#include "precompiled.hpp"
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
#include "gc_implementation/g1/heapRegion.hpp"
|
|
|
|
#include "gc_implementation/g1/heapRegionSeq.inline.hpp"
|
2014-03-14 10:15:46 +01:00
|
|
|
#include "gc_implementation/g1/heapRegionSet.hpp"
|
2010-11-23 13:22:55 -08:00
|
|
|
#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
|
|
|
|
#include "memory/allocation.hpp"
|
2008-06-05 15:57:56 -07:00
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
// Private
|
2008-06-05 15:57:56 -07:00
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint HeapRegionSeq::find_contiguous_from(uint from, uint num) {
|
|
|
|
uint len = length();
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(num > 1, "use this only for sequences of length 2 or greater");
|
|
|
|
assert(from <= len,
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("from: %u should be valid and <= than %u", from, len));
|
2008-06-05 15:57:56 -07:00
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint curr = from;
|
|
|
|
uint first = G1_NULL_HRS_INDEX;
|
|
|
|
uint num_so_far = 0;
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
while (curr < len && num_so_far < num) {
|
|
|
|
if (at(curr)->is_empty()) {
|
|
|
|
if (first == G1_NULL_HRS_INDEX) {
|
2011-01-19 19:30:42 -05:00
|
|
|
first = curr;
|
|
|
|
num_so_far = 1;
|
|
|
|
} else {
|
|
|
|
num_so_far += 1;
|
|
|
|
}
|
|
|
|
} else {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
first = G1_NULL_HRS_INDEX;
|
2011-01-19 19:30:42 -05:00
|
|
|
num_so_far = 0;
|
|
|
|
}
|
|
|
|
curr += 1;
|
|
|
|
}
|
|
|
|
assert(num_so_far <= num, "post-condition");
|
|
|
|
if (num_so_far == num) {
|
2011-03-04 17:13:19 -05:00
|
|
|
// we found enough space for the humongous object
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(from <= first && first < len, "post-condition");
|
|
|
|
assert(first < curr && (curr - first) == num, "post-condition");
|
2012-04-18 07:21:15 -04:00
|
|
|
for (uint i = first; i < first + num; ++i) {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(at(i)->is_empty(), "post-condition");
|
2011-01-19 19:30:42 -05:00
|
|
|
}
|
|
|
|
return first;
|
|
|
|
} else {
|
|
|
|
// we failed to find enough space for the humongous object
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
return G1_NULL_HRS_INDEX;
|
2011-01-19 19:30:42 -05:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
// Public
|
2011-01-19 19:30:42 -05:00
|
|
|
|
2013-09-25 13:25:24 +02:00
|
|
|
void HeapRegionSeq::initialize(HeapWord* bottom, HeapWord* end) {
|
2012-04-18 07:21:15 -04:00
|
|
|
assert((uintptr_t) bottom % HeapRegion::GrainBytes == 0,
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
"bottom should be heap region aligned");
|
2012-04-18 07:21:15 -04:00
|
|
|
assert((uintptr_t) end % HeapRegion::GrainBytes == 0,
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
"end should be heap region aligned");
|
2008-06-05 15:57:56 -07:00
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
_next_search_index = 0;
|
|
|
|
_allocated_length = 0;
|
|
|
|
|
2013-09-25 13:25:24 +02:00
|
|
|
_regions.initialize(bottom, end, HeapRegion::GrainBytes);
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
}
|
2008-06-05 15:57:56 -07:00
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
MemRegion HeapRegionSeq::expand_by(HeapWord* old_end,
|
|
|
|
HeapWord* new_end,
|
|
|
|
FreeRegionList* list) {
|
|
|
|
assert(old_end < new_end, "don't call it otherwise");
|
|
|
|
G1CollectedHeap* g1h = G1CollectedHeap::heap();
|
2008-06-05 15:57:56 -07:00
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
HeapWord* next_bottom = old_end;
|
2013-09-25 13:25:24 +02:00
|
|
|
assert(heap_bottom() <= next_bottom, "invariant");
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
while (next_bottom < new_end) {
|
2013-09-25 13:25:24 +02:00
|
|
|
assert(next_bottom < heap_end(), "invariant");
|
2012-04-18 07:21:15 -04:00
|
|
|
uint index = length();
|
2008-06-05 15:57:56 -07:00
|
|
|
|
2013-09-25 13:25:24 +02:00
|
|
|
assert(index < max_length(), "otherwise we cannot expand further");
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
if (index == 0) {
|
|
|
|
// We have not allocated any regions so far
|
2013-09-25 13:25:24 +02:00
|
|
|
assert(next_bottom == heap_bottom(), "invariant");
|
2008-06-05 15:57:56 -07:00
|
|
|
} else {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
// next_bottom should match the end of the last/previous region
|
|
|
|
assert(next_bottom == at(index - 1)->end(), "invariant");
|
|
|
|
}
|
|
|
|
|
|
|
|
if (index == _allocated_length) {
|
|
|
|
// We have to allocate a new HeapRegion.
|
|
|
|
HeapRegion* new_hr = g1h->new_heap_region(index, next_bottom);
|
|
|
|
if (new_hr == NULL) {
|
|
|
|
// allocation failed, we bail out and return what we have done so far
|
|
|
|
return MemRegion(old_end, next_bottom);
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
2013-09-25 13:25:24 +02:00
|
|
|
assert(_regions.get_by_index(index) == NULL, "invariant");
|
|
|
|
_regions.set_by_index(index, new_hr);
|
2013-05-06 21:30:34 +02:00
|
|
|
increment_allocated_length();
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
// Have to increment the length first, otherwise we will get an
|
|
|
|
// assert failure at(index) below.
|
2013-05-06 21:30:34 +02:00
|
|
|
increment_length();
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
HeapRegion* hr = at(index);
|
|
|
|
list->add_as_tail(hr);
|
|
|
|
|
|
|
|
next_bottom = hr->end();
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(next_bottom == new_end, "post-condition");
|
|
|
|
return MemRegion(old_end, next_bottom);
|
|
|
|
}
|
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint HeapRegionSeq::free_suffix() {
|
|
|
|
uint res = 0;
|
|
|
|
uint index = length();
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
while (index > 0) {
|
|
|
|
index -= 1;
|
|
|
|
if (!at(index)->is_empty()) {
|
|
|
|
break;
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
res += 1;
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint HeapRegionSeq::find_contiguous(uint num) {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(num > 1, "use this only for sequences of length 2 or greater");
|
|
|
|
assert(_next_search_index <= length(),
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("_next_search_index: %u should be valid and <= than %u",
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
_next_search_index, length()));
|
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint start = _next_search_index;
|
|
|
|
uint res = find_contiguous_from(start, num);
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
if (res == G1_NULL_HRS_INDEX && start > 0) {
|
|
|
|
// Try starting from the beginning. If _next_search_index was 0,
|
|
|
|
// no point in doing this again.
|
|
|
|
res = find_contiguous_from(0, num);
|
|
|
|
}
|
|
|
|
if (res != G1_NULL_HRS_INDEX) {
|
2012-04-18 07:21:15 -04:00
|
|
|
assert(res < length(), err_msg("res: %u should be valid", res));
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
_next_search_index = res + num;
|
|
|
|
assert(_next_search_index <= length(),
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("_next_search_index: %u should be valid and <= than %u",
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
_next_search_index, length()));
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
void HeapRegionSeq::iterate(HeapRegionClosure* blk) const {
|
|
|
|
iterate_from((HeapRegion*) NULL, blk);
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
void HeapRegionSeq::iterate_from(HeapRegion* hr, HeapRegionClosure* blk) const {
|
2012-04-18 07:21:15 -04:00
|
|
|
uint hr_index = 0;
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
if (hr != NULL) {
|
2012-04-18 07:21:15 -04:00
|
|
|
hr_index = hr->hrs_index();
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
}
|
|
|
|
|
2012-04-18 07:21:15 -04:00
|
|
|
uint len = length();
|
|
|
|
for (uint i = hr_index; i < len; i += 1) {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
bool res = blk->doHeapRegion(at(i));
|
|
|
|
if (res) {
|
2008-06-05 15:57:56 -07:00
|
|
|
blk->incomplete();
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
2012-04-18 07:21:15 -04:00
|
|
|
for (uint i = 0; i < hr_index; i += 1) {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
bool res = blk->doHeapRegion(at(i));
|
|
|
|
if (res) {
|
2008-06-05 15:57:56 -07:00
|
|
|
blk->incomplete();
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-05-06 21:30:34 +02:00
|
|
|
uint HeapRegionSeq::shrink_by(uint num_regions_to_remove) {
|
2011-01-19 19:30:42 -05:00
|
|
|
// Reset this in case it's currently pointing into the regions that
|
|
|
|
// we just removed.
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
_next_search_index = 0;
|
2011-01-19 19:30:42 -05:00
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
assert(length() > 0, "the region sequence should not be empty");
|
|
|
|
assert(length() <= _allocated_length, "invariant");
|
|
|
|
assert(_allocated_length > 0, "we should have at least one region committed");
|
2013-05-06 21:30:34 +02:00
|
|
|
assert(num_regions_to_remove < length(), "We should never remove all regions");
|
2008-06-05 15:57:56 -07:00
|
|
|
|
2013-05-06 21:30:34 +02:00
|
|
|
uint i = 0;
|
|
|
|
for (; i < num_regions_to_remove; i++) {
|
|
|
|
HeapRegion* cur = at(length() - 1);
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
|
2013-05-06 21:30:34 +02:00
|
|
|
if (!cur->is_empty()) {
|
|
|
|
// We have to give up if the region can not be moved
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
assert(!cur->isHumongous(), "Humongous regions should not be empty");
|
|
|
|
|
|
|
|
decrement_length();
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
2013-05-06 21:30:34 +02:00
|
|
|
return i;
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
|
|
|
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
#ifndef PRODUCT
|
|
|
|
void HeapRegionSeq::verify_optional() {
|
2013-09-25 13:25:24 +02:00
|
|
|
guarantee(length() <= _allocated_length,
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("invariant: _length: %u _allocated_length: %u",
|
2013-09-25 13:25:24 +02:00
|
|
|
length(), _allocated_length));
|
|
|
|
guarantee(_allocated_length <= max_length(),
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("invariant: _allocated_length: %u _max_length: %u",
|
2013-09-25 13:25:24 +02:00
|
|
|
_allocated_length, max_length()));
|
|
|
|
guarantee(_next_search_index <= length(),
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("invariant: _next_search_index: %u _length: %u",
|
2013-09-25 13:25:24 +02:00
|
|
|
_next_search_index, length()));
|
2008-06-05 15:57:56 -07:00
|
|
|
|
2013-09-25 13:25:24 +02:00
|
|
|
HeapWord* prev_end = heap_bottom();
|
2012-04-18 07:21:15 -04:00
|
|
|
for (uint i = 0; i < _allocated_length; i += 1) {
|
2013-09-25 13:25:24 +02:00
|
|
|
HeapRegion* hr = _regions.get_by_index(i);
|
2012-04-18 07:21:15 -04:00
|
|
|
guarantee(hr != NULL, err_msg("invariant: i: %u", i));
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
guarantee(hr->bottom() == prev_end,
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("invariant i: %u "HR_FORMAT" prev_end: "PTR_FORMAT,
|
2014-05-09 16:50:54 -04:00
|
|
|
i, HR_FORMAT_PARAMS(hr), p2i(prev_end)));
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
guarantee(hr->hrs_index() == i,
|
2012-04-18 07:21:15 -04:00
|
|
|
err_msg("invariant: i: %u hrs_index(): %u", i, hr->hrs_index()));
|
2013-09-25 13:25:24 +02:00
|
|
|
if (i < length()) {
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
// Asserts will fire if i is >= _length
|
|
|
|
HeapWord* addr = hr->bottom();
|
|
|
|
guarantee(addr_to_region(addr) == hr, "sanity");
|
|
|
|
} else {
|
|
|
|
guarantee(hr->is_empty(), "sanity");
|
|
|
|
guarantee(!hr->isHumongous(), "sanity");
|
|
|
|
// using assert instead of guarantee here since containing_set()
|
|
|
|
// is only available in non-product builds.
|
|
|
|
assert(hr->containing_set() == NULL, "sanity");
|
|
|
|
}
|
|
|
|
if (hr->startsHumongous()) {
|
|
|
|
prev_end = hr->orig_end();
|
|
|
|
} else {
|
|
|
|
prev_end = hr->end();
|
|
|
|
}
|
|
|
|
}
|
2013-09-25 13:25:24 +02:00
|
|
|
for (uint i = _allocated_length; i < max_length(); i += 1) {
|
|
|
|
guarantee(_regions.get_by_index(i) == NULL, err_msg("invariant i: %u", i));
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
}
|
2008-06-05 15:57:56 -07:00
|
|
|
}
|
7045330: G1: Simplify/fix the HeapRegionSeq class
7042285: G1: native memory leak during humongous object allocation
6804436: G1: heap region indices should be size_t
A series of fixes and improvements to the HeapRegionSeq class: a) replace the _regions growable array with a standard C array, b) avoid de-allocating / re-allocating HeapRegion instances when the heap shrinks / grows (fix for 7042285), c) introduce fast method to map address to HeapRegion via a "biased" array pointer, d) embed the _hrs object in G1CollectedHeap, instead of pointing to it via an indirection, e) assume that all the regions added to the HeapRegionSeq instance are contiguous, f) replace int's with size_t's for indexes (and expand that to HeapRegion as part of 6804436), g) remove unnecessary / unused methods, h) rename a couple of fields (_alloc_search_start and _seq_bottom), i) fix iterate_from() not to always start from index 0 irrespective of the region passed to it, j) add a verification method to check the HeapRegionSeq assumptions, k) always call the wrappers for _hrs.iterate(), _hrs_length(), and _hrs.at() from G1CollectedHeap, not those methods directly, and l) unify the code that expands the sequence (by either re-using or creating a new HeapRegion) and make it robust wrt to a HeapRegion allocation failing.
Reviewed-by: stefank, johnc, brutisso
2011-06-10 13:16:40 -04:00
|
|
|
#endif // PRODUCT
|