• file
• revisions
• annotate
• diff
• comparison
• raw

src/share/vm/gc_implementation/shared/gcUtil.cpp@b2ef234911c9 (annotated)

src/share/vm/gc_implementation/shared/gcUtil.cpp

Thu, 20 Sep 2012 09:52:56 -0700

author

johnc

date

Thu, 20 Sep 2012 09:52:56 -0700

changeset 4067

b2ef234911c9

parent 3763

78a1b285cda8

child 4153

b9a9ed0f8eeb

permissions

-rw-r--r--

7190666: G1: assert(_unused == 0) failed: Inconsistency in PLAB stats
Summary: Reset the fields in ParGCAllocBuffer, that are used for accumulating values for the ResizePLAB sensors in PLABStats, to zero after flushing the values to the PLABStats fields. Flush PLABStats values only when retiring the final allocation buffers prior to disposing of a G1ParScanThreadState object, rather than when retiring every allocation buffer.
Reviewed-by: jwilhelm, jmasa, ysr

duke@435	1	/*
trims@2708	2	* Copyright (c) 2002, 2011, Oracle and/or its affiliates. All rights reserved.
duke@435	3	* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
duke@435	4	*
duke@435	5	* This code is free software; you can redistribute it and/or modify it
duke@435	6	* under the terms of the GNU General Public License version 2 only, as
duke@435	7	* published by the Free Software Foundation.
duke@435	8	*
duke@435	9	* This code is distributed in the hope that it will be useful, but WITHOUT
duke@435	10	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
duke@435	11	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
duke@435	12	* version 2 for more details (a copy is included in the LICENSE file that
duke@435	13	* accompanied this code).
duke@435	14	*
duke@435	15	* You should have received a copy of the GNU General Public License version
duke@435	16	* 2 along with this work; if not, write to the Free Software Foundation,
duke@435	17	* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
duke@435	18	*
trims@1907	19	* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
trims@1907	20	* or visit www.oracle.com if you need additional information or have any
trims@1907	21	* questions.
duke@435	22	*
duke@435	23	*/
duke@435	24
stefank@2314	25	#include "precompiled.hpp"
stefank@2314	26	#include "gc_implementation/shared/gcUtil.hpp"
duke@435	27
duke@435	28	// Catch-all file for utility classes
duke@435	29
duke@435	30	float AdaptiveWeightedAverage::compute_adaptive_average(float new_sample,
duke@435	31	float average) {
duke@435	32	// We smooth the samples by not using weight() directly until we've
duke@435	33	// had enough data to make it meaningful. We'd like the first weight
mikael@3763	34	// used to be 1, the second to be 1/2, etc until we have
mikael@3763	35	// OLD_THRESHOLD/weight samples.
mikael@3763	36	unsigned count_weight = 0;
mikael@3763	37
mikael@3763	38	// Avoid division by zero if the counter wraps (7158457)
mikael@3763	39	if (!is_old()) {
mikael@3763	40	count_weight = OLD_THRESHOLD/count();
mikael@3763	41	}
mikael@3763	42
duke@435	43	unsigned adaptive_weight = (MAX2(weight(), count_weight));
duke@435	44
duke@435	45	float new_avg = exp_avg(average, new_sample, adaptive_weight);
duke@435	46
duke@435	47	return new_avg;
duke@435	48	}
duke@435	49
duke@435	50	void AdaptiveWeightedAverage::sample(float new_sample) {
duke@435	51	increment_count();
duke@435	52
duke@435	53	// Compute the new weighted average
duke@435	54	float new_avg = compute_adaptive_average(new_sample, average());
duke@435	55	set_average(new_avg);
duke@435	56	_last_sample = new_sample;
duke@435	57	}
duke@435	58
ysr@1580	59	void AdaptiveWeightedAverage::print() const {
ysr@1580	60	print_on(tty);
ysr@1580	61	}
ysr@1580	62
ysr@1580	63	void AdaptiveWeightedAverage::print_on(outputStream* st) const {
ysr@1580	64	guarantee(false, "NYI");
ysr@1580	65	}
ysr@1580	66
ysr@1580	67	void AdaptivePaddedAverage::print() const {
ysr@1580	68	print_on(tty);
ysr@1580	69	}
ysr@1580	70
ysr@1580	71	void AdaptivePaddedAverage::print_on(outputStream* st) const {
ysr@1580	72	guarantee(false, "NYI");
ysr@1580	73	}
ysr@1580	74
ysr@1580	75	void AdaptivePaddedNoZeroDevAverage::print() const {
ysr@1580	76	print_on(tty);
ysr@1580	77	}
ysr@1580	78
ysr@1580	79	void AdaptivePaddedNoZeroDevAverage::print_on(outputStream* st) const {
ysr@1580	80	guarantee(false, "NYI");
ysr@1580	81	}
ysr@1580	82
duke@435	83	void AdaptivePaddedAverage::sample(float new_sample) {
ysr@1580	84	// Compute new adaptive weighted average based on new sample.
duke@435	85	AdaptiveWeightedAverage::sample(new_sample);
duke@435	86
ysr@1580	87	// Now update the deviation and the padded average.
duke@435	88	float new_avg = average();
duke@435	89	float new_dev = compute_adaptive_average(fabsd(new_sample - new_avg),
duke@435	90	deviation());
duke@435	91	set_deviation(new_dev);
duke@435	92	set_padded_average(new_avg + padding() * new_dev);
duke@435	93	_last_sample = new_sample;
duke@435	94	}
duke@435	95
duke@435	96	void AdaptivePaddedNoZeroDevAverage::sample(float new_sample) {
duke@435	97	// Compute our parent classes sample information
duke@435	98	AdaptiveWeightedAverage::sample(new_sample);
duke@435	99
duke@435	100	float new_avg = average();
duke@435	101	if (new_sample != 0) {
duke@435	102	// We only create a new deviation if the sample is non-zero
duke@435	103	float new_dev = compute_adaptive_average(fabsd(new_sample - new_avg),
duke@435	104	deviation());
duke@435	105
duke@435	106	set_deviation(new_dev);
duke@435	107	}
duke@435	108	set_padded_average(new_avg + padding() * deviation());
duke@435	109	_last_sample = new_sample;
duke@435	110	}
duke@435	111
duke@435	112	LinearLeastSquareFit::LinearLeastSquareFit(unsigned weight) :
phh@2505	113	_sum_x(0), _sum_x_squared(0), _sum_y(0), _sum_xy(0),
phh@2505	114	_intercept(0), _slope(0), _mean_x(weight), _mean_y(weight) {}
duke@435	115
duke@435	116	void LinearLeastSquareFit::update(double x, double y) {
duke@435	117	_sum_x = _sum_x + x;
duke@435	118	_sum_x_squared = _sum_x_squared + x * x;
duke@435	119	_sum_y = _sum_y + y;
duke@435	120	_sum_xy = _sum_xy + x * y;
duke@435	121	_mean_x.sample(x);
duke@435	122	_mean_y.sample(y);
duke@435	123	assert(_mean_x.count() == _mean_y.count(), "Incorrect count");
duke@435	124	if ( _mean_x.count() > 1 ) {
duke@435	125	double slope_denominator;
duke@435	126	slope_denominator = (_mean_x.count() * _sum_x_squared - _sum_x * _sum_x);
duke@435	127	// Some tolerance should be injected here. A denominator that is
duke@435	128	// nearly 0 should be avoided.
duke@435	129
duke@435	130	if (slope_denominator != 0.0) {
duke@435	131	double slope_numerator;
duke@435	132	slope_numerator = (_mean_x.count() * _sum_xy - _sum_x * _sum_y);
duke@435	133	_slope = slope_numerator / slope_denominator;
duke@435	134
duke@435	135	// The _mean_y and _mean_x are decaying averages and can
duke@435	136	// be used to discount earlier data. If they are used,
duke@435	137	// first consider whether all the quantities should be
duke@435	138	// kept as decaying averages.
duke@435	139	// _intercept = _mean_y.average() - _slope * _mean_x.average();
duke@435	140	_intercept = (_sum_y - _slope * _sum_x) / ((double) _mean_x.count());
duke@435	141	}
duke@435	142	}
duke@435	143	}
duke@435	144
duke@435	145	double LinearLeastSquareFit::y(double x) {
duke@435	146	double new_y;
duke@435	147
duke@435	148	if ( _mean_x.count() > 1 ) {
duke@435	149	new_y = (_intercept + _slope * x);
duke@435	150	return new_y;
duke@435	151	} else {
duke@435	152	return _mean_y.average();
duke@435	153	}
duke@435	154	}
duke@435	155
duke@435	156	// Both decrement_will_decrease() and increment_will_decrease() return
duke@435	157	// true for a slope of 0. That is because a change is necessary before
duke@435	158	// a slope can be calculated and a 0 slope will, in general, indicate
duke@435	159	// that no calculation of the slope has yet been done. Returning true
duke@435	160	// for a slope equal to 0 reflects the intuitive expectation of the
duke@435	161	// dependence on the slope. Don't use the complement of these functions
duke@435	162	// since that untuitive expectation is not built into the complement.
duke@435	163	bool LinearLeastSquareFit::decrement_will_decrease() {
duke@435	164	return (_slope >= 0.00);
duke@435	165	}
duke@435	166
duke@435	167	bool LinearLeastSquareFit::increment_will_decrease() {
duke@435	168	return (_slope <= 0.00);
duke@435	169	}