1 /*

     2  * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.

     3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

     4  *

     5  * This code is free software; you can redistribute it and/or modify it

     6  * under the terms of the GNU General Public License version 2 only, as

     7  * published by the Free Software Foundation.

     8  *

     9  * This code is distributed in the hope that it will be useful, but WITHOUT

    10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

    11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

    12  * version 2 for more details (a copy is included in the LICENSE file that

    13  * accompanied this code).

    14  *

    15  * You should have received a copy of the GNU General Public License version

    16  * 2 along with this work; if not, write to the Free Software Foundation,

    17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

    18  *

    19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

    20  * or visit www.oracle.com if you need additional information or have any

    21  * questions.

    22  *

    23  */

    24 

    25 #ifndef SHARE_VM_GC_SHARED_GCUTIL_HPP

    26 #define SHARE_VM_GC_SHARED_GCUTIL_HPP

    27 

    28 #include "memory/allocation.hpp"

    29 #include "runtime/timer.hpp"

    30 #include "utilities/debug.hpp"

    31 #include "utilities/globalDefinitions.hpp"

    32 #include "utilities/ostream.hpp"

    33 

    34 // Catch-all file for utility classes

    35 

    36 // A weighted average maintains a running, weighted average

    37 // of some float value (templates would be handy here if we

    38 // need different types).

    39 //

    40 // The average is adaptive in that we smooth it for the

    41 // initial samples; we don't use the weight until we have

    42 // enough samples for it to be meaningful.

    43 //

    44 // This serves as our best estimate of a future unknown.

    45 //

    46 class AdaptiveWeightedAverage : public CHeapObj<mtGC> {

    47  private:

    48   float            _average;        // The last computed average

    49   unsigned         _sample_count;   // How often we've sampled this average

    50   unsigned         _weight;         // The weight used to smooth the averages

    51                                     //   A higher weight favors the most

    52                                     //   recent data.

    53   bool             _is_old;         // Has enough historical data

    54 

    55   const static unsigned OLD_THRESHOLD = 100;

    56 

    57  protected:

    58   float            _last_sample;    // The last value sampled.

    59 

    60   void  increment_count() {

    61     _sample_count++;

    62     if (!_is_old && _sample_count > OLD_THRESHOLD) {

    63       _is_old = true;

    64     }

    65   }

    66 

    67   void  set_average(float avg)  { _average = avg;        }

    68 

    69   // Helper function, computes an adaptive weighted average

    70   // given a sample and the last average

    71   float compute_adaptive_average(float new_sample, float average);

    72 

    73  public:

    74   // Input weight must be between 0 and 100

    75   AdaptiveWeightedAverage(unsigned weight, float avg = 0.0) :

    76     _average(avg), _sample_count(0), _weight(weight), _last_sample(0.0),

    77     _is_old(false) {

    78   }

    79 

    80   void clear() {

    81     _average = 0;

    82     _sample_count = 0;

    83     _last_sample = 0;

    84     _is_old = false;

    85   }

    86 

    87   // Useful for modifying static structures after startup.

    88   void  modify(size_t avg, unsigned wt, bool force = false)  {

    89     assert(force, "Are you sure you want to call this?");

    90     _average = (float)avg;

    91     _weight  = wt;

    92   }

    93 

    94   // Accessors

    95   float    average() const       { return _average;       }

    96   unsigned weight()  const       { return _weight;        }

    97   unsigned count()   const       { return _sample_count;  }

    98   float    last_sample() const   { return _last_sample;   }

    99   bool     is_old()  const       { return _is_old;        }

   100 

   101   // Update data with a new sample.

   102   void sample(float new_sample);

   103 

   104   static inline float exp_avg(float avg, float sample,

   105                                unsigned int weight) {

   106     assert(weight <= 100, "weight must be a percent");

   107     return (100.0F - weight) * avg / 100.0F + weight * sample / 100.0F;

   108   }

   109   static inline size_t exp_avg(size_t avg, size_t sample,

   110                                unsigned int weight) {

   111     // Convert to float and back to avoid integer overflow.

   112     return (size_t)exp_avg((float)avg, (float)sample, weight);

   113   }

   114 

   115   // Printing

   116   void print_on(outputStream* st) const;

   117   void print() const;

   118 };

   119 

   120 

   121 // A weighted average that includes a deviation from the average,

   122 // some multiple of which is added to the average.

   123 //

   124 // This serves as our best estimate of an upper bound on a future

   125 // unknown.

   126 class AdaptivePaddedAverage : public AdaptiveWeightedAverage {

   127  private:

   128   float          _padded_avg;     // The last computed padded average

   129   float          _deviation;      // Running deviation from the average

   130   unsigned       _padding;        // A multiple which, added to the average,

   131                                   // gives us an upper bound guess.

   132 

   133  protected:

   134   void set_padded_average(float avg)  { _padded_avg = avg;  }

   135   void set_deviation(float dev)       { _deviation  = dev;  }

   136 

   137  public:

   138   AdaptivePaddedAverage() :

   139     AdaptiveWeightedAverage(0),

   140     _padded_avg(0.0), _deviation(0.0), _padding(0) {}

   141 

   142   AdaptivePaddedAverage(unsigned weight, unsigned padding) :

   143     AdaptiveWeightedAverage(weight),

   144     _padded_avg(0.0), _deviation(0.0), _padding(padding) {}

   145 

   146   // Placement support

   147   void* operator new(size_t ignored, void* p) throw() { return p; }

   148   // Allocator

   149   void* operator new(size_t size) throw() { return CHeapObj<mtGC>::operator new(size); }

   150 

   151   // Accessor

   152   float padded_average() const         { return _padded_avg; }

   153   float deviation()      const         { return _deviation;  }

   154   unsigned padding()     const         { return _padding;    }

   155 

   156   void clear() {

   157     AdaptiveWeightedAverage::clear();

   158     _padded_avg = 0;

   159     _deviation = 0;

   160   }

   161 

   162   // Override

   163   void  sample(float new_sample);

   164 

   165   // Printing

   166   void print_on(outputStream* st) const;

   167   void print() const;

   168 };

   169 

   170 // A weighted average that includes a deviation from the average,

   171 // some multiple of which is added to the average.

   172 //

   173 // This serves as our best estimate of an upper bound on a future

   174 // unknown.

   175 // A special sort of padded average:  it doesn't update deviations

   176 // if the sample is zero. The average is allowed to change. We're

   177 // preventing the zero samples from drastically changing our padded

   178 // average.

   179 class AdaptivePaddedNoZeroDevAverage : public AdaptivePaddedAverage {

   180 public:

   181   AdaptivePaddedNoZeroDevAverage(unsigned weight, unsigned padding) :

   182     AdaptivePaddedAverage(weight, padding)  {}

   183   // Override

   184   void  sample(float new_sample);

   185 

   186   // Printing

   187   void print_on(outputStream* st) const;

   188   void print() const;

   189 };

   190 

   191 // Use a least squares fit to a set of data to generate a linear

   192 // equation.

   193 //              y = intercept + slope * x

   194 

   195 class LinearLeastSquareFit : public CHeapObj<mtGC> {

   196   double _sum_x;        // sum of all independent data points x

   197   double _sum_x_squared; // sum of all independent data points x**2

   198   double _sum_y;        // sum of all dependent data points y

   199   double _sum_xy;       // sum of all x * y.

   200   double _intercept;     // constant term

   201   double _slope;        // slope

   202   // The weighted averages are not currently used but perhaps should

   203   // be used to get decaying averages.

   204   AdaptiveWeightedAverage _mean_x; // weighted mean of independent variable

   205   AdaptiveWeightedAverage _mean_y; // weighted mean of dependent variable

   206 

   207  public:

   208   LinearLeastSquareFit(unsigned weight);

   209   void update(double x, double y);

   210   double y(double x);

   211   double slope() { return _slope; }

   212   // Methods to decide if a change in the dependent variable will

   213   // achieve a desired goal.  Note that these methods are not

   214   // complementary and both are needed.

   215   bool decrement_will_decrease();

   216   bool increment_will_decrease();

   217 };

   218 

   219 #endif // SHARE_VM_GC_SHARED_GCUTIL_HPP