You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by Luc Maisonobe <Lu...@free.fr> on 2012/03/21 08:53:40 UTC

Re: svn commit: r1303290 - /commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java

Le 21/03/2012 06:26, celestin@apache.org a écrit :
> Author: celestin
> Date: Wed Mar 21 05:26:45 2012
> New Revision: 1303290
> 
> URL: http://svn.apache.org/viewvc?rev=1303290&view=rev
> Log:
> In org.apache.commons.math3.linear.SymmLQ.State, created
>   - boolean bEqualsNullVector()
>   - boolean betaIsZero()
> to improve data encapsulation (see MATH-761).
> 
> Modified:
>     commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java
> 
> Modified: commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java
> URL: http://svn.apache.org/viewvc/commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java?rev=1303290&r1=1303289&r2=1303290&view=diff
> ==============================================================================
> --- commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java (original)
> +++ commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java Wed Mar 21 05:26:45 2012
> @@ -226,6 +226,11 @@ public class SymmLQ
>       * </p>
>       */
>      private static class State {
> +        /** The cubic root of {@link #MACH_PREC}. */
> +        static final double CBRT_MACH_PREC;
> +
> +        /** The machine precision. */
> +        static final double MACH_PREC;
>  
>          /** Reference to the linear operator. */
>          private final RealLinearOperator a;
> @@ -334,6 +339,14 @@ public class SymmLQ
>          /** The value of zeta[1]^2 + ... + zeta[k-1]^2. */
>          private double ynorm2;
>  
> +        /** The value of {@code b == 0} (exact floating-point equality). */
> +        private boolean bIsNull;
> +
> +        static {
> +            MACH_PREC = Math.ulp(1.);
> +            CBRT_MACH_PREC = Math.cbrt(MACH_PREC);

Why not using FastMath ?

Luc

> +        }
> +
>          /**
>           * Creates and inits to k = 1 a new instance of this class.
>           *
> @@ -390,7 +403,7 @@ public class SymmLQ
>              throws NonSelfAdjointOperatorException {
>              final double s = y.dotProduct(y);
>              final double t = x.dotProduct(z);
> -            final double epsa = (s + SymmLQ.MACH_PREC) * SymmLQ.CBRT_MACH_PREC;
> +            final double epsa = (s + MACH_PREC) * CBRT_MACH_PREC;
>              if (FastMath.abs(s - t) > epsa) {
>                  final NonSelfAdjointOperatorException e;
>                  e = new NonSelfAdjointOperatorException();
> @@ -525,8 +538,10 @@ public class SymmLQ
>              }
>              if (this.beta1 == 0.) {
>                  /* If b = 0 exactly, stop with x = 0. */
> +                this.bIsNull = true;
>                  return;
>              }
> +            this.bIsNull = false;
>              this.beta1 = FastMath.sqrt(this.beta1);
>              /* At this point
>               *   r1 = b,
> @@ -766,6 +781,25 @@ public class SymmLQ
>          public boolean hasConverged() {
>              return hasConverged;
>          }
> +
> +        /**
> +         * Returns {@code true} if the right-hand side vector is zero exactly.
> +         *
> +         * @return the boolean value of {@code b == 0}
> +         */
> +        public boolean bEqualsNullVector() {
> +            return bIsNull;
> +        }
> +
> +        /**
> +         * Returns {@code true} if {@code beta} is essentially zero. This method
> +         * is used to check for early stop of the iterations.
> +         *
> +         * @return {@code true} if {@code beta < }{@link #MACH_PREC}
> +         */
> +        public boolean betaEqualsZero() {
> +            return beta < MACH_PREC;
> +        }
>      }
>  
>      /**
> @@ -820,12 +854,6 @@ public class SymmLQ
>          }
>      }
>  
> -    /** The cubic root of {@link #MACH_PREC}. */
> -    static final double CBRT_MACH_PREC;
> -
> -    /** The machine precision. */
> -    static final double MACH_PREC;
> -
>      /** Key for the exception context. */
>      private static final String OPERATOR = "operator";
>  
> @@ -885,11 +913,6 @@ public class SymmLQ
>          this.check = check;
>      }
>  
> -    static {
> -        MACH_PREC = Math.ulp(1.);
> -        CBRT_MACH_PREC = Math.cbrt(MACH_PREC);
> -    }
> -
>      /**
>       * Returns {@code true} if symmetry of the matrix, and symmetry as well as
>       * positive definiteness of the preconditioner should be checked.
> @@ -1159,14 +1182,14 @@ public class SymmLQ
>  
>          final State state = new State(a, minv, b, x, goodb, shift, delta, check);
>          final IterativeLinearSolverEvent event = new SymmLQEvent(this, state);
> -        if (state.beta1 == 0.) {
> +        if (state.bEqualsNullVector()) {
>              /* If b = 0 exactly, stop with x = 0. */
>              manager.fireTerminationEvent(event);
>              return x;
>          }
>          /* Cause termination if beta is essentially zero. */
>          final boolean earlyStop;
> -        earlyStop = (state.beta < MACH_PREC) || (state.hasConverged());
> +        earlyStop = state.betaEqualsZero() || state.hasConverged();
>          manager.fireInitializationEvent(event);
>          if (!earlyStop) {
>              do {
> 
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Re: svn commit: r1303290 - /commons/proper/math/trunk/src/main/java/org/apache/commons/math3/linear/SymmLQ.java

Posted by Sébastien Brisard <se...@m4x.org>.
Hi Luc,
>>
>> +        /** The value of {@code b == 0} (exact floating-point equality). */
>> +        private boolean bIsNull;
>> +
>> +        static {
>> +            MACH_PREC = Math.ulp(1.);
>> +            CBRT_MACH_PREC = Math.cbrt(MACH_PREC);
>
> Why not using FastMath ?
>
> Luc
>
Because I'm silly ;-)
Thanks for pointing that out, will correct that. Meanwhile, should you
find some time to review the refactoring itself (still ongoing), any
comment would be most welcome.

Sébastien


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org