diff options
Diffstat (limited to '2.3-1/src/c/elementaryFunctions/sqrt/csqrts.c')
-rw-r--r-- | 2.3-1/src/c/elementaryFunctions/sqrt/csqrts.c | 111 |
1 files changed, 111 insertions, 0 deletions
diff --git a/2.3-1/src/c/elementaryFunctions/sqrt/csqrts.c b/2.3-1/src/c/elementaryFunctions/sqrt/csqrts.c new file mode 100644 index 00000000..a24f9558 --- /dev/null +++ b/2.3-1/src/c/elementaryFunctions/sqrt/csqrts.c @@ -0,0 +1,111 @@ +/*
+ * Scilab ( http://www.scilab.org/ ) - This file is part of Scilab
+ * Copyright (C) 2008-2008 - INRIA - Bruno JOFRET
+ * Copyright (C) Bruno Pincon
+ *
+ * This file must be used under the terms of the CeCILL.
+ * This source file is licensed as described in the file COPYING, which
+ * you should have received as part of this distribution. The terms
+ * are also available at
+ * http://www.cecill.info/licences/Licence_CeCILL_V2-en.txt
+ *
+ */
+
+#include <math.h>
+#include "sqrt.h"
+#include "lapack.h"
+#include "abs.h"
+#include "sign.h"
+#include "pythag.h"
+
+#ifdef _MSC_VER
+#include <float.h>
+#define isnan(x) _isnan((double)x)
+#endif
+
+#define _sign(a, b) b >=0 ? a : -a
+
+floatComplex csqrts(floatComplex in) {
+ float RMax = (float) getOverflowThreshold();
+ float BRMin = 2.0f * (float) getUnderflowThreshold();
+
+ float RealIn = creals(in);
+ float ImgIn = cimags(in);
+
+ float RealOut = 0;
+ float ImgOut = 0;
+
+ if(RealIn == 0)
+ {/* pure imaginary case */
+ if(dabss(ImgIn >= BRMin))
+ RealOut = ssqrts(0.5f * sabss(ImgIn));
+ else
+ RealOut = ssqrts(sabss(ImgIn)) * ssqrts(0.5);
+
+ ImgOut = _sign(1, ImgIn) * RealOut;
+ }
+ else if( sabss(RealIn) <= RMax && sabss(ImgIn) <= RMax)
+ {/* standard case : a (not zero) and b are finite */
+ float Temp = ssqrts(2.0f * (sabss(RealIn) + spythags(RealIn, ImgIn)));
+ /* overflow test */
+ if(Temp > RMax)
+ {/* handle (spurious) overflow by scaling a and b */
+ float RealTemp = RealIn / 16.0f;
+ float ImgTemp = ImgIn / 16.0f;
+ Temp = ssqrts(2.0f * (sabss(RealIn) + spythags(RealIn, ImgTemp)));
+ if(RealTemp >= 0)
+ {
+ RealOut = 2 * Temp;
+ ImgOut = 4 * ImgTemp / Temp;
+ }
+ else
+ {
+ RealOut = 4 * sabss(ImgIn) / Temp;
+ ImgOut = _sign(2, ImgIn) * Temp;
+ }
+ }
+ else if(RealIn >= 0) /* classic switch to get the stable formulas */
+ {
+ RealOut = 0.5f * Temp;
+ ImgOut = ImgIn / Temp;
+ }
+ else
+ {
+ RealOut = sabss(ImgIn) / Temp;
+ ImgOut = (_sign(0.5f, ImgIn)) * Temp;
+ }
+ }
+ else
+ {
+ /*
+ //Here we treat the special cases where a and b are +- 00 or NaN.
+ //The following is the treatment recommended by the C99 standard
+ //with the simplification of returning NaN + i NaN if the
+ //the real part or the imaginary part is NaN (C99 recommends
+ //something more complicated)
+ */
+
+ if(isnan(RealIn) == 1 || isnan(ImgIn) == 1)
+ {/* got NaN + i NaN */
+ RealOut = RealIn + ImgIn;
+ ImgOut = RealOut;
+ }
+ else if( dabss(ImgIn) > RMax)
+ {/* case a +- i oo -> result must be +oo +- i oo for all a (finite or not) */
+ RealOut = sabss(ImgIn);
+ ImgOut = ImgIn;
+ }
+ else if(RealIn < -RMax)
+ {/* here a is -Inf and b is finite */
+ RealOut = 0;
+ ImgOut = _sign(1, ImgIn) * sabss(RealIn);
+ }
+ else
+ {/* here a is +Inf and b is finite */
+ RealOut = RealIn;
+ ImgOut = 0;
+ }
+ }
+
+ return FloatComplex(RealOut, ImgOut);
+}
|