backend: add double version of tan

cp from fdlibm and pass the cft after refined

Signed-off-by: rander <rander.wang@intel.com>
Tested-by: Yang Rong <rong.r.yang@intel.com>

2 files changed, 128 insertions, 0 deletions

diff --git a/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl b/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
index 378b3b03..a705e2ff 100644
--- a/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
+++ b/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
@@ -1871,6 +1871,133 @@ OVERLOADABLE double sqrt(double x)
     return z;
 }
 
+double __kernel_tan(double x, double y, int iy)
+{
+		 const double xxx[] = {
+			 3.33333333333334091986e-01,	/* 3FD55555, 55555563 */
+			 1.33333333333201242699e-01,	/* 3FC11111, 1110FE7A */
+			 5.39682539762260521377e-02,	/* 3FABA1BA, 1BB341FE */
+			 2.18694882948595424599e-02,	/* 3F9664F4, 8406D637 */
+			 8.86323982359930005737e-03,	/* 3F8226E3, E96E8493 */
+			 3.59207910759131235356e-03,	/* 3F6D6D22, C9560328 */
+			 1.45620945432529025516e-03,	/* 3F57DBC8, FEE08315 */
+			 5.88041240820264096874e-04,	/* 3F4344D8, F2F26501 */
+			 2.46463134818469906812e-04,	/* 3F3026F7, 1A8D1068 */
+			 7.81794442939557092300e-05,	/* 3F147E88, A03792A6 */
+			 7.14072491382608190305e-05,	/* 3F12B80F, 32F0A7E9 */
+			-1.85586374855275456654e-05,	/* BEF375CB, DB605373 */
+			 2.59073051863633712884e-05,	/* 3EFB2A70, 74BF7AD4 */
+	/* one */	1.00000000000000000000e+00,	/* 3FF00000, 00000000 */
+	/* pio4 */   7.85398163397448278999e-01,	/* 3FE921FB, 54442D18 */
+	/* pio4lo */	 3.06161699786838301793e-17 /* 3C81A626, 33145C07 */
+	};
+#define	one	xxx[13]
+#define	pio4	xxx[14]
+#define	pio4lo	xxx[15]
+#define	T	xxx
+
+	double z, r, v, w, s;
+	int ix, hx;
+
+	hx = __HI(x);		/* high word of x */
+	ix = hx & 0x7fffffff;			/* high word of |x| */
+	if (ix < 0x3e300000) {			/* x < 2**-28 */
+		if ((int) x == 0) {		/* generate inexact */
+			if (((ix | __LO(x)) | (iy + 1)) == 0)
+				return one / fabs(x);
+			else {
+				if (iy == 1)
+					return x;
+				else {	/* compute -1 / (x+y) carefully */
+					double a, t;
+
+					z = w = x + y;
+					__setLow(&z, 0);
+					v = y - (z - x);
+					t = a = -one / w;
+					__setLow(&t, 0);
+					s = one + t * z;
+					return t + a * (s + t * v);
+				}
+			}
+		}
+	}
+	if (ix >= 0x3FE59428) {	/* |x| >= 0.6744 */
+		if (hx < 0) {
+			x = -x;
+			y = -y;
+		}
+		z = pio4 - x;
+		w = pio4lo - y;
+		x = z + w;
+		y = 0.0;
+	}
+	z = x * x;
+	w = z * z;
+	/*
+	 * Break x^5*(T[1]+x^2*T[2]+...) into
+	 * x^5(T[1]+x^4*T[3]+...+x^20*T[11]) +
+	 * x^5(x^2*(T[2]+x^4*T[4]+...+x^22*[T12]))
+	 */
+	r = T[1] + w * (T[3] + w * (T[5] + w * (T[7] + w * (T[9] +
+		w * T[11]))));
+	v = z * (T[2] + w * (T[4] + w * (T[6] + w * (T[8] + w * (T[10] +
+		w * T[12])))));
+	s = z * x;
+	r = y + z * (s * (r + v) + y);
+	r += T[0] * s;
+	w = x + r;
+	if (ix >= 0x3FE59428) {
+		v = (double) iy;
+		return (double) (1 - ((hx >> 30) & 2)) *
+			(v - 2.0 * (x - (w * w / (w + v) - r)));
+	}
+	if (iy == 1)
+		return w;
+	else {
+		/*
+		 * if allow error up to 2 ulp, simply return
+		 * -1.0 / (x+r) here
+		 */
+		/* compute -1.0 / (x+r) accurately */
+		double a, t;
+		z = w;
+		__setLow(&z, 0);
+		v = r - (z - x);	/* z+v = r+x */
+		t = a = -1.0 / w;	/* a = -1.0/w */
+		__setLow(&t, 0);
+		s = 1.0 + t * z;
+		return t + a * (s + t * v);
+	}
+
+#undef	one
+#undef	pio4
+#undef	pio4lo
+#undef	T
+}
+
+OVERLOADABLE double tan(double x)
+{
+	double y[2],z=0.0;
+	int n, ix;
+
+	/* High word of x. */
+	ix = __HI(x);
+
+	/* |x| ~< pi/4 */
+	ix &= 0x7fffffff;
+	if(ix <= 0x3fe921fb) return __kernel_tan(x,z,1);
+
+	/* tan(Inf or NaN) is NaN */
+	else if (ix>=0x7ff00000) return x-x;		/* NaN */
+
+	/* argument reduction needed */
+	else {
+		n = __ieee754_rem_pio2(x,y);
+		return __kernel_tan(y[0],y[1],1-((n&1)<<1)); /*   1 -- n even
+							-1 -- n odd */
+	}
+}
 
 OVERLOADABLE double trunc(double x)
 {
diff --git a/backend/src/libocl/tmpl/ocl_math_common.tmpl.h b/backend/src/libocl/tmpl/ocl_math_common.tmpl.h
index a7959d33..9ac3b31f 100644
--- a/backend/src/libocl/tmpl/ocl_math_common.tmpl.h
+++ b/backend/src/libocl/tmpl/ocl_math_common.tmpl.h
@@ -57,6 +57,7 @@ OVERLOADABLE double round(double x);
 OVERLOADABLE double sin(double x);
 OVERLOADABLE double sinh(double x);
 OVERLOADABLE double sqrt(double x);
+OVERLOADABLE double tan(double x);
 OVERLOADABLE double trunc(double x);