2007-09-28 Patrick Mansfield <patmans@us.ibm.com>

	* libm/machine/spu/Makefile.am: Add new C files.
	* libm/machine/spu/Makefile.in: Regenerate.
	* libm/machine/spu/headers/acos.h: New file.
	* libm/machine/spu/headers/acosd2.h: Ditto.
	* libm/machine/spu/headers/acosf.h: Ditto.
	* libm/machine/spu/headers/acosf4.h: Ditto.
	* libm/machine/spu/headers/acosh.h: Ditto.
	* libm/machine/spu/headers/acoshd2.h: Ditto.
	* libm/machine/spu/headers/acoshf.h: Ditto.
	* libm/machine/spu/headers/acoshf4.h: Ditto.
	* libm/machine/spu/headers/asin.h: Ditto.
	* libm/machine/spu/headers/asind2.h: Ditto.
	* libm/machine/spu/headers/asinf.h: Ditto.
	* libm/machine/spu/headers/asinf4.h: Ditto.
	* libm/machine/spu/headers/asinh.h: Ditto.
	* libm/machine/spu/headers/asinhd2.h: Ditto.
	* libm/machine/spu/headers/asinhf.h: Ditto.
	* libm/machine/spu/headers/asinhf4.h: Ditto.
	* libm/machine/spu/headers/atan.h: Ditto.
	* libm/machine/spu/headers/atan2.h: Ditto.
	* libm/machine/spu/headers/atan2d2.h: Ditto.
	* libm/machine/spu/headers/atan2f.h: Ditto.
	* libm/machine/spu/headers/atan2f4.h: Ditto.
	* libm/machine/spu/headers/atand2.h: Ditto.
	* libm/machine/spu/headers/atanf4.h: Ditto.
	* libm/machine/spu/headers/atanh.h: Ditto.
	* libm/machine/spu/headers/atanhd2.h: Ditto.
	* libm/machine/spu/headers/atanhf.h: Ditto.
	* libm/machine/spu/headers/atanhf4.h: Ditto.
	* libm/machine/spu/headers/cos.h: Ditto.
	* libm/machine/spu/headers/cos_sin.h: Ditto.
	* libm/machine/spu/headers/cosd2.h: Ditto.
	* libm/machine/spu/headers/cosf.h: Ditto.
	* libm/machine/spu/headers/cosf4.h: Ditto.
	* libm/machine/spu/headers/cosh.h: Ditto.
	* libm/machine/spu/headers/coshd2.h: Ditto.
	* libm/machine/spu/headers/coshf.h: Ditto.
	* libm/machine/spu/headers/coshf4.h: Ditto.
	* libm/machine/spu/headers/divd2.h: Ditto.
	* libm/machine/spu/headers/divf4.h: Ditto.
	* libm/machine/spu/headers/erf.h: Ditto.
	* libm/machine/spu/headers/erf_utils.h: Ditto.
	* libm/machine/spu/headers/erfc.h: Ditto.
	* libm/machine/spu/headers/erfcd2.h: Ditto.
	* libm/machine/spu/headers/erfcf.h: Ditto.
	* libm/machine/spu/headers/erfcf4.h: Ditto.
	* libm/machine/spu/headers/erfd2.h: Ditto.
	* libm/machine/spu/headers/erff.h: Ditto.
	* libm/machine/spu/headers/erff4.h: Ditto.
	* libm/machine/spu/headers/exp.h: Ditto.
	* libm/machine/spu/headers/exp2.h: Ditto.
	* libm/machine/spu/headers/exp2d2.h: Ditto.
	* libm/machine/spu/headers/exp2f4.h: Ditto.
	* libm/machine/spu/headers/expd2.h: Ditto.
	* libm/machine/spu/headers/expf4.h: Ditto.
	* libm/machine/spu/headers/expm1.h: Ditto.
	* libm/machine/spu/headers/expm1d2.h: Ditto.
	* libm/machine/spu/headers/expm1f.h: Ditto.
	* libm/machine/spu/headers/expm1f4.h: Ditto.
	* libm/machine/spu/headers/floord2.h: Ditto.
	* libm/machine/spu/headers/floorf4.h: Ditto.
	* libm/machine/spu/headers/hypot.h: Ditto.
	* libm/machine/spu/headers/hypotd2.h: Ditto.
	* libm/machine/spu/headers/hypotf.h: Ditto.
	* libm/machine/spu/headers/hypotf4.h: Ditto.
	* libm/machine/spu/headers/isnan.h: Ditto.
	* libm/machine/spu/headers/isnand2.h: Ditto.
	* libm/machine/spu/headers/isnanf.h: Ditto.
	* libm/machine/spu/headers/isnanf4.h: Ditto.
	* libm/machine/spu/headers/ldexpd2.h: Ditto.
	* libm/machine/spu/headers/ldexpf4.h: Ditto.
	* libm/machine/spu/headers/lgamma.h: Ditto.
	* libm/machine/spu/headers/lgammad2.h: Ditto.
	* libm/machine/spu/headers/lgammaf.h: Ditto.
	* libm/machine/spu/headers/lgammaf4.h: Ditto.
	* libm/machine/spu/headers/log.h: Ditto.
	* libm/machine/spu/headers/log10.h: Ditto.
	* libm/machine/spu/headers/log10d2.h: Ditto.
	* libm/machine/spu/headers/log1p.h: Ditto.
	* libm/machine/spu/headers/log1pd2.h: Ditto.
	* libm/machine/spu/headers/log1pf.h: Ditto.
	* libm/machine/spu/headers/log1pf4.h: Ditto.
	* libm/machine/spu/headers/log2.h: Ditto.
	* libm/machine/spu/headers/log2d2.h: Ditto.
	* libm/machine/spu/headers/log2f4.h: Ditto.
	* libm/machine/spu/headers/logbf.h: Ditto.
	* libm/machine/spu/headers/logbf4.h: Ditto.
	* libm/machine/spu/headers/logd2.h: Ditto.
	* libm/machine/spu/headers/logf4.h: Ditto.
	* libm/machine/spu/headers/nearbyintf.h: Ditto.
	* libm/machine/spu/headers/nearbyintf4.h: Ditto.
	* libm/machine/spu/headers/nextafter.h: Ditto.
	* libm/machine/spu/headers/nextafterd2.h: Ditto.
	* libm/machine/spu/headers/nextafterf.h: Ditto.
	* libm/machine/spu/headers/nextafterf4.h: Ditto.
	* libm/machine/spu/headers/pow.h: Ditto.
	* libm/machine/spu/headers/powd2.h: Ditto.
	* libm/machine/spu/headers/powf.h: Ditto.
	* libm/machine/spu/headers/powf4.h: Ditto.
	* libm/machine/spu/headers/recipd2.h: Ditto.
	* libm/machine/spu/headers/recipf4.h: Ditto.
	* libm/machine/spu/headers/rintf.h: Ditto.
	* libm/machine/spu/headers/rintf4.h: Ditto.
	* libm/machine/spu/headers/scalbnf4.h: Ditto.
	* libm/machine/spu/headers/signbit.h: Ditto.
	* libm/machine/spu/headers/signbitd2.h: Ditto.
	* libm/machine/spu/headers/simdmath.h: Ditto.
	* libm/machine/spu/headers/sin.h: Ditto.
	* libm/machine/spu/headers/sincos.h: Ditto.
	* libm/machine/spu/headers/sincosd2.h: Ditto.
	* libm/machine/spu/headers/sincosf.h: Ditto.
	* libm/machine/spu/headers/sincosf4.h: Ditto.
	* libm/machine/spu/headers/sind2.h: Ditto.
	* libm/machine/spu/headers/sinf.h: Ditto.
	* libm/machine/spu/headers/sinf4.h: Ditto.
	* libm/machine/spu/headers/sinh.h: Ditto.
	* libm/machine/spu/headers/sinhd2.h: Ditto.
	* libm/machine/spu/headers/sinhf.h: Ditto.
	* libm/machine/spu/headers/sinhf4.h: Ditto.
	* libm/machine/spu/headers/sqrtd2.h: Ditto.
	* libm/machine/spu/headers/sqrtf.h: Ditto.
	* libm/machine/spu/headers/sqrtf4.h: Ditto.
	* libm/machine/spu/headers/tan.h: Ditto.
	* libm/machine/spu/headers/tand2.h: Ditto.
	* libm/machine/spu/headers/tanf.h: Ditto.
	* libm/machine/spu/headers/tanf4.h: Ditto.
	* libm/machine/spu/headers/tanh.h: Ditto.
	* libm/machine/spu/headers/tanhd2.h: Ditto.
	* libm/machine/spu/headers/tanhf.h: Ditto.
	* libm/machine/spu/headers/tanhf4.h: Ditto.
	* libm/machine/spu/headers/tgamma.h: Ditto.
	* libm/machine/spu/headers/tgammad2.h: Ditto.
	* libm/machine/spu/headers/tgammaf.h: Ditto.
	* libm/machine/spu/headers/tgammaf4.h: Ditto.
	* libm/machine/spu/headers/truncd2.h: Ditto.
	* libm/machine/spu/headers/truncf4.h: Ditto.
	* libm/machine/spu/log2.c: Ditto.
	* libm/machine/spu/s_asinh.c: Ditto.
	* libm/machine/spu/s_atan.c: Ditto.
	* libm/machine/spu/s_cos.c: Ditto.
	* libm/machine/spu/s_erf.c: Ditto.
	* libm/machine/spu/s_expm1.c: Ditto.
	* libm/machine/spu/s_isnan.c: Ditto.
	* libm/machine/spu/s_log1p.c: Ditto.
	* libm/machine/spu/s_nextafter.c: Ditto.
	* libm/machine/spu/s_sin.c: Ditto.
	* libm/machine/spu/s_tan.c: Ditto.
	* libm/machine/spu/s_tanh.c: Ditto.
	* libm/machine/spu/sf_asinh.c: Ditto.
	* libm/machine/spu/sf_cos.c: Ditto.
	* libm/machine/spu/sf_erf.c: Ditto.
	* libm/machine/spu/sf_expm1.c: Ditto.
	* libm/machine/spu/sf_log1p.c: Ditto.
	* libm/machine/spu/sf_logb.c: Ditto.
	* libm/machine/spu/sf_nearbyint.c: Ditto.
	* libm/machine/spu/sf_nextafter.c: Ditto.
	* libm/machine/spu/sf_rint.c: Ditto.
	* libm/machine/spu/sf_sin.c: Ditto.
	* libm/machine/spu/sf_tan.c: Ditto.
	* libm/machine/spu/sf_tanh.c: Ditto.
	* libm/machine/spu/w_acos.c: Ditto.
	* libm/machine/spu/w_acosh.c: Ditto.
	* libm/machine/spu/w_asin.c: Ditto.
	* libm/machine/spu/w_atan2.c: Ditto.
	* libm/machine/spu/w_atanh.c: Ditto.
	* libm/machine/spu/w_cosh.c: Ditto.
	* libm/machine/spu/w_exp.c: Ditto.
	* libm/machine/spu/w_exp2.c: Ditto.
	* libm/machine/spu/w_hypot.c: Ditto.
	* libm/machine/spu/w_lgamma.c: Ditto.
	* libm/machine/spu/w_log.c: Ditto.
	* libm/machine/spu/w_log10.c: Ditto.
	* libm/machine/spu/w_pow.c: Ditto.
	* libm/machine/spu/w_sincos.c: Ditto.
	* libm/machine/spu/w_sinh.c: Ditto.
	* libm/machine/spu/w_tgamma.c: Ditto.
	* libm/machine/spu/wf_acos.c: Ditto.
	* libm/machine/spu/wf_acosh.c: Ditto.
	* libm/machine/spu/wf_asin.c: Ditto.
	* libm/machine/spu/wf_atan2.c: Ditto.
	* libm/machine/spu/wf_atanh.c: Ditto.
	* libm/machine/spu/wf_cosh.c: Ditto.
	* libm/machine/spu/wf_hypot.c: Ditto.
	* libm/machine/spu/wf_lgamma.c: Ditto.
	* libm/machine/spu/wf_pow.c: Ditto.
	* libm/machine/spu/wf_sincos.c: Ditto.
	* libm/machine/spu/wf_sinh.c: Ditto.
	* libm/machine/spu/wf_sqrt.c: Ditto.
	* libm/machine/spu/wf_tgamma.c: Ditto.

1 files changed, 244 insertions, 0 deletions

diff --git a/newlib/libm/machine/spu/headers/divd2.h b/newlib/libm/machine/spu/headers/divd2.h
new file mode 100644
index 000000000..b04d3c592
--- /dev/null
+++ b/newlib/libm/machine/spu/headers/divd2.h
@@ -0,0 +1,244 @@
+/* --------------------------------------------------------------  */
+/* (C)Copyright 2006,2007,                                         */
+/* International Business Machines Corporation,                    */
+/* Sony Computer Entertainment, Incorporated,                      */
+/* Toshiba Corporation,                                            */
+/*                                                                 */
+/* All Rights Reserved.                                            */
+/*                                                                 */
+/* Redistribution and use in source and binary forms, with or      */
+/* without modification, are permitted provided that the           */
+/* following conditions are met:                                   */
+/*                                                                 */
+/* - Redistributions of source code must retain the above copyright*/
+/*   notice, this list of conditions and the following disclaimer. */
+/*                                                                 */
+/* - Redistributions in binary form must reproduce the above       */
+/*   copyright notice, this list of conditions and the following   */
+/*   disclaimer in the documentation and/or other materials        */
+/*   provided with the distribution.                               */
+/*                                                                 */
+/* - Neither the name of IBM Corporation nor the names of its      */
+/*   contributors may be used to endorse or promote products       */
+/*   derived from this software without specific prior written     */
+/*   permission.                                                   */
+/* Redistributions of source code must retain the above copyright  */
+/* notice, this list of conditions and the following disclaimer.   */
+/*                                                                 */
+/* Redistributions in binary form must reproduce the above         */
+/* copyright notice, this list of conditions and the following     */
+/* disclaimer in the documentation and/or other materials          */
+/* provided with the distribution.                                 */
+/*                                                                 */
+/* Neither the name of IBM Corporation nor the names of its        */
+/* contributors may be used to endorse or promote products         */
+/* derived from this software without specific prior written       */
+/* permission.                                                     */
+/*                                                                 */
+/* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND          */
+/* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,     */
+/* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF        */
+/* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE        */
+/* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR            */
+/* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,    */
+/* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT    */
+/* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;    */
+/* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)        */
+/* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN       */
+/* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR    */
+/* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,  */
+/* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.              */
+/* --------------------------------------------------------------  */
+/* PROLOG END TAG zYx                                              */
+#ifdef __SPU__
+
+#ifndef _DIVD2_H_
+#define _DIVD2_H_		 1
+
+#include <spu_intrinsics.h>
+
+/*
+ * FUNCTION
+ * 	vector double _divd2(vector double a, vector double b)
+ * 
+ * DESCRIPTION
+ * 	_divd2 divides the vector dividend a by the vector divisor b and 
+ *      returns the resulting vector quotient.  Maximum error 0.5 ULPS for 
+ *      normalized results, 1ulp for denorm results, over entire double 
+ *      range including denorms, compared to true result in round-to-nearest
+ *      rounding mode.  Handles Inf or NaN operands and results correctly.
+ */
+static __inline vector double _divd2(vector double a, vector double b)
+{
+
+
+  /* Variables
+   */
+  vec_float4 inv_bf, mant_bf;
+  vec_double2 mant_a, mant_b, inv_b, q0, q1, q2, mult;
+  vec_int4 exp, tmp;
+  vec_uint4 exp_a, exp_b, exp_q1, overflow, nounderflow, normal, utmp,
+    sign_a, sign_b, a_frac, b_frac, a_frac_0, b_frac_0, a_exp_0, b_exp_0, 
+    a_exp_ones, b_exp_ones, a_nan, b_nan, a_inf, b_inf, a_zero, b_zero, 
+    res_nan, sign_res;
+
+  /* Constants
+   */
+  vec_float4 onef = spu_splats(1.0f);
+  vec_double2 one = spu_splats(1.0);
+  vec_uint4 exp_mask = (vec_uint4) { 0x7FF00000, 0, 0x7FF00000, 0 };
+  vec_uint4 sign_mask = (vec_uint4) { 0x80000000, 0, 0x80000000, 0};
+  vec_uint4 sign_exp_mask = (vec_uint4) { 0xFFF00000, 0, 0xFFF00000,0};
+  vec_uint4 frac_mask =(vec_uint4) { 0x000FFFFF, 0xFFFFFFFF, 0x000FFFFF, 0xFFFFFFFF };
+  vec_uchar16 swap32 = (vec_uchar16) ((vec_uint4) { 0x04050607, 0x00010203, 0x0C0D0E0F, 0x08090A0B} );
+  vec_uint4 zero = (vec_uint4) { 0, 0, 0, 0 };
+  vec_int4 e1022 = (vec_int4) { 0x000003FE, 0, 0x000003FE, 0 };
+  vec_int4 emax  = (vec_int4) { 0x000007FE, 0, 0x000007FE, 0 };
+  vec_int4 e1    = (vec_int4) { 0x00000001, 0, 0x00000001, 0 };
+
+  vec_uint4 nan  = (vec_uint4) { 0x7FF80000, 0, 0x7FF80000, 0};
+
+  /* Extract exponents and underflow denorm arguments to signed zero.
+   */
+  exp_a = spu_and((vec_uint4)a, exp_mask);
+  exp_b = spu_and((vec_uint4)b, exp_mask);
+
+  sign_a = spu_and((vec_uint4)a, sign_mask);
+  sign_b = spu_and((vec_uint4)b, sign_mask);
+
+  a_exp_0 = spu_cmpeq (exp_a, 0);
+  utmp = spu_shuffle (a_exp_0, a_exp_0, swap32);
+  a_exp_0 = spu_and (a_exp_0, utmp);
+  b_exp_0 = spu_cmpeq (exp_b, 0);
+  utmp = spu_shuffle (b_exp_0, b_exp_0, swap32);
+  b_exp_0 = spu_and (b_exp_0, utmp);
+
+  a = spu_sel(a, (vec_double2)sign_a, (vec_ullong2)a_exp_0);
+  b = spu_sel(b, (vec_double2)sign_b, (vec_ullong2)b_exp_0);
+
+  /* Force the divisor and dividend into the range [1.0,2.0).
+     (Unless they're zero.)
+  */
+  mant_a = spu_sel(a, one, (vec_ullong2)sign_exp_mask);
+  mant_b = spu_sel(b, one, (vec_ullong2)sign_exp_mask);
+
+  /* Approximate the single reciprocal of b by using
+   * the single precision reciprocal estimate followed by one 
+   * single precision iteration of Newton-Raphson.
+   */
+  mant_bf = spu_roundtf(mant_b);
+  inv_bf = spu_re(mant_bf);
+  inv_bf = spu_madd(spu_nmsub(mant_bf, inv_bf, onef), inv_bf, inv_bf);
+
+  /* Perform 2 more Newton-Raphson iterations in double precision.
+   */
+  inv_b = spu_extend(inv_bf);
+  inv_b = spu_madd(spu_nmsub(mant_b, inv_b, one), inv_b, inv_b);
+  q0 = spu_mul(mant_a, inv_b);
+  q1 = spu_madd(spu_nmsub(mant_b, q0, mant_a), inv_b, q0);
+
+  /* Compute the quotient's expected exponent. If the exponent
+   * is out of range, then force the resulting exponent to 0.
+   * (1023 with the bias). We correct for the out of range 
+   * values by computing a multiplier (mult) that will force the 
+   * result to the correct out of range value and set the 
+   * correct exception flag (UNF, OVF, or neither).
+   */
+  exp_q1 = spu_and((vec_uint4)q1, exp_mask);
+  exp = spu_sub((vec_int4)exp_a, (vec_int4)exp_b);
+  exp = spu_rlmaska(exp, -20);  // shift right to allow enough bits for working
+  tmp = spu_rlmaska((vec_int4)exp_q1, -20);
+  exp = spu_add(exp, tmp);  // biased exponent of result (right justified)
+
+  /* The default multiplier is 1.0. If an underflow is detected (the computed 
+   * exponent is less than or equal to a biased 0), force the multiplier to 0.0.
+   * If exp<=0 set mult = 2**(unbiased exp + 1022) and unbiased exp = -1022
+   * = biased 1, the smallest normalized exponent.  If exp<-51 set 
+   * mult = 2**(-1074) to ensure underflowing result.  Otherwise mult=1.
+   */
+  normal = spu_cmpgt(exp, 0);
+  nounderflow = spu_cmpgt(exp, -52);
+  tmp = spu_add(exp, e1022);
+  mult = (vec_double2)spu_sl(tmp, 20);
+  mult = spu_sel(mult, one, (vec_ullong2)normal);
+  mult = spu_sel((vec_double2)e1, mult, (vec_ullong2)nounderflow);
+  exp = spu_sel(e1, exp, normal);  // unbiased -1022 is biased 1
+
+  /* Force the multiplier to positive infinity (exp_mask) and the biased 
+   * exponent to 1022, if the computed biased exponent is > emax.
+   */
+  overflow = spu_cmpgt(exp, (vec_int4)emax);
+  exp = spu_sel(exp, (vec_int4)e1022, overflow);
+  mult = spu_sel(mult, (vec_double2)exp_mask, (vec_ullong2)overflow);
+
+  /* Determine if a, b are Inf, NaN, or zero.
+   * Since these are rare, it would improve speed if these could be detected
+   * quickly and a branch used to avoid slowing down the main path.  However
+   * most of the work seems to be in the detection.
+   */
+  a_exp_ones = spu_cmpeq (exp_a, exp_mask);
+  utmp = spu_shuffle (a_exp_ones, a_exp_ones, swap32);
+  a_exp_ones = spu_and (a_exp_ones, utmp);
+
+  a_frac = spu_and ((vec_uint4)a, frac_mask);
+  a_frac_0 = spu_cmpeq (a_frac, 0);
+  utmp = spu_shuffle (a_frac_0, a_frac_0, swap32);
+  a_frac_0 = spu_and (a_frac_0, utmp);
+
+  a_zero = spu_and (a_exp_0, a_frac_0);
+  a_inf = spu_and (a_exp_ones, a_frac_0);
+  a_nan = spu_andc (a_exp_ones, a_frac_0);
+
+  b_exp_ones = spu_cmpeq (exp_b, exp_mask);
+  utmp = spu_shuffle (b_exp_ones, b_exp_ones, swap32);
+  b_exp_ones = spu_and (b_exp_ones, utmp);
+
+  b_frac = spu_and ((vec_uint4)b, frac_mask);
+  b_frac_0 = spu_cmpeq (b_frac, 0);
+  utmp = spu_shuffle (b_frac_0, b_frac_0, swap32);
+  b_frac_0 = spu_and (b_frac_0, utmp);
+
+  b_zero = spu_and (b_exp_0, b_frac_0);
+  b_inf = spu_and (b_exp_ones, b_frac_0);
+  b_nan = spu_andc (b_exp_ones, b_frac_0);
+
+  /* Handle exception cases */
+
+  /* Result is 0 for 0/x, x!=0, or x/Inf, x!=Inf.
+   * Set mult=0 for 0/0 or Inf/Inf now, since it will be replaced 
+   * with NaN later.
+   */
+  utmp = spu_or (a_zero, b_inf);
+  mult = spu_sel(mult, (vec_double2)zero, (vec_ullong2)utmp);
+
+  /* Result is Inf for x/0, x!=0.  Set mult=Inf for 0/0 now, since it
+   * will be replaced with NaN later.
+   */
+  mult = spu_sel(mult, (vec_double2)exp_mask, (vec_ullong2)b_zero);
+
+  /* Result is NaN if either operand is, or Inf/Inf, or 0/0.
+   */
+  res_nan = spu_or (a_nan, b_nan);
+  utmp = spu_and (a_inf, b_inf);
+  res_nan = spu_or (res_nan, utmp);
+  utmp = spu_and (a_zero, b_zero);
+  res_nan = spu_or (res_nan, utmp);
+  mult = spu_sel(mult, (vec_double2)nan, (vec_ullong2)res_nan);
+  
+  /* Insert sign of result into mult.
+   */
+  sign_res = spu_xor (sign_a, sign_b);
+  mult = spu_or (mult, (vec_double2)sign_res);
+
+  /* Insert the sign and exponent into the result and perform the
+   * final multiplication.
+   */
+  exp = spu_sl(exp, 20);
+  q2 = spu_sel(q1, (vec_double2)exp, (vec_ullong2)exp_mask);
+  q2 = spu_mul(q2, mult);
+
+  return (q2);
+}
+
+#endif /* _DIVD2_H_ */
+#endif /* __SPU__ */