Chris@10: /*
Chris@10:  * Copyright (c) 2003, 2007-11 Matteo Frigo
Chris@10:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
Chris@10:  *
Chris@10:  * This program is free software; you can redistribute it and/or modify
Chris@10:  * it under the terms of the GNU General Public License as published by
Chris@10:  * the Free Software Foundation; either version 2 of the License, or
Chris@10:  * (at your option) any later version.
Chris@10:  *
Chris@10:  * This program is distributed in the hope that it will be useful,
Chris@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@10:  * GNU General Public License for more details.
Chris@10:  *
Chris@10:  * You should have received a copy of the GNU General Public License
Chris@10:  * along with this program; if not, write to the Free Software
Chris@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@10:  *
Chris@10:  */
Chris@10: 
Chris@10: /* This file was automatically generated --- DO NOT EDIT */
Chris@10: /* Generated on Sun Nov 25 07:36:58 EST 2012 */
Chris@10: 
Chris@10: #include "codelet-dft.h"
Chris@10: 
Chris@10: #ifdef HAVE_FMA
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include n1b.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 30 FP additions, 24 FP multiplications,
Chris@10:  * (or, 9 additions, 3 multiplications, 21 fused multiply/add),
Chris@10:  * 37 stack variables, 6 constants, and 14 memory accesses
Chris@10:  */
Chris@10: #include "n1b.h"
Chris@10: 
Chris@10: static void n1bv_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@10:      DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
Chris@10:      DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
Chris@10:      DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@10:      DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
Chris@10:      DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  const R *xi;
Chris@10: 	  R *xo;
Chris@10: 	  xi = ii;
Chris@10: 	  xo = io;
Chris@10: 	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(14, is), MAKE_VOLATILE_STRIDE(14, os)) {
Chris@10: 	       V T1, T2, T3, T8, T9, T5, T6;
Chris@10: 	       T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@10: 	       T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@10: 	       T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       T9 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@10: 	       T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@10: 	       T6 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       {
Chris@10: 		    V Tg, T4, Te, Ta, Tf, T7;
Chris@10: 		    Tg = VSUB(T2, T3);
Chris@10: 		    T4 = VADD(T2, T3);
Chris@10: 		    Te = VSUB(T8, T9);
Chris@10: 		    Ta = VADD(T8, T9);
Chris@10: 		    Tf = VSUB(T5, T6);
Chris@10: 		    T7 = VADD(T5, T6);
Chris@10: 		    {
Chris@10: 			 V Tr, Tj, Tm, Th, To, Tb;
Chris@10: 			 Tr = VFMA(LDK(KP554958132), Te, Tg);
Chris@10: 			 Tj = VFNMS(LDK(KP356895867), T4, Ta);
Chris@10: 			 Tm = VFMA(LDK(KP554958132), Tf, Te);
Chris@10: 			 Th = VFNMS(LDK(KP554958132), Tg, Tf);
Chris@10: 			 ST(&(xo[0]), VADD(T1, VADD(T4, VADD(T7, Ta))), ovs, &(xo[0]));
Chris@10: 			 To = VFNMS(LDK(KP356895867), T7, T4);
Chris@10: 			 Tb = VFNMS(LDK(KP356895867), Ta, T7);
Chris@10: 			 {
Chris@10: 			      V Ts, Tk, Tn, Ti;
Chris@10: 			      Ts = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Tr, Tf));
Chris@10: 			      Tk = VFNMS(LDK(KP692021471), Tj, T7);
Chris@10: 			      Tn = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tm, Tg));
Chris@10: 			      Ti = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Th, Te));
Chris@10: 			      {
Chris@10: 				   V Tp, Tc, Tl, Tq, Td;
Chris@10: 				   Tp = VFNMS(LDK(KP692021471), To, Ta);
Chris@10: 				   Tc = VFNMS(LDK(KP692021471), Tb, T4);
Chris@10: 				   Tl = VFNMS(LDK(KP900968867), Tk, T1);
Chris@10: 				   Tq = VFNMS(LDK(KP900968867), Tp, T1);
Chris@10: 				   Td = VFNMS(LDK(KP900968867), Tc, T1);
Chris@10: 				   ST(&(xo[WS(os, 5)]), VFNMSI(Tn, Tl), ovs, &(xo[WS(os, 1)]));
Chris@10: 				   ST(&(xo[WS(os, 2)]), VFMAI(Tn, Tl), ovs, &(xo[0]));
Chris@10: 				   ST(&(xo[WS(os, 6)]), VFNMSI(Ts, Tq), ovs, &(xo[0]));
Chris@10: 				   ST(&(xo[WS(os, 1)]), VFMAI(Ts, Tq), ovs, &(xo[WS(os, 1)]));
Chris@10: 				   ST(&(xo[WS(os, 4)]), VFNMSI(Ti, Td), ovs, &(xo[0]));
Chris@10: 				   ST(&(xo[WS(os, 3)]), VFMAI(Ti, Td), ovs, &(xo[WS(os, 1)]));
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {9, 3, 21, 0}, &GENUS, 0, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_n1bv_7) (planner *p) {
Chris@10:      X(kdft_register) (p, n1bv_7, &desc);
Chris@10: }
Chris@10: 
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include n1b.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 30 FP additions, 18 FP multiplications,
Chris@10:  * (or, 18 additions, 6 multiplications, 12 fused multiply/add),
Chris@10:  * 24 stack variables, 6 constants, and 14 memory accesses
Chris@10:  */
Chris@10: #include "n1b.h"
Chris@10: 
Chris@10: static void n1bv_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
Chris@10:      DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@10:      DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
Chris@10:      DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
Chris@10:      DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
Chris@10:      DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  const R *xi;
Chris@10: 	  R *xo;
Chris@10: 	  xi = ii;
Chris@10: 	  xo = io;
Chris@10: 	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(14, is), MAKE_VOLATILE_STRIDE(14, os)) {
Chris@10: 	       V Tb, T9, Tc, T3, Te, T6, Td, T7, T8, Ti, Tj;
Chris@10: 	       Tb = LD(&(xi[0]), ivs, &(xi[0]));
Chris@10: 	       T7 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@10: 	       T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       T9 = VSUB(T7, T8);
Chris@10: 	       Tc = VADD(T7, T8);
Chris@10: 	       {
Chris@10: 		    V T1, T2, T4, T5;
Chris@10: 		    T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 		    T2 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@10: 		    T3 = VSUB(T1, T2);
Chris@10: 		    Te = VADD(T1, T2);
Chris@10: 		    T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 		    T5 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@10: 		    T6 = VSUB(T4, T5);
Chris@10: 		    Td = VADD(T4, T5);
Chris@10: 	       }
Chris@10: 	       ST(&(xo[0]), VADD(Tb, VADD(Te, VADD(Tc, Td))), ovs, &(xo[0]));
Chris@10: 	       Ti = VBYI(VFNMS(LDK(KP781831482), T6, VFNMS(LDK(KP433883739), T9, VMUL(LDK(KP974927912), T3))));
Chris@10: 	       Tj = VFMA(LDK(KP623489801), Td, VFNMS(LDK(KP900968867), Tc, VFNMS(LDK(KP222520933), Te, Tb)));
Chris@10: 	       ST(&(xo[WS(os, 2)]), VADD(Ti, Tj), ovs, &(xo[0]));
Chris@10: 	       ST(&(xo[WS(os, 5)]), VSUB(Tj, Ti), ovs, &(xo[WS(os, 1)]));
Chris@10: 	       {
Chris@10: 		    V Ta, Tf, Tg, Th;
Chris@10: 		    Ta = VBYI(VFMA(LDK(KP433883739), T3, VFNMS(LDK(KP781831482), T9, VMUL(LDK(KP974927912), T6))));
Chris@10: 		    Tf = VFMA(LDK(KP623489801), Tc, VFNMS(LDK(KP222520933), Td, VFNMS(LDK(KP900968867), Te, Tb)));
Chris@10: 		    ST(&(xo[WS(os, 3)]), VADD(Ta, Tf), ovs, &(xo[WS(os, 1)]));
Chris@10: 		    ST(&(xo[WS(os, 4)]), VSUB(Tf, Ta), ovs, &(xo[0]));
Chris@10: 		    Tg = VBYI(VFMA(LDK(KP781831482), T3, VFMA(LDK(KP974927912), T9, VMUL(LDK(KP433883739), T6))));
Chris@10: 		    Th = VFMA(LDK(KP623489801), Te, VFNMS(LDK(KP900968867), Td, VFNMS(LDK(KP222520933), Tc, Tb)));
Chris@10: 		    ST(&(xo[WS(os, 1)]), VADD(Tg, Th), ovs, &(xo[WS(os, 1)]));
Chris@10: 		    ST(&(xo[WS(os, 6)]), VSUB(Th, Tg), ovs, &(xo[0]));
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {18, 6, 12, 0}, &GENUS, 0, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_n1bv_7) (planner *p) {
Chris@10:      X(kdft_register) (p, n1bv_7, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */