cannam@95: /*
cannam@95:  * Copyright (c) 2003, 2007-11 Matteo Frigo
cannam@95:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
cannam@95:  *
cannam@95:  * This program is free software; you can redistribute it and/or modify
cannam@95:  * it under the terms of the GNU General Public License as published by
cannam@95:  * the Free Software Foundation; either version 2 of the License, or
cannam@95:  * (at your option) any later version.
cannam@95:  *
cannam@95:  * This program is distributed in the hope that it will be useful,
cannam@95:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@95:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
cannam@95:  * GNU General Public License for more details.
cannam@95:  *
cannam@95:  * You should have received a copy of the GNU General Public License
cannam@95:  * along with this program; if not, write to the Free Software
cannam@95:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
cannam@95:  *
cannam@95:  */
cannam@95: 
cannam@95: /* This file was automatically generated --- DO NOT EDIT */
cannam@95: /* Generated on Sun Nov 25 07:42:30 EST 2012 */
cannam@95: 
cannam@95: #include "codelet-rdft.h"
cannam@95: 
cannam@95: #ifdef HAVE_FMA
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 16 -dif -sign 1 -name hc2cbdftv_16 -include hc2cbv.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 103 FP additions, 80 FP multiplications,
cannam@95:  * (or, 53 additions, 30 multiplications, 50 fused multiply/add),
cannam@95:  * 123 stack variables, 3 constants, and 32 memory accesses
cannam@95:  */
cannam@95: #include "hc2cbv.h"
cannam@95: 
cannam@95: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
cannam@95:      DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
cannam@95:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 30)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(64, rs)) {
cannam@95: 	       V T1D, T1F, TV, TW, T17, T18, T1B, T1A, T1H, T1G;
cannam@95: 	       {
cannam@95: 		    V T8, Tv, Tb, TF, Tl, TJ, TP, T1w, TE, T1t, T10, T1p, TG, Te, Tg;
cannam@95: 		    V Th, T2, T3, Ts, Tt, T5, T6, Tp, Tq, T9, TA, T4, TC, Tu, TN;
cannam@95: 		    V T7, TB, Tr, Ta, Tj, Tk, Tc, Td, TY, TD, TO, TZ, T1Q, T19, T1I;
cannam@95: 		    V T1d, Tf, T11, TH, TQ, Ti, TI, T1k, T1K, T1S, T1r, T14, T16, TU, Ty;
cannam@95: 		    V T1z, TX, T1o, T1, TK, TR, Tm, T12, T1C, Tz, T15;
cannam@95: 		    T2 = LD(&(Rp[0]), ms, &(Rp[0]));
cannam@95: 		    T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    Ts = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0]));
cannam@95: 		    Tt = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T5 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
cannam@95: 		    T6 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    Tp = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
cannam@95: 		    Tq = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T9 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    TA = VFNMSCONJ(T3, T2);
cannam@95: 		    T4 = VFMACONJ(T3, T2);
cannam@95: 		    TC = VFMSCONJ(Tt, Ts);
cannam@95: 		    Tu = VFMACONJ(Tt, Ts);
cannam@95: 		    TN = VFNMSCONJ(T6, T5);
cannam@95: 		    T7 = VFMACONJ(T6, T5);
cannam@95: 		    TB = VFNMSCONJ(Tq, Tp);
cannam@95: 		    Tr = VFMACONJ(Tq, Tp);
cannam@95: 		    Ta = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0]));
cannam@95: 		    Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Tk = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
cannam@95: 		    Tc = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Td = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
cannam@95: 		    T8 = VSUB(T4, T7);
cannam@95: 		    TY = VADD(T4, T7);
cannam@95: 		    TD = VADD(TB, TC);
cannam@95: 		    TO = VSUB(TB, TC);
cannam@95: 		    Tv = VSUB(Tr, Tu);
cannam@95: 		    TZ = VADD(Tr, Tu);
cannam@95: 		    Tb = VFMACONJ(Ta, T9);
cannam@95: 		    TF = VFNMSCONJ(Ta, T9);
cannam@95: 		    Tl = VFMACONJ(Tk, Tj);
cannam@95: 		    TJ = VFNMSCONJ(Tk, Tj);
cannam@95: 		    TP = VFMA(LDK(KP707106781), TO, TN);
cannam@95: 		    T1w = VFNMS(LDK(KP707106781), TO, TN);
cannam@95: 		    TE = VFMA(LDK(KP707106781), TD, TA);
cannam@95: 		    T1t = VFNMS(LDK(KP707106781), TD, TA);
cannam@95: 		    T10 = VADD(TY, TZ);
cannam@95: 		    T1p = VSUB(TY, TZ);
cannam@95: 		    TG = VFNMSCONJ(Td, Tc);
cannam@95: 		    Te = VFMACONJ(Td, Tc);
cannam@95: 		    Tg = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Th = LD(&(Rm[0]), -ms, &(Rm[0]));
cannam@95: 		    T1Q = LDW(&(W[TWVL * 22]));
cannam@95: 		    T19 = LDW(&(W[TWVL * 26]));
cannam@95: 		    T1I = LDW(&(W[TWVL * 2]));
cannam@95: 		    T1d = LDW(&(W[TWVL * 28]));
cannam@95: 		    Tf = VSUB(Tb, Te);
cannam@95: 		    T11 = VADD(Tb, Te);
cannam@95: 		    TH = VFNMS(LDK(KP414213562), TG, TF);
cannam@95: 		    TQ = VFMA(LDK(KP414213562), TF, TG);
cannam@95: 		    Ti = VFMACONJ(Th, Tg);
cannam@95: 		    TI = VFMSCONJ(Th, Tg);
cannam@95: 		    T1k = LDW(&(W[0]));
cannam@95: 		    T1K = LDW(&(W[TWVL * 4]));
cannam@95: 		    T1S = LDW(&(W[TWVL * 24]));
cannam@95: 		    TX = LDW(&(W[TWVL * 14]));
cannam@95: 		    T1o = LDW(&(W[TWVL * 6]));
cannam@95: 		    T1 = LDW(&(W[TWVL * 10]));
cannam@95: 		    TK = VFMA(LDK(KP414213562), TJ, TI);
cannam@95: 		    TR = VFNMS(LDK(KP414213562), TI, TJ);
cannam@95: 		    Tm = VSUB(Ti, Tl);
cannam@95: 		    T12 = VADD(Ti, Tl);
cannam@95: 		    T1C = LDW(&(W[TWVL * 18]));
cannam@95: 		    Tz = LDW(&(W[TWVL * 12]));
cannam@95: 		    T15 = LDW(&(W[TWVL * 16]));
cannam@95: 		    {
cannam@95: 			 V T1v, T1y, T1N, T1g, T1J, T1c, T1U, T1V, T1m, T1n, T1s, TS, T1u, TL, T1x;
cannam@95: 			 V T13, T1q, Tn, Tw, T1L, T1f, TT, T1M, T1e, TM, T1R, T1j, T1b, Tx, T1a;
cannam@95: 			 V To, T1T, T1l, T1E, T1O, T1P, T1h, T1i;
cannam@95: 			 T1s = LDW(&(W[TWVL * 8]));
cannam@95: 			 TS = VADD(TQ, TR);
cannam@95: 			 T1u = VSUB(TQ, TR);
cannam@95: 			 TL = VADD(TH, TK);
cannam@95: 			 T1x = VSUB(TH, TK);
cannam@95: 			 T13 = VADD(T11, T12);
cannam@95: 			 T1q = VSUB(T11, T12);
cannam@95: 			 Tn = VADD(Tf, Tm);
cannam@95: 			 Tw = VSUB(Tf, Tm);
cannam@95: 			 T1L = VFMA(LDK(KP923879532), T1u, T1t);
cannam@95: 			 T1v = VFNMS(LDK(KP923879532), T1u, T1t);
cannam@95: 			 T1f = VFMA(LDK(KP923879532), TS, TP);
cannam@95: 			 TT = VFNMS(LDK(KP923879532), TS, TP);
cannam@95: 			 T1M = VFNMS(LDK(KP923879532), T1x, T1w);
cannam@95: 			 T1y = VFMA(LDK(KP923879532), T1x, T1w);
cannam@95: 			 T1e = VFMA(LDK(KP923879532), TL, TE);
cannam@95: 			 TM = VFNMS(LDK(KP923879532), TL, TE);
cannam@95: 			 T1r = VZMUL(T1o, VFMAI(T1q, T1p));
cannam@95: 			 T1R = VZMUL(T1Q, VFNMSI(T1q, T1p));
cannam@95: 			 T14 = VZMUL(TX, VSUB(T10, T13));
cannam@95: 			 T1j = VADD(T10, T13);
cannam@95: 			 T1b = VFMA(LDK(KP707106781), Tw, Tv);
cannam@95: 			 Tx = VFNMS(LDK(KP707106781), Tw, Tv);
cannam@95: 			 T1a = VFMA(LDK(KP707106781), Tn, T8);
cannam@95: 			 To = VFNMS(LDK(KP707106781), Tn, T8);
cannam@95: 			 T1T = VZMULI(T1S, VFMAI(T1M, T1L));
cannam@95: 			 T1N = VZMULI(T1K, VFNMSI(T1M, T1L));
cannam@95: 			 T16 = VZMULI(T15, VFMAI(TT, TM));
cannam@95: 			 TU = VZMULI(Tz, VFNMSI(TT, TM));
cannam@95: 			 T1l = VZMULI(T1k, VFMAI(T1f, T1e));
cannam@95: 			 T1g = VZMULI(T1d, VFNMSI(T1f, T1e));
cannam@95: 			 T1D = VZMUL(T1C, VFMAI(Tx, To));
cannam@95: 			 Ty = VZMUL(T1, VFNMSI(Tx, To));
cannam@95: 			 T1J = VZMUL(T1I, VFMAI(T1b, T1a));
cannam@95: 			 T1c = VZMUL(T19, VFNMSI(T1b, T1a));
cannam@95: 			 T1U = VCONJ(VSUB(T1R, T1T));
cannam@95: 			 T1V = VADD(T1R, T1T);
cannam@95: 			 T1m = VCONJ(VSUB(T1j, T1l));
cannam@95: 			 T1n = VADD(T1j, T1l);
cannam@95: 			 T1z = VZMULI(T1s, VFMAI(T1y, T1v));
cannam@95: 			 T1E = LDW(&(W[TWVL * 20]));
cannam@95: 			 T1O = VCONJ(VSUB(T1J, T1N));
cannam@95: 			 T1P = VADD(T1J, T1N);
cannam@95: 			 T1h = VCONJ(VSUB(T1c, T1g));
cannam@95: 			 T1i = VADD(T1c, T1g);
cannam@95: 			 ST(&(Rp[WS(rs, 6)]), T1V, ms, &(Rp[0]));
cannam@95: 			 ST(&(Rm[WS(rs, 6)]), T1U, -ms, &(Rm[0]));
cannam@95: 			 ST(&(Rp[0]), T1n, ms, &(Rp[0]));
cannam@95: 			 ST(&(Rm[0]), T1m, -ms, &(Rm[0]));
cannam@95: 			 ST(&(Rp[WS(rs, 1)]), T1P, ms, &(Rp[WS(rs, 1)]));
cannam@95: 			 ST(&(Rm[WS(rs, 1)]), T1O, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 			 ST(&(Rp[WS(rs, 7)]), T1i, ms, &(Rp[WS(rs, 1)]));
cannam@95: 			 ST(&(Rm[WS(rs, 7)]), T1h, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 			 T1F = VZMULI(T1E, VFNMSI(T1y, T1v));
cannam@95: 		    }
cannam@95: 		    TV = VCONJ(VSUB(Ty, TU));
cannam@95: 		    TW = VADD(Ty, TU);
cannam@95: 		    T17 = VCONJ(VSUB(T14, T16));
cannam@95: 		    T18 = VADD(T14, T16);
cannam@95: 		    T1B = VADD(T1r, T1z);
cannam@95: 		    T1A = VCONJ(VSUB(T1r, T1z));
cannam@95: 	       }
cannam@95: 	       T1H = VADD(T1D, T1F);
cannam@95: 	       T1G = VCONJ(VSUB(T1D, T1F));
cannam@95: 	       ST(&(Rm[WS(rs, 3)]), TV, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 	       ST(&(Rp[WS(rs, 3)]), TW, ms, &(Rp[WS(rs, 1)]));
cannam@95: 	       ST(&(Rm[WS(rs, 4)]), T17, -ms, &(Rm[0]));
cannam@95: 	       ST(&(Rm[WS(rs, 2)]), T1A, -ms, &(Rm[0]));
cannam@95: 	       ST(&(Rp[WS(rs, 2)]), T1B, ms, &(Rp[0]));
cannam@95: 	       ST(&(Rp[WS(rs, 4)]), T18, ms, &(Rp[0]));
cannam@95: 	       ST(&(Rp[WS(rs, 5)]), T1H, ms, &(Rp[WS(rs, 1)]));
cannam@95: 	       ST(&(Rm[WS(rs, 5)]), T1G, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 	  }
cannam@95:      }
cannam@95:      VLEAVE();
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      VTW(1, 1),
cannam@95:      VTW(1, 2),
cannam@95:      VTW(1, 3),
cannam@95:      VTW(1, 4),
cannam@95:      VTW(1, 5),
cannam@95:      VTW(1, 6),
cannam@95:      VTW(1, 7),
cannam@95:      VTW(1, 8),
cannam@95:      VTW(1, 9),
cannam@95:      VTW(1, 10),
cannam@95:      VTW(1, 11),
cannam@95:      VTW(1, 12),
cannam@95:      VTW(1, 13),
cannam@95:      VTW(1, 14),
cannam@95:      VTW(1, 15),
cannam@95:      {TW_NEXT, VL, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {53, 30, 50, 0} };
cannam@95: 
cannam@95: void XSIMD(codelet_hc2cbdftv_16) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cbdftv_16, &desc, HC2C_VIA_DFT);
cannam@95: }
cannam@95: #else				/* HAVE_FMA */
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 16 -dif -sign 1 -name hc2cbdftv_16 -include hc2cbv.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 103 FP additions, 42 FP multiplications,
cannam@95:  * (or, 99 additions, 38 multiplications, 4 fused multiply/add),
cannam@95:  * 83 stack variables, 3 constants, and 32 memory accesses
cannam@95:  */
cannam@95: #include "hc2cbv.h"
cannam@95: 
cannam@95: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
cannam@95:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
cannam@95:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 30)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(64, rs)) {
cannam@95: 	       V Tf, T16, TZ, T1C, TI, T1a, TV, T1D, T1F, T1G, Ty, T19, TC, T17, TS;
cannam@95: 	       V T10;
cannam@95: 	       {
cannam@95: 		    V T2, TD, T4, TF, Tc, Tb, Td, T6, T8, T9, T3, TE, Ta, T7, T5;
cannam@95: 		    V Te, TX, TY, TG, TH, TT, TU, Tj, TM, Tw, TQ, Tn, TN, Ts, TP;
cannam@95: 		    V Tg, Ti, Th, Tt, Tv, Tu, Tk, Tm, Tl, Tr, Tq, Tp, To, Tx, TA;
cannam@95: 		    V TB, TO, TR;
cannam@95: 		    T2 = LD(&(Rp[0]), ms, &(Rp[0]));
cannam@95: 		    TD = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
cannam@95: 		    T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T4 = VCONJ(T3);
cannam@95: 		    TE = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    TF = VCONJ(TE);
cannam@95: 		    Tc = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0]));
cannam@95: 		    Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    Tb = VCONJ(Ta);
cannam@95: 		    Td = VSUB(Tb, Tc);
cannam@95: 		    T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
cannam@95: 		    T7 = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T8 = VCONJ(T7);
cannam@95: 		    T9 = VSUB(T6, T8);
cannam@95: 		    T5 = VSUB(T2, T4);
cannam@95: 		    Te = VMUL(LDK(KP707106781), VADD(T9, Td));
cannam@95: 		    Tf = VADD(T5, Te);
cannam@95: 		    T16 = VSUB(T5, Te);
cannam@95: 		    TX = VADD(T2, T4);
cannam@95: 		    TY = VADD(TD, TF);
cannam@95: 		    TZ = VSUB(TX, TY);
cannam@95: 		    T1C = VADD(TX, TY);
cannam@95: 		    TG = VSUB(TD, TF);
cannam@95: 		    TH = VMUL(LDK(KP707106781), VSUB(T9, Td));
cannam@95: 		    TI = VADD(TG, TH);
cannam@95: 		    T1a = VSUB(TH, TG);
cannam@95: 		    TT = VADD(T6, T8);
cannam@95: 		    TU = VADD(Tb, Tc);
cannam@95: 		    TV = VSUB(TT, TU);
cannam@95: 		    T1D = VADD(TT, TU);
cannam@95: 		    Tg = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Th = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0]));
cannam@95: 		    Ti = VCONJ(Th);
cannam@95: 		    Tj = VSUB(Tg, Ti);
cannam@95: 		    TM = VADD(Tg, Ti);
cannam@95: 		    Tt = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Tu = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
cannam@95: 		    Tv = VCONJ(Tu);
cannam@95: 		    Tw = VSUB(Tt, Tv);
cannam@95: 		    TQ = VADD(Tt, Tv);
cannam@95: 		    Tk = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Tl = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
cannam@95: 		    Tm = VCONJ(Tl);
cannam@95: 		    Tn = VSUB(Tk, Tm);
cannam@95: 		    TN = VADD(Tk, Tm);
cannam@95: 		    Tr = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    Tp = LD(&(Rm[0]), -ms, &(Rm[0]));
cannam@95: 		    Tq = VCONJ(Tp);
cannam@95: 		    Ts = VSUB(Tq, Tr);
cannam@95: 		    TP = VADD(Tq, Tr);
cannam@95: 		    T1F = VADD(TM, TN);
cannam@95: 		    T1G = VADD(TP, TQ);
cannam@95: 		    To = VFNMS(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tj));
cannam@95: 		    Tx = VFMA(LDK(KP923879532), Ts, VMUL(LDK(KP382683432), Tw));
cannam@95: 		    Ty = VADD(To, Tx);
cannam@95: 		    T19 = VSUB(To, Tx);
cannam@95: 		    TA = VFMA(LDK(KP382683432), Tj, VMUL(LDK(KP923879532), Tn));
cannam@95: 		    TB = VFNMS(LDK(KP382683432), Ts, VMUL(LDK(KP923879532), Tw));
cannam@95: 		    TC = VADD(TA, TB);
cannam@95: 		    T17 = VSUB(TA, TB);
cannam@95: 		    TO = VSUB(TM, TN);
cannam@95: 		    TR = VSUB(TP, TQ);
cannam@95: 		    TS = VMUL(LDK(KP707106781), VSUB(TO, TR));
cannam@95: 		    T10 = VMUL(LDK(KP707106781), VADD(TO, TR));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V T21, T1W, T1u, T20, T1I, T1O, TK, T1S, T12, T1e, T1k, T1A, T1o, T1w, T1c;
cannam@95: 		    V T1M, T1U, T1V, T1T, T1s, T1t, T1r, T1Z, T1E, T1H, T1B, T1N, Tz, TJ, T1;
cannam@95: 		    V T1R, TW, T11, TL, T1d, T1i, T1j, T1h, T1z, T1m, T1n, T1l, T1v, T18, T1b;
cannam@95: 		    V T15, T1L, T13, T1g, T1X, T23, T14, T1f, T1Y, T22, T1p, T1y, T1J, T1Q, T1q;
cannam@95: 		    V T1x, T1K, T1P;
cannam@95: 		    T1U = VADD(T1C, T1D);
cannam@95: 		    T1V = VADD(T1F, T1G);
cannam@95: 		    T21 = VADD(T1U, T1V);
cannam@95: 		    T1T = LDW(&(W[TWVL * 14]));
cannam@95: 		    T1W = VZMUL(T1T, VSUB(T1U, T1V));
cannam@95: 		    T1s = VADD(Tf, Ty);
cannam@95: 		    T1t = VBYI(VADD(TI, TC));
cannam@95: 		    T1r = LDW(&(W[TWVL * 28]));
cannam@95: 		    T1u = VZMULI(T1r, VSUB(T1s, T1t));
cannam@95: 		    T1Z = LDW(&(W[0]));
cannam@95: 		    T20 = VZMULI(T1Z, VADD(T1s, T1t));
cannam@95: 		    T1E = VSUB(T1C, T1D);
cannam@95: 		    T1H = VBYI(VSUB(T1F, T1G));
cannam@95: 		    T1B = LDW(&(W[TWVL * 22]));
cannam@95: 		    T1I = VZMUL(T1B, VSUB(T1E, T1H));
cannam@95: 		    T1N = LDW(&(W[TWVL * 6]));
cannam@95: 		    T1O = VZMUL(T1N, VADD(T1E, T1H));
cannam@95: 		    Tz = VSUB(Tf, Ty);
cannam@95: 		    TJ = VBYI(VSUB(TC, TI));
cannam@95: 		    T1 = LDW(&(W[TWVL * 12]));
cannam@95: 		    TK = VZMULI(T1, VADD(Tz, TJ));
cannam@95: 		    T1R = LDW(&(W[TWVL * 16]));
cannam@95: 		    T1S = VZMULI(T1R, VSUB(Tz, TJ));
cannam@95: 		    TW = VBYI(VSUB(TS, TV));
cannam@95: 		    T11 = VSUB(TZ, T10);
cannam@95: 		    TL = LDW(&(W[TWVL * 10]));
cannam@95: 		    T12 = VZMUL(TL, VADD(TW, T11));
cannam@95: 		    T1d = LDW(&(W[TWVL * 18]));
cannam@95: 		    T1e = VZMUL(T1d, VSUB(T11, TW));
cannam@95: 		    T1i = VBYI(VADD(T1a, T19));
cannam@95: 		    T1j = VADD(T16, T17);
cannam@95: 		    T1h = LDW(&(W[TWVL * 4]));
cannam@95: 		    T1k = VZMULI(T1h, VADD(T1i, T1j));
cannam@95: 		    T1z = LDW(&(W[TWVL * 24]));
cannam@95: 		    T1A = VZMULI(T1z, VSUB(T1j, T1i));
cannam@95: 		    T1m = VBYI(VADD(TV, TS));
cannam@95: 		    T1n = VADD(TZ, T10);
cannam@95: 		    T1l = LDW(&(W[TWVL * 2]));
cannam@95: 		    T1o = VZMUL(T1l, VADD(T1m, T1n));
cannam@95: 		    T1v = LDW(&(W[TWVL * 26]));
cannam@95: 		    T1w = VZMUL(T1v, VSUB(T1n, T1m));
cannam@95: 		    T18 = VSUB(T16, T17);
cannam@95: 		    T1b = VBYI(VSUB(T19, T1a));
cannam@95: 		    T15 = LDW(&(W[TWVL * 20]));
cannam@95: 		    T1c = VZMULI(T15, VSUB(T18, T1b));
cannam@95: 		    T1L = LDW(&(W[TWVL * 8]));
cannam@95: 		    T1M = VZMULI(T1L, VADD(T1b, T18));
cannam@95: 		    T13 = VADD(TK, T12);
cannam@95: 		    ST(&(Rp[WS(rs, 3)]), T13, ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    T1g = VCONJ(VSUB(T1e, T1c));
cannam@95: 		    ST(&(Rm[WS(rs, 5)]), T1g, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T1X = VADD(T1S, T1W);
cannam@95: 		    ST(&(Rp[WS(rs, 4)]), T1X, ms, &(Rp[0]));
cannam@95: 		    T23 = VCONJ(VSUB(T21, T20));
cannam@95: 		    ST(&(Rm[0]), T23, -ms, &(Rm[0]));
cannam@95: 		    T14 = VCONJ(VSUB(T12, TK));
cannam@95: 		    ST(&(Rm[WS(rs, 3)]), T14, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T1f = VADD(T1c, T1e);
cannam@95: 		    ST(&(Rp[WS(rs, 5)]), T1f, ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    T1Y = VCONJ(VSUB(T1W, T1S));
cannam@95: 		    ST(&(Rm[WS(rs, 4)]), T1Y, -ms, &(Rm[0]));
cannam@95: 		    T22 = VADD(T20, T21);
cannam@95: 		    ST(&(Rp[0]), T22, ms, &(Rp[0]));
cannam@95: 		    T1p = VADD(T1k, T1o);
cannam@95: 		    ST(&(Rp[WS(rs, 1)]), T1p, ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    T1y = VCONJ(VSUB(T1w, T1u));
cannam@95: 		    ST(&(Rm[WS(rs, 7)]), T1y, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T1J = VADD(T1A, T1I);
cannam@95: 		    ST(&(Rp[WS(rs, 6)]), T1J, ms, &(Rp[0]));
cannam@95: 		    T1Q = VCONJ(VSUB(T1O, T1M));
cannam@95: 		    ST(&(Rm[WS(rs, 2)]), T1Q, -ms, &(Rm[0]));
cannam@95: 		    T1q = VCONJ(VSUB(T1o, T1k));
cannam@95: 		    ST(&(Rm[WS(rs, 1)]), T1q, -ms, &(Rm[WS(rs, 1)]));
cannam@95: 		    T1x = VADD(T1u, T1w);
cannam@95: 		    ST(&(Rp[WS(rs, 7)]), T1x, ms, &(Rp[WS(rs, 1)]));
cannam@95: 		    T1K = VCONJ(VSUB(T1I, T1A));
cannam@95: 		    ST(&(Rm[WS(rs, 6)]), T1K, -ms, &(Rm[0]));
cannam@95: 		    T1P = VADD(T1M, T1O);
cannam@95: 		    ST(&(Rp[WS(rs, 2)]), T1P, ms, &(Rp[0]));
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95:      VLEAVE();
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      VTW(1, 1),
cannam@95:      VTW(1, 2),
cannam@95:      VTW(1, 3),
cannam@95:      VTW(1, 4),
cannam@95:      VTW(1, 5),
cannam@95:      VTW(1, 6),
cannam@95:      VTW(1, 7),
cannam@95:      VTW(1, 8),
cannam@95:      VTW(1, 9),
cannam@95:      VTW(1, 10),
cannam@95:      VTW(1, 11),
cannam@95:      VTW(1, 12),
cannam@95:      VTW(1, 13),
cannam@95:      VTW(1, 14),
cannam@95:      VTW(1, 15),
cannam@95:      {TW_NEXT, VL, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {99, 38, 4, 0} };
cannam@95: 
cannam@95: void XSIMD(codelet_hc2cbdftv_16) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cbdftv_16, &desc, HC2C_VIA_DFT);
cannam@95: }
cannam@95: #endif				/* HAVE_FMA */