q1fv_8.c

快速fft变换

/*
 * Copyright (c) 2003, 2007-8 Matteo Frigo
 * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 */

/* This file was automatically generated --- DO NOT EDIT */
/* Generated on Sat Nov 15 20:54:15 EST 2008 */

#include "codelet-dft.h"

#ifdef HAVE_FMA

/* Generated by: ../../../genfft/gen_twidsq_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1fv_8 -include q1f.h */

/*
 * This function contains 264 FP additions, 192 FP multiplications,
 * (or, 184 additions, 112 multiplications, 80 fused multiply/add),
 * 117 stack variables, 1 constants, and 128 memory accesses
 */
#include "q1f.h"

static void q1fv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
{
     DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
     INT m;
     R *x;
     x = ri;
     for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(rs), MAKE_VOLATILE_STRIDE(vs)) {
	  V T42, T43, T1U, T1V, T2Y, T2Z, TT, TS;
	  {
	       V T3, Te, T1E, T1P, Tu, Tp, T25, T20, T2b, T2m, T3M, T2x, T2C, T3X, TA;
	       V TL, T48, T4d, T17, T11, TW, T1i, T2I, T1y, T1t, T2T, T3f, T3q, T34, T39;
	       V T3G, T3B, Ts, Tv, Tf, Ta, T23, T26, T1Q, T1L, T2A, T2D, T2n, T2i, T4b;
	       V T4e, T3Y, T3T, TZ, T12, TM, TH, T35, T2L, T3j, T1w, T1z, T1j, T1e, T36;
	       V T2O, T3C, T3i, T3k;
	       {
		    V T3d, T32, T3e, T3o, T3p, T33;
		    {
			 V T2v, T2w, T3V, T46, T3W;
			 {
			      V T1, T2, Tc, Td, T1C, T1D, T1N, T1O;
			      T1 = LD(&(x[0]), ms, &(x[0]));
			      T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
			      Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
			      Td = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
			      T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
			      T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
			      T1N = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
			      T1O = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
			      {
				   V T29, T1Y, T1Z, T2a, T2k, T2l, Tn, To, T3K, T3L;
				   T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
				   T3 = VSUB(T1, T2);
				   Tn = VADD(T1, T2);
				   Te = VSUB(Tc, Td);
				   To = VADD(Tc, Td);
				   T1E = VSUB(T1C, T1D);
				   T1Y = VADD(T1C, T1D);
				   T1P = VSUB(T1N, T1O);
				   T1Z = VADD(T1N, T1O);
				   T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
				   T2k = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
				   T2l = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
				   Tu = VSUB(Tn, To);
				   Tp = VADD(Tn, To);
				   T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
				   T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
				   T25 = VSUB(T1Y, T1Z);
				   T20 = VADD(T1Y, T1Z);
				   T2v = VADD(T29, T2a);
				   T2b = VSUB(T29, T2a);
				   T2w = VADD(T2k, T2l);
				   T2m = VSUB(T2k, T2l);
				   T3V = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
				   T46 = VADD(T3K, T3L);
				   T3M = VSUB(T3K, T3L);
				   T3W = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
			      }
			 }
			 {
			      V T15, TU, T16, T1g, TV, T1h;
			      {
				   V Ty, Tz, TJ, TK, T47;
				   Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
				   Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
				   TJ = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
				   T2x = VADD(T2v, T2w);
				   T2C = VSUB(T2v, T2w);
				   TK = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
				   T47 = VADD(T3V, T3W);
				   T3X = VSUB(T3V, T3W);
				   T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
				   TA = VSUB(Ty, Tz);
				   TU = VADD(Ty, Tz);
				   T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
				   T1g = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
				   TL = VSUB(TJ, TK);
				   TV = VADD(TJ, TK);
				   T48 = VADD(T46, T47);
				   T4d = VSUB(T46, T47);
				   T1h = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
			      }
			      {
				   V T2G, T1r, T2H, T2R, T1s, T2S;
				   T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
				   T17 = VSUB(T15, T16);
				   T1r = VADD(T15, T16);
				   T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
				   T11 = VSUB(TU, TV);
				   TW = VADD(TU, TV);
				   T2R = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
				   T1i = VSUB(T1g, T1h);
				   T1s = VADD(T1g, T1h);
				   T2S = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
				   T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
				   T2I = VSUB(T2G, T2H);
				   T32 = VADD(T2G, T2H);
				   T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
				   T3o = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
				   T3p = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
				   T1y = VSUB(T1r, T1s);
				   T1t = VADD(T1r, T1s);
				   T33 = VADD(T2R, T2S);
				   T2T = VSUB(T2R, T2S);
			      }
			 }
		    }
		    {
			 V T2y, T2e, T3Q, T2z, T2h, T49, T3P, T3R;
			 {
			      V T6, Tq, T1I, Tr, T9, T21, T1H, T1J;
			      {
				   V T4, T3z, T3A, T5, T7, T8, T1F, T1G;
				   T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
				   T3f = VSUB(T3d, T3e);
				   T3z = VADD(T3d, T3e);
				   T3q = VSUB(T3o, T3p);
				   T3A = VADD(T3o, T3p);
				   T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
				   T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
				   T34 = VADD(T32, T33);
				   T39 = VSUB(T32, T33);
				   T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
				   T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
				   T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
				   T3G = VSUB(T3z, T3A);
				   T3B = VADD(T3z, T3A);
				   T6 = VSUB(T4, T5);
				   Tq = VADD(T4, T5);
				   T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
				   Tr = VADD(T7, T8);
				   T9 = VSUB(T7, T8);
				   T21 = VADD(T1F, T1G);
				   T1H = VSUB(T1F, T1G);
				   T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
			      }
			      {
				   V T2f, T22, T1K, T2g, T2c, T2d, T3N, T3O;
				   T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
				   T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
				   T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
				   Ts = VADD(Tq, Tr);
				   Tv = VSUB(Tr, Tq);
				   Tf = VSUB(T9, T6);
				   Ta = VADD(T6, T9);
				   T22 = VADD(T1I, T1J);
				   T1K = VSUB(T1I, T1J);
				   T2y = VADD(T2c, T2d);
				   T2e = VSUB(T2c, T2d);
				   T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
				   T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
				   T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
				   T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
				   T23 = VADD(T21, T22);
				   T26 = VSUB(T22, T21);
				   T1Q = VSUB(T1K, T1H);
				   T1L = VADD(T1H, T1K);
				   T2z = VADD(T2f, T2g);
				   T2h = VSUB(T2f, T2g);
				   T49 = VADD(T3N, T3O);
				   T3P = VSUB(T3N, T3O);
				   T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
			      }
			 }
			 {
			      V TX, TD, T1b, TY, TG, T1u, T1a, T1c;
			      {
				   V TE, T4a, T3S, TF, TB, TC, T18, T19;
				   TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
				   TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
				   TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
				   T2A = VADD(T2y, T2z);
				   T2D = VSUB(T2z, T2y);
				   T2n = VSUB(T2h, T2e);
				   T2i = VADD(T2e, T2h);
				   T4a = VADD(T3Q, T3R);
				   T3S = VSUB(T3Q, T3R);
				   TX = VADD(TB, TC);
				   TD = VSUB(TB, TC);
				   TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
				   T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
				   T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
				   T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
				   T4b = VADD(T49, T4a);
				   T4e = VSUB(T4a, T49);
				   T3Y = VSUB(T3S, T3P);
				   T3T = VADD(T3P, T3S);
				   TY = VADD(TE, TF);
				   TG = VSUB(TE, TF);
				   T1u = VADD(T18, T19);
				   T1a = VSUB(T18, T19);
				   T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
			      }
			      {
				   V T2M, T1v, T1d, T2N, T2J, T2K, T3g, T3h;
				   T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
				   T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
				   T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
				   TZ = VADD(TX, TY);
				   T12 = VSUB(TY, TX);
				   TM = VSUB(TG, TD);
				   TH = VADD(TD, TG);
				   T1v = VADD(T1b, T1c);
				   T1d = VSUB(T1b, T1c);
				   T35 = VADD(T2J, T2K);
				   T2L = VSUB(T2J, T2K);
				   T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
				   T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
				   T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
				   T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
				   T1w = VADD(T1u, T1v);
				   T1z = VSUB(T1v, T1u);
				   T1j = VSUB(T1d, T1a);
				   T1e = VADD(T1a, T1d);
				   T36 = VADD(T2M, T2N);
				   T2O = VSUB(T2M, T2N);
				   T3C = VADD(T3g, T3h);
				   T3i = VSUB(T3g, T3h);
				   T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
			      }
			 }
		    }
	       }
	       {
		    V T3a, T2U, T2P, T3H, T3r, T3m, T13, T27, T3b, T4f;
		    {
			 V T37, T3E, T2B, T24;
			 {
			      V T3D, T3l, Tt, T4c;
			      ST(&(x[0]), VADD(Tp, Ts), ms, &(x[0]));
			      ST(&(x[WS(rs, 2)]), VADD(T1t, T1w), ms, &(x[0]));
			      ST(&(x[WS(rs, 7)]), VADD(T48, T4b), ms, &(x[WS(rs, 1)]));
			      T37 = VADD(T35, T36);
			      T3a = VSUB(T36, T35);
			      T2U = VSUB(T2O, T2L);
			      T2P = VADD(T2L, T2O);
			      T3D = VADD(T3j, T3k);
			      T3l = VSUB(T3j, T3k);
			      ST(&(x[WS(rs, 4)]), VADD(T2x, T2A), ms, &(x[0]));
			      ST(&(x[WS(rs, 3)]), VADD(T20, T23), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 5)]), VADD(T34, T37), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 1)]), VADD(TW, TZ), ms, &(x[WS(rs, 1)]));
			      Tt = BYTWJ(&(W[TWVL * 6]), VSUB(Tp, Ts));
			      T4c = BYTWJ(&(W[TWVL * 6]), VSUB(T48, T4b));
			      T3E = VADD(T3C, T3D);
			      T3H = VSUB(T3D, T3C);
			      T3r = VSUB(T3l, T3i);
			      T3m = VADD(T3i, T3l);
			      T2B = BYTWJ(&(W[TWVL * 6]), VSUB(T2x, T2A));
			      T24 = BYTWJ(&(W[TWVL * 6]), VSUB(T20, T23));
			      ST(&(x[WS(vs, 4)]), Tt, ms, &(x[WS(vs, 4)]));
			      ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
			      ST(&(x[WS(rs, 6)]), VADD(T3B, T3E), ms, &(x[0]));
			 }
			 {
			      V T38, T1A, Tw, T10, T1x, T3F, T2E, T3I;
			      T10 = BYTWJ(&(W[TWVL * 6]), VSUB(TW, TZ));
			      T1x = BYTWJ(&(W[TWVL * 6]), VSUB(T1t, T1w));
			      T3F = BYTWJ(&(W[TWVL * 6]), VSUB(T3B, T3E));
			      ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 4)]));
			      ST(&(x[WS(vs, 4) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
			      T38 = BYTWJ(&(W[TWVL * 6]), VSUB(T34, T37));
			      T1A = BYTWJ(&(W[TWVL * 10]), VFNMSI(T1z, T1y));
			      Tw = BYTWJ(&(W[TWVL * 10]), VFNMSI(Tv, Tu));
			      ST(&(x[WS(vs, 4) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
			      ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 4)]));
			      ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 4)]));
			      T2E = BYTWJ(&(W[TWVL * 10]), VFNMSI(T2D, T2C));
			      T3I = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3H, T3G));
			      ST(&(x[WS(vs, 4) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
			      ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1A, ms, &(x[WS(vs, 6)]));
			      ST(&(x[WS(vs, 6)]), Tw, ms, &(x[WS(vs, 6)]));
			      T13 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T12, T11));
			      T27 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T26, T25));
			      T3b = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3a, T39));
			      ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2E, ms, &(x[WS(vs, 6)]));
			      ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3I, ms, &(x[WS(vs, 6)]));
			      T4f = BYTWJ(&(W[TWVL * 10]), VFNMSI(T4e, T4d));
			 }
		    }
		    {
			 V Tj, Tk, T2r, T2j, Ti, Th, T2o, T2s, T1M, T1R, T41, T40;
			 {
			      V T3c, T4g, T3J, T2F, Tx, T1B;
			      Tx = BYTWJ(&(W[TWVL * 2]), VFMAI(Tv, Tu));
			      T1B = BYTWJ(&(W[TWVL * 2]), VFMAI(T1z, T1y));
			      ST(&(x[WS(vs, 6) + WS(rs, 1)]), T13, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
			      ST(&(x[WS(vs, 6) + WS(rs, 3)]), T27, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
			      ST(&(x[WS(vs, 6) + WS(rs, 5)]), T3b, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
			      T3c = BYTWJ(&(W[TWVL * 2]), VFMAI(T3a, T39));
			      T4g = BYTWJ(&(W[TWVL * 2]), VFMAI(T4e, T4d));
			      ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4f, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
			      ST(&(x[WS(vs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
			      ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 2)]));