r2cbiii_25.c

快速fft变换

			 }
		    }
	       }
	  }
	  R0[WS(rs, 1)] = FNMS(KP1_752613360, T24, T1P);
	  T25 = FMA(KP438153340, T24, T1P);
	  T27 = FMA(KP979740652, T26, T25);
	  T2b = FNMS(KP979740652, T26, T25);
	  R1[WS(rs, 8)] = -(FMA(KP1_606007150, T2a, T27));
	  R0[WS(rs, 6)] = FNMS(KP1_606007150, T2a, T27);
	  R1[WS(rs, 3)] = -(FMA(KP1_666834356, T2c, T2b));
	  R0[WS(rs, 11)] = FNMS(KP1_666834356, T2c, T2b);
     }
}

static const kr2c_desc desc = { 25, "r2cbIII_25", {32, 0, 120, 0}, &GENUS };

void X(codelet_r2cbIII_25) (planner *p) {
     X(kr2c_register) (p, r2cbIII_25, &desc);
}

#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_r2cb -compact -variables 4 -pipeline-latency 4 -sign 1 -n 25 -name r2cbIII_25 -dft-III -include r2cbIII.h */

/*
 * This function contains 152 FP additions, 98 FP multiplications,
 * (or, 100 additions, 46 multiplications, 52 fused multiply/add),
 * 65 stack variables, 21 constants, and 50 memory accesses
 */
#include "r2cbIII.h"

static void r2cbIII_25(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
{
     DK(KP968583161, +0.968583161128631119490168375464735813836012403);
     DK(KP248689887, +0.248689887164854788242283746006447968417567406);
     DK(KP684547105, +0.684547105928688673732283357621209269889519233);
     DK(KP728968627, +0.728968627421411523146730319055259111372571664);
     DK(KP062790519, +0.062790519529313376076178224565631133122484832);
     DK(KP998026728, +0.998026728428271561952336806863450553336905220);
     DK(KP876306680, +0.876306680043863587308115903922062583399064238);
     DK(KP481753674, +0.481753674101715274987191502872129653528542010);
     DK(KP535826794, +0.535826794978996618271308767867639978063575346);
     DK(KP844327925, +0.844327925502015078548558063966681505381659241);
     DK(KP904827052, +0.904827052466019527713668647932697593970413911);
     DK(KP425779291, +0.425779291565072648862502445744251703979973042);
     DK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DK(KP951056516, +0.951056516295153572116439333379382143405698634);
     DK(KP587785252, +0.587785252292473129168705954639072768597652438);
     DK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DK(KP500000000, +0.500000000000000000000000000000000000000000000);
     DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
     DK(KP1_118033988, +1.118033988749894848204586834365638117720309180);
     DK(KP1_175570504, +1.175570504584946258337411909278145537195304875);
     DK(KP1_902113032, +1.902113032590307144232878666758764286811397268);
     INT i;
     for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(rs), MAKE_VOLATILE_STRIDE(csr), MAKE_VOLATILE_STRIDE(csi)) {
	  E TS, T1O, T5, TP, T1N, TI, TH, Te, T17, T2h, T1y, T1V, T10, T2g, T1x;
	  E T1S, Tz, Ty, Tn, T1m, T2e, T1B, T22, T1f, T2d, T1A, T1Z, TQ, TR;
	  TQ = Ci[WS(csi, 2)];
	  TR = Ci[WS(csi, 7)];
	  TS = FNMS(KP1_175570504, TR, KP1_902113032 * TQ);
	  T1O = FMA(KP1_902113032, TR, KP1_175570504 * TQ);
	  {
	       E T1, T4, TN, T2, T3, TO;
	       T1 = Cr[WS(csr, 12)];
	       T2 = Cr[WS(csr, 7)];
	       T3 = Cr[WS(csr, 2)];
	       T4 = T2 + T3;
	       TN = KP1_118033988 * (T3 - T2);
	       T5 = FMA(KP2_000000000, T4, T1);
	       TO = FMS(KP500000000, T4, T1);
	       TP = TN - TO;
	       T1N = TO + TN;
	  }
	  {
	       E T6, Td, T15, TU, T14, T11, TX, TY;
	       T6 = Cr[WS(csr, 11)];
	       TI = Ci[WS(csi, 11)];
	       {
		    E T7, T8, T9, Ta, Tb, Tc;
		    T7 = Cr[WS(csr, 6)];
		    T8 = Cr[WS(csr, 8)];
		    T9 = T7 + T8;
		    Ta = Cr[WS(csr, 1)];
		    Tb = Cr[WS(csr, 3)];
		    Tc = Ta + Tb;
		    Td = T9 + Tc;
		    T15 = Ta - Tb;
		    TU = KP559016994 * (Tc - T9);
		    T14 = T8 - T7;
	       }
	       {
		    E TB, TC, TD, TE, TF, TG;
		    TB = Ci[WS(csi, 6)];
		    TC = Ci[WS(csi, 8)];
		    TD = TB - TC;
		    TE = Ci[WS(csi, 1)];
		    TF = Ci[WS(csi, 3)];
		    TG = TE - TF;
		    TH = TD + TG;
		    T11 = KP559016994 * (TD - TG);
		    TX = TB + TC;
		    TY = TE + TF;
	       }
	       Te = T6 + Td;
	       {
		    E T16, T1T, T13, T1U, T12;
		    T16 = FMA(KP587785252, T14, KP951056516 * T15);
		    T1T = FNMS(KP587785252, T15, KP951056516 * T14);
		    T12 = FNMS(KP250000000, TH, TI);
		    T13 = T11 - T12;
		    T1U = T11 + T12;
		    T17 = T13 - T16;
		    T2h = T1T - T1U;
		    T1y = T16 + T13;
		    T1V = T1T + T1U;
	       }
	       {
		    E TZ, T1R, TW, T1Q, TV;
		    TZ = FNMS(KP951056516, TY, KP587785252 * TX);
		    T1R = FMA(KP951056516, TX, KP587785252 * TY);
		    TV = FMS(KP250000000, Td, T6);
		    TW = TU - TV;
		    T1Q = TV + TU;
		    T10 = TW + TZ;
		    T2g = T1Q + T1R;
		    T1x = TZ - TW;
		    T1S = T1Q - T1R;
	       }
	  }
	  {
	       E Tf, Tm, T1k, T19, T1j, T1g, T1c, T1d;
	       Tf = Cr[WS(csr, 10)];
	       Tz = Ci[WS(csi, 10)];
	       {
		    E Tg, Th, Ti, Tj, Tk, Tl;
		    Tg = Cr[WS(csr, 5)];
		    Th = Cr[WS(csr, 9)];
		    Ti = Tg + Th;
		    Tj = Cr[0];
		    Tk = Cr[WS(csr, 4)];
		    Tl = Tj + Tk;
		    Tm = Ti + Tl;
		    T1k = Tj - Tk;
		    T19 = KP559016994 * (Tl - Ti);
		    T1j = Th - Tg;
	       }
	       {
		    E Ts, Tt, Tu, Tv, Tw, Tx;
		    Ts = Ci[WS(csi, 4)];
		    Tt = Ci[0];
		    Tu = Ts - Tt;
		    Tv = Ci[WS(csi, 5)];
		    Tw = Ci[WS(csi, 9)];
		    Tx = Tv - Tw;
		    Ty = Tu - Tx;
		    T1g = KP559016994 * (Tx + Tu);
		    T1c = Tv + Tw;
		    T1d = Tt + Ts;
	       }
	       Tn = Tf + Tm;
	       {
		    E T1l, T20, T1i, T21, T1h;
		    T1l = FMA(KP587785252, T1j, KP951056516 * T1k);
		    T20 = FNMS(KP587785252, T1k, KP951056516 * T1j);
		    T1h = FMA(KP250000000, Ty, Tz);
		    T1i = T1g - T1h;
		    T21 = T1g + T1h;
		    T1m = T1i - T1l;
		    T2e = T21 - T20;
		    T1B = T1l + T1i;
		    T22 = T20 + T21;
	       }
	       {
		    E T1e, T1Y, T1b, T1X, T1a;
		    T1e = FNMS(KP951056516, T1d, KP587785252 * T1c);
		    T1Y = FMA(KP951056516, T1c, KP587785252 * T1d);
		    T1a = FMS(KP250000000, Tm, Tf);
		    T1b = T19 - T1a;
		    T1X = T1a + T19;
		    T1f = T1b + T1e;
		    T2d = T1X + T1Y;
		    T1A = T1e - T1b;
		    T1Z = T1X - T1Y;
	       }
	  }
	  {
	       E Tq, To, Tp, TK, TM, TA, TJ, TL, Tr;
	       Tq = KP1_118033988 * (Tn - Te);
	       To = Te + Tn;
	       Tp = FMS(KP500000000, To, T5);
	       TA = Ty - Tz;
	       TJ = TH + TI;
	       TK = FNMS(KP1_902113032, TJ, KP1_175570504 * TA);
	       TM = FMA(KP1_175570504, TJ, KP1_902113032 * TA);
	       R0[0] = FMA(KP2_000000000, To, T5);
	       TL = Tq - Tp;
	       R0[WS(rs, 5)] = TL + TM;
	       R1[WS(rs, 7)] = TM - TL;
	       Tr = Tp + Tq;
	       R1[WS(rs, 2)] = Tr + TK;
	       R0[WS(rs, 10)] = TK - Tr;
	  }
	  {
	       E T2q, T2s, T2k, T2j, T2l, T2m, T2r, T2n;
	       {
		    E T2o, T2p, T2f, T2i;
		    T2o = FNMS(KP904827052, T2d, KP425779291 * T2e);
		    T2p = FNMS(KP535826794, T2h, KP844327925 * T2g);
		    T2q = FNMS(KP1_902113032, T2p, KP1_175570504 * T2o);
		    T2s = FMA(KP1_175570504, T2p, KP1_902113032 * T2o);
		    T2k = T1N + T1O;
		    T2f = FMA(KP425779291, T2d, KP904827052 * T2e);
		    T2i = FMA(KP535826794, T2g, KP844327925 * T2h);
		    T2j = T2f - T2i;
		    T2l = FMA(KP500000000, T2j, T2k);
		    T2m = KP1_118033988 * (T2i + T2f);
	       }
	       R0[WS(rs, 2)] = FMS(KP2_000000000, T2j, T2k);
	       T2r = T2m - T2l;
	       R0[WS(rs, 7)] = T2r + T2s;
	       R1[WS(rs, 9)] = T2s - T2r;
	       T2n = T2l + T2m;
	       R1[WS(rs, 4)] = T2n + T2q;
	       R0[WS(rs, 12)] = T2q - T2n;
	  }
	  {
	       E T1u, T1w, TT, T1o, T1p, T1q, T1v, T1r;
	       {
		    E T1s, T1t, T18, T1n;
		    T1s = FMA(KP481753674, T10, KP876306680 * T17);
		    T1t = FMA(KP844327925, T1f, KP535826794 * T1m);
		    T1u = FMA(KP1_902113032, T1s, KP1_175570504 * T1t);
		    T1w = FNMS(KP1_175570504, T1s, KP1_902113032 * T1t);
		    TT = TP - TS;
		    T18 = FNMS(KP481753674, T17, KP876306680 * T10);
		    T1n = FNMS(KP844327925, T1m, KP535826794 * T1f);
		    T1o = T18 + T1n;
		    T1p = FMS(KP500000000, T1o, TT);
		    T1q = KP1_118033988 * (T1n - T18);
	       }
	       R0[WS(rs, 1)] = FMA(KP2_000000000, T1o, TT);
	       T1v = T1q - T1p;
	       R0[WS(rs, 6)] = T1v + T1w;
	       R1[WS(rs, 8)] = T1w - T1v;
	       T1r = T1p + T1q;
	       R1[WS(rs, 3)] = T1r + T1u;
	       R0[WS(rs, 11)] = T1u - T1r;
	  }
	  {
	       E T1H, T1L, T1E, T1D, T1I, T1J, T1M, T1K;
	       {
		    E T1F, T1G, T1z, T1C;
		    T1F = FNMS(KP062790519, T1B, KP998026728 * T1A);
		    T1G = FNMS(KP684547105, T1x, KP728968627 * T1y);
		    T1H = FNMS(KP1_902113032, T1G, KP1_175570504 * T1F);
		    T1L = FMA(KP1_175570504, T1G, KP1_902113032 * T1F);
		    T1E = TP + TS;
		    T1z = FMA(KP728968627, T1x, KP684547105 * T1y);
		    T1C = FMA(KP062790519, T1A, KP998026728 * T1B);
		    T1D = T1z + T1C;
		    T1I = FMA(KP500000000, T1D, T1E);
		    T1J = KP1_118033988 * (T1C - T1z);
	       }
	       R1[WS(rs, 1)] = FMS(KP2_000000000, T1D, T1E);
	       T1M = T1J - T1I;
	       R0[WS(rs, 9)] = T1L - T1M;
	       R1[WS(rs, 6)] = T1L + T1M;
	       T1K = T1I + T1J;
	       R1[WS(rs, 11)] = T1H - T1K;
	       R0[WS(rs, 4)] = T1H + T1K;
	  }
	  {
	       E T2a, T2c, T1P, T24, T25, T26, T2b, T27;
	       {
		    E T28, T29, T1W, T23;
		    T28 = FMA(KP248689887, T1S, KP968583161 * T1V);
		    T29 = FMA(KP481753674, T1Z, KP876306680 * T22);
		    T2a = FMA(KP1_902113032, T28, KP1_175570504 * T29);
		    T2c = FNMS(KP1_175570504, T28, KP1_902113032 * T29);
		    T1P = T1N - T1O;
		    T1W = FNMS(KP248689887, T1V, KP968583161 * T1S);
		    T23 = FNMS(KP481753674, T22, KP876306680 * T1Z);
		    T24 = T1W + T23;
		    T25 = FMS(KP500000000, T24, T1P);
		    T26 = KP1_118033988 * (T23 - T1W);
	       }
	       R1[0] = FMA(KP2_000000000, T24, T1P);
	       T2b = T26 - T25;
	       R1[WS(rs, 5)] = T2b + T2c;
	       R0[WS(rs, 8)] = T2c - T2b;
	       T27 = T25 + T26;
	       R0[WS(rs, 3)] = T27 + T2a;
	       R1[WS(rs, 10)] = T2a - T27;
	  }
     }
}

static const kr2c_desc desc = { 25, "r2cbIII_25", {100, 46, 52, 0}, &GENUS };

void X(codelet_r2cbIII_25) (planner *p) {
     X(kr2c_register) (p, r2cbIII_25, &desc);
}

#endif				/* HAVE_FMA */