hc2cbdftv_10.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 21:14:16 EST 2008 */

#include "codelet-rdft.h"

#ifdef HAVE_FMA

/* Generated by: ../../../genfft/gen_hc2cdft_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 10 -dif -sign 1 -name hc2cbdftv_10 -include hc2cbv.h */

/*
 * This function contains 61 FP additions, 50 FP multiplications,
 * (or, 33 additions, 22 multiplications, 28 fused multiply/add),
 * 76 stack variables, 4 constants, and 20 memory accesses
 */
#include "hc2cbv.h"

static void hc2cbdftv_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     INT m;
     for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 18)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(rs)) {
	  V Ts, T4, TR, T1, TZ, TD, Ty, Tn, Ti, TT, T11, TJ, T15, Tr, TN;
	  V TE, Tv, To, Tb, T8, Tw, Te, Tx, Th, Tt, T7, T9, T2, T3, Tc;
	  V Td, Tf, Tg, T5, T6, Tu, Ta;
	  T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	  T3 = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
	  Tc = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
	  Td = LD(&(Rm[0]), -ms, &(Rm[0]));
	  Tf = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	  Tg = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
	  T5 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
	  T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
	  T8 = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
	  Ts = VFMACONJ(T3, T2);
	  T4 = VFNMSCONJ(T3, T2);
	  Tw = VFMACONJ(Td, Tc);
	  Te = VFNMSCONJ(Td, Tc);
	  Tx = VFMACONJ(Tg, Tf);
	  Th = VFMSCONJ(Tg, Tf);
	  Tt = VFMACONJ(T6, T5);
	  T7 = VFNMSCONJ(T6, T5);
	  T9 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	  TR = LDW(&(W[TWVL * 8]));
	  T1 = LDW(&(W[TWVL * 4]));
	  TZ = LDW(&(W[TWVL * 12]));
	  TD = VSUB(Tw, Tx);
	  Ty = VADD(Tw, Tx);
	  Tn = VSUB(Te, Th);
	  Ti = VADD(Te, Th);
	  Tu = VFMACONJ(T9, T8);
	  Ta = VFMSCONJ(T9, T8);
	  TT = LDW(&(W[TWVL * 6]));
	  T11 = LDW(&(W[TWVL * 10]));
	  TJ = LDW(&(W[TWVL * 16]));
	  T15 = LDW(&(W[0]));
	  Tr = LDW(&(W[TWVL * 2]));
	  TN = LDW(&(W[TWVL * 14]));
	  TE = VSUB(Tt, Tu);
	  Tv = VADD(Tt, Tu);
	  To = VSUB(T7, Ta);
	  Tb = VADD(T7, Ta);
	  {
	       V TV, TF, Tz, TB, TL, Tp, Tj, Tl, T17, TA, TS, Tk, TC, TU, TK;
	       V Tm, TO, TG, T12, TW, T16, TM, T10, Tq, TX, TY, T18, T19, TQ, TP;
	       V T13, T14, TI, TH;
	       TV = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TD, TE));
	       TF = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TD));
	       Tz = VADD(Tv, Ty);
	       TB = VSUB(Tv, Ty);
	       TL = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, To));
	       Tp = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), To, Tn));
	       Tj = VADD(Tb, Ti);
	       Tl = VSUB(Tb, Ti);
	       T17 = VADD(Ts, Tz);
	       TA = VFNMS(LDK(KP250000000), Tz, Ts);
	       TS = VZMULI(TR, VADD(T4, Tj));
	       Tk = VFNMS(LDK(KP250000000), Tj, T4);
	       TC = VFNMS(LDK(KP559016994), TB, TA);
	       TU = VFMA(LDK(KP559016994), TB, TA);
	       TK = VFMA(LDK(KP559016994), Tl, Tk);
	       Tm = VFNMS(LDK(KP559016994), Tl, Tk);
	       TO = VZMUL(TN, VFMAI(TF, TC));
	       TG = VZMUL(Tr, VFNMSI(TF, TC));
	       T12 = VZMUL(T11, VFMAI(TV, TU));
	       TW = VZMUL(TT, VFNMSI(TV, TU));
	       T16 = VZMULI(T15, VFMAI(TL, TK));
	       TM = VZMULI(TJ, VFNMSI(TL, TK));
	       T10 = VZMULI(TZ, VFNMSI(Tp, Tm));
	       Tq = VZMULI(T1, VFMAI(Tp, Tm));
	       TX = VADD(TS, TW);
	       TY = VCONJ(VSUB(TW, TS));
	       T18 = VADD(T16, T17);
	       T19 = VCONJ(VSUB(T17, T16));
	       TQ = VCONJ(VSUB(TO, TM));
	       TP = VADD(TM, TO);
	       T13 = VADD(T10, T12);
	       T14 = VCONJ(VSUB(T12, T10));
	       TI = VCONJ(VSUB(TG, Tq));
	       TH = VADD(Tq, TG);
	       ST(&(Rp[WS(rs, 2)]), TX, ms, &(Rp[0]));
	       ST(&(Rm[WS(rs, 2)]), TY, -ms, &(Rm[0]));
	       ST(&(Rp[0]), T18, ms, &(Rp[0]));
	       ST(&(Rm[0]), T19, -ms, &(Rm[0]));
	       ST(&(Rm[WS(rs, 4)]), TQ, -ms, &(Rm[0]));
	       ST(&(Rp[WS(rs, 4)]), TP, ms, &(Rp[0]));
	       ST(&(Rp[WS(rs, 3)]), T13, ms, &(Rp[WS(rs, 1)]));
	       ST(&(Rm[WS(rs, 3)]), T14, -ms, &(Rm[WS(rs, 1)]));
	       ST(&(Rm[WS(rs, 1)]), TI, -ms, &(Rm[WS(rs, 1)]));
	       ST(&(Rp[WS(rs, 1)]), TH, ms, &(Rp[WS(rs, 1)]));
	  }
     }
}

static const tw_instr twinstr[] = {
     VTW(1, 1),
     VTW(1, 2),
     VTW(1, 3),
     VTW(1, 4),
     VTW(1, 5),
     VTW(1, 6),
     VTW(1, 7),
     VTW(1, 8),
     VTW(1, 9),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 10, "hc2cbdftv_10", twinstr, &GENUS, {33, 22, 28, 0} };

void X(codelet_hc2cbdftv_10) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_10, &desc, HC2C_VIA_DFT);
}
#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_hc2cdft_c -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 10 -dif -sign 1 -name hc2cbdftv_10 -include hc2cbv.h */

/*
 * This function contains 61 FP additions, 30 FP multiplications,
 * (or, 55 additions, 24 multiplications, 6 fused multiply/add),
 * 81 stack variables, 4 constants, and 20 memory accesses
 */
#include "hc2cbv.h"

static void hc2cbdftv_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     INT m;
     for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 18)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(rs)) {
	  V T5, TE, Ts, Tt, TC, Tz, TH, TJ, To, Tq, T2, T4, T3, T9, Tx;
	  V Tm, TB, Td, Ty, Ti, TA, T6, T8, T7, Tl, Tk, Tj, Tc, Tb, Ta;
	  V Tf, Th, Tg, TF, TG, Te, Tn;
	  T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	  T3 = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
	  T4 = VCONJ(T3);
	  T5 = VSUB(T2, T4);
	  TE = VADD(T2, T4);
	  T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
	  T7 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
	  T8 = VCONJ(T7);
	  T9 = VSUB(T6, T8);
	  Tx = VADD(T6, T8);
	  Tl = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	  Tj = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
	  Tk = VCONJ(Tj);
	  Tm = VSUB(Tk, Tl);
	  TB = VADD(Tk, Tl);
	  Tc = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
	  Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	  Tb = VCONJ(Ta);
	  Td = VSUB(Tb, Tc);
	  Ty = VADD(Tb, Tc);
	  Tf = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
	  Tg = LD(&(Rm[0]), -ms, &(Rm[0]));
	  Th = VCONJ(Tg);
	  Ti = VSUB(Tf, Th);
	  TA = VADD(Tf, Th);
	  Ts = VSUB(T9, Td);
	  Tt = VSUB(Ti, Tm);
	  TC = VSUB(TA, TB);
	  Tz = VSUB(Tx, Ty);
	  TF = VADD(Tx, Ty);
	  TG = VADD(TA, TB);
	  TH = VADD(TF, TG);
	  TJ = VMUL(LDK(KP559016994), VSUB(TF, TG));
	  Te = VADD(T9, Td);
	  Tn = VADD(Ti, Tm);
	  To = VADD(Te, Tn);
	  Tq = VMUL(LDK(KP559016994), VSUB(Te, Tn));
	  {
	       V T1c, TX, Tv, T1b, TR, T15, TL, T17, TT, T11, TW, Tu, TQ, Tr, TP;
	       V Tp, T1, T1a, TO, T14, TD, T10, TK, TZ, TI, Tw, T16, TS, TY, TM;
	       V TU, T1e, TN, T1d, T19, T13, TV, T18, T12;
	       T1c = VADD(TE, TH);
	       TW = LDW(&(W[TWVL * 8]));
	       TX = VZMULI(TW, VADD(T5, To));
	       Tu = VBYI(VFNMS(LDK(KP951056516), Tt, VMUL(LDK(KP587785252), Ts)));
	       TQ = VBYI(VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tt)));
	       Tp = VFNMS(LDK(KP250000000), To, T5);
	       Tr = VSUB(Tp, Tq);
	       TP = VADD(Tq, Tp);
	       T1 = LDW(&(W[TWVL * 4]));
	       Tv = VZMULI(T1, VSUB(Tr, Tu));
	       T1a = LDW(&(W[0]));
	       T1b = VZMULI(T1a, VADD(TQ, TP));
	       TO = LDW(&(W[TWVL * 16]));
	       TR = VZMULI(TO, VSUB(TP, TQ));
	       T14 = LDW(&(W[TWVL * 12]));
	       T15 = VZMULI(T14, VADD(Tu, Tr));
	       TD = VBYI(VFNMS(LDK(KP951056516), TC, VMUL(LDK(KP587785252), Tz)));
	       T10 = VBYI(VFMA(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), TC)));
	       TI = VFNMS(LDK(KP250000000), TH, TE);
	       TK = VSUB(TI, TJ);
	       TZ = VADD(TJ, TI);
	       Tw = LDW(&(W[TWVL * 2]));
	       TL = VZMUL(Tw, VADD(TD, TK));
	       T16 = LDW(&(W[TWVL * 10]));
	       T17 = VZMUL(T16, VADD(T10, TZ));
	       TS = LDW(&(W[TWVL * 14]));
	       TT = VZMUL(TS, VSUB(TK, TD));
	       TY = LDW(&(W[TWVL * 6]));
	       T11 = VZMUL(TY, VSUB(TZ, T10));
	       TM = VADD(Tv, TL);
	       ST(&(Rp[WS(rs, 1)]), TM, ms, &(Rp[WS(rs, 1)]));
	       TU = VADD(TR, TT);
	       ST(&(Rp[WS(rs, 4)]), TU, ms, &(Rp[0]));
	       T1e = VCONJ(VSUB(T1c, T1b));
	       ST(&(Rm[0]), T1e, -ms, &(Rm[0]));
	       TN = VCONJ(VSUB(TL, Tv));
	       ST(&(Rm[WS(rs, 1)]), TN, -ms, &(Rm[WS(rs, 1)]));
	       T1d = VADD(T1b, T1c);
	       ST(&(Rp[0]), T1d, ms, &(Rp[0]));
	       T19 = VCONJ(VSUB(T17, T15));
	       ST(&(Rm[WS(rs, 3)]), T19, -ms, &(Rm[WS(rs, 1)]));
	       T13 = VCONJ(VSUB(T11, TX));
	       ST(&(Rm[WS(rs, 2)]), T13, -ms, &(Rm[0]));
	       TV = VCONJ(VSUB(TT, TR));
	       ST(&(Rm[WS(rs, 4)]), TV, -ms, &(Rm[0]));
	       T18 = VADD(T15, T17);
	       ST(&(Rp[WS(rs, 3)]), T18, ms, &(Rp[WS(rs, 1)]));
	       T12 = VADD(TX, T11);
	       ST(&(Rp[WS(rs, 2)]), T12, ms, &(Rp[0]));
	  }
     }
}

static const tw_instr twinstr[] = {
     VTW(1, 1),
     VTW(1, 2),
     VTW(1, 3),
     VTW(1, 4),
     VTW(1, 5),
     VTW(1, 6),
     VTW(1, 7),
     VTW(1, 8),
     VTW(1, 9),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 10, "hc2cbdftv_10", twinstr, &GENUS, {55, 24, 6, 0} };

void X(codelet_hc2cbdftv_10) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_10, &desc, HC2C_VIA_DFT);
}
#endif				/* HAVE_FMA */