annotate src/fftw-3.3.8/rdft/simd/common/hc2cbdftv_8.c @ 82:d0c2a83c1364

Add FFTW 3.3.8 source, and a Linux build

/*
 * Copyright (c) 2003, 2007-14 Matteo Frigo
 * Copyright (c) 2003, 2007-14 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., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
 *
 */

/* This file was automatically generated --- DO NOT EDIT */
/* Generated on Thu May 24 08:08:11 EDT 2018 */

#include "rdft/codelet-rdft.h"

#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)

/* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 8 -dif -sign 1 -name hc2cbdftv_8 -include rdft/simd/hc2cbv.h */

/*
 * This function contains 41 FP additions, 32 FP multiplications,
 * (or, 23 additions, 14 multiplications, 18 fused multiply/add),
 * 51 stack variables, 1 constants, and 16 memory accesses
 */
#include "rdft/simd/hc2cbv.h"

static void hc2cbdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
     {
	  INT m;
	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
	       V Tm, Tp, TF, TE, Th, Tv, Tc, Tu, T4, Tk, Tf, Tl, T7, Tn, Ta;
	       V To, T2, T3, Td, Te, T5, T6, T8, T9, Tg, Tb, TL, TK, TJ, TM;
	       V TN, TC, TG, TB, TD, TH, TI, Ti, Tq, T1, Tj, Tr, Ts, Tw, Ty;
	       V Tt, Tx, Tz, TA;
	       T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	       T3 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
	       T4 = VFNMSCONJ(T3, T2);
	       Tk = VFMACONJ(T3, T2);
	       Td = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
	       Te = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	       Tf = VFNMSCONJ(Te, Td);
	       Tl = VFMACONJ(Te, Td);
	       T5 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	       T6 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
	       T7 = VFNMSCONJ(T6, T5);
	       Tn = VFMACONJ(T6, T5);
	       T8 = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
	       T9 = LD(&(Rm[0]), -ms, &(Rm[0]));
	       Ta = VFMSCONJ(T9, T8);
	       To = VFMACONJ(T9, T8);
	       Tm = VSUB(Tk, Tl);
	       Tp = VSUB(Tn, To);
	       TF = VADD(Tn, To);
	       TE = VADD(Tk, Tl);
	       Tg = VSUB(T7, Ta);
	       Th = VFMA(LDK(KP707106781), Tg, Tf);
	       Tv = VFNMS(LDK(KP707106781), Tg, Tf);
	       Tb = VADD(T7, Ta);
	       Tc = VFMA(LDK(KP707106781), Tb, T4);
	       Tu = VFNMS(LDK(KP707106781), Tb, T4);
	       TL = VADD(TE, TF);
	       TJ = LDW(&(W[0]));
	       TK = VZMULI(TJ, VFMAI(Th, Tc));
	       TM = VADD(TK, TL);
	       ST(&(Rp[0]), TM, ms, &(Rp[0]));
	       TN = VCONJ(VSUB(TL, TK));
	       ST(&(Rm[0]), TN, -ms, &(Rm[0]));
	       TB = LDW(&(W[TWVL * 8]));
	       TC = VZMULI(TB, VFMAI(Tv, Tu));
	       TD = LDW(&(W[TWVL * 6]));
	       TG = VZMUL(TD, VSUB(TE, TF));
	       TH = VADD(TC, TG);
	       ST(&(Rp[WS(rs, 2)]), TH, ms, &(Rp[0]));
	       TI = VCONJ(VSUB(TG, TC));
	       ST(&(Rm[WS(rs, 2)]), TI, -ms, &(Rm[0]));
	       T1 = LDW(&(W[TWVL * 12]));
	       Ti = VZMULI(T1, VFNMSI(Th, Tc));
	       Tj = LDW(&(W[TWVL * 10]));
	       Tq = VZMUL(Tj, VFNMSI(Tp, Tm));
	       Tr = VADD(Ti, Tq);
	       ST(&(Rp[WS(rs, 3)]), Tr, ms, &(Rp[WS(rs, 1)]));
	       Ts = VCONJ(VSUB(Tq, Ti));
	       ST(&(Rm[WS(rs, 3)]), Ts, -ms, &(Rm[WS(rs, 1)]));
	       Tt = LDW(&(W[TWVL * 4]));
	       Tw = VZMULI(Tt, VFNMSI(Tv, Tu));
	       Tx = LDW(&(W[TWVL * 2]));
	       Ty = VZMUL(Tx, VFMAI(Tp, Tm));
	       Tz = VADD(Tw, Ty);
	       ST(&(Rp[WS(rs, 1)]), Tz, ms, &(Rp[WS(rs, 1)]));
	       TA = VCONJ(VSUB(Ty, Tw));
	       ST(&(Rm[WS(rs, 1)]), TA, -ms, &(Rm[WS(rs, 1)]));
	  }
     }
     VLEAVE();
}

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),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cbdftv_8"), twinstr, &GENUS, {23, 14, 18, 0} };

void XSIMD(codelet_hc2cbdftv_8) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_8, &desc, HC2C_VIA_DFT);
}
#else

/* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 8 -dif -sign 1 -name hc2cbdftv_8 -include rdft/simd/hc2cbv.h */

/*
 * This function contains 41 FP additions, 16 FP multiplications,
 * (or, 41 additions, 16 multiplications, 0 fused multiply/add),
 * 55 stack variables, 1 constants, and 16 memory accesses
 */
#include "rdft/simd/hc2cbv.h"

static void hc2cbdftv_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
     {
	  INT m;
	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 14)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(32, rs)) {
	       V T5, Tj, Tq, TI, Te, Tk, Tt, TJ, T2, Tg, T4, Ti, T3, Th, To;
	       V Tp, T6, Tc, T8, Tb, T7, Ta, T9, Td, Tr, Ts, TP, Tu, Tm, TO;
	       V Tn, Tf, Tl, T1, TN, Tv, TR, Tw, TQ, TC, TK, TA, TG, TB, TH;
	       V Ty, Tz, Tx, TF, TD, TM, TE, TL;
	       T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	       Tg = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
	       T3 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
	       T4 = VCONJ(T3);
	       Th = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	       Ti = VCONJ(Th);
	       T5 = VSUB(T2, T4);
	       Tj = VSUB(Tg, Ti);
	       To = VADD(T2, T4);
	       Tp = VADD(Tg, Ti);
	       Tq = VSUB(To, Tp);
	       TI = VADD(To, Tp);
	       T6 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	       Tc = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
	       T7 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
	       T8 = VCONJ(T7);
	       Ta = LD(&(Rm[0]), -ms, &(Rm[0]));
	       Tb = VCONJ(Ta);
	       T9 = VSUB(T6, T8);
	       Td = VSUB(Tb, Tc);
	       Te = VMUL(LDK(KP707106781), VADD(T9, Td));
	       Tk = VMUL(LDK(KP707106781), VSUB(T9, Td));
	       Tr = VADD(T6, T8);
	       Ts = VADD(Tb, Tc);
	       Tt = VBYI(VSUB(Tr, Ts));
	       TJ = VADD(Tr, Ts);
	       TP = VADD(TI, TJ);
	       Tn = LDW(&(W[TWVL * 10]));
	       Tu = VZMUL(Tn, VSUB(Tq, Tt));
	       Tf = VADD(T5, Te);
	       Tl = VBYI(VADD(Tj, Tk));
	       T1 = LDW(&(W[TWVL * 12]));
	       Tm = VZMULI(T1, VSUB(Tf, Tl));
	       TN = LDW(&(W[0]));
	       TO = VZMULI(TN, VADD(Tl, Tf));
	       Tv = VADD(Tm, Tu);
	       ST(&(Rp[WS(rs, 3)]), Tv, ms, &(Rp[WS(rs, 1)]));
	       TR = VCONJ(VSUB(TP, TO));
	       ST(&(Rm[0]), TR, -ms, &(Rm[0]));
	       Tw = VCONJ(VSUB(Tu, Tm));
	       ST(&(Rm[WS(rs, 3)]), Tw, -ms, &(Rm[WS(rs, 1)]));
	       TQ = VADD(TO, TP);
	       ST(&(Rp[0]), TQ, ms, &(Rp[0]));
	       TB = LDW(&(W[TWVL * 2]));
	       TC = VZMUL(TB, VADD(Tq, Tt));
	       TH = LDW(&(W[TWVL * 6]));
	       TK = VZMUL(TH, VSUB(TI, TJ));
	       Ty = VBYI(VSUB(Tk, Tj));
	       Tz = VSUB(T5, Te);
	       Tx = LDW(&(W[TWVL * 4]));
	       TA = VZMULI(Tx, VADD(Ty, Tz));
	       TF = LDW(&(W[TWVL * 8]));
	       TG = VZMULI(TF, VSUB(Tz, Ty));
	       TD = VADD(TA, TC);
	       ST(&(Rp[WS(rs, 1)]), TD, ms, &(Rp[WS(rs, 1)]));
	       TM = VCONJ(VSUB(TK, TG));
	       ST(&(Rm[WS(rs, 2)]), TM, -ms, &(Rm[0]));
	       TE = VCONJ(VSUB(TC, TA));
	       ST(&(Rm[WS(rs, 1)]), TE, -ms, &(Rm[WS(rs, 1)]));
	       TL = VADD(TG, TK);
	       ST(&(Rp[WS(rs, 2)]), TL, ms, &(Rp[0]));
	  }
     }
     VLEAVE();
}

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),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 8, XSIMD_STRING("hc2cbdftv_8"), twinstr, &GENUS, {41, 16, 0, 0} };

void XSIMD(codelet_hc2cbdftv_8) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_8, &desc, HC2C_VIA_DFT);
}
#endif