488 lines
14 KiB
C
488 lines
14 KiB
C
/*
|
|
* 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 Tue Sep 14 10:46:13 EDT 2021 */
|
|
|
|
#include "rdft/codelet-rdft.h"
|
|
|
|
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include rdft/scalar/hf.h */
|
|
|
|
/*
|
|
* This function contains 96 FP additions, 88 FP multiplications,
|
|
* (or, 24 additions, 16 multiplications, 72 fused multiply/add),
|
|
* 55 stack variables, 10 constants, and 36 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hf.h"
|
|
|
|
static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DK(KP852868531, +0.852868531952443209628250963940074071936020296);
|
|
DK(KP492403876, +0.492403876506104029683371512294761506835321626);
|
|
DK(KP984807753, +0.984807753012208059366743024589523013670643252);
|
|
DK(KP777861913, +0.777861913430206160028177977318626690410586096);
|
|
DK(KP839099631, +0.839099631177280011763127298123181364687434283);
|
|
DK(KP954188894, +0.954188894138671133499268364187245676532219158);
|
|
DK(KP363970234, +0.363970234266202361351047882776834043890471784);
|
|
DK(KP176326980, +0.176326980708464973471090386868618986121633062);
|
|
DK(KP866025403, +0.866025403784438646763723170752936183471402627);
|
|
DK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
|
|
E T1, T1P, Te, T1S, T10, T1Q, T1a, T1d, Ty, T18, Tl, T13, T19, T1c, T1l;
|
|
E T1r, TS, T1p, TF, T1o, T1g, T1q;
|
|
T1 = cr[0];
|
|
T1P = ci[0];
|
|
{
|
|
E T3, T6, T4, TW, T9, Tc, Ta, TY, T2, T8;
|
|
T3 = cr[WS(rs, 3)];
|
|
T6 = ci[WS(rs, 3)];
|
|
T2 = W[4];
|
|
T4 = T2 * T3;
|
|
TW = T2 * T6;
|
|
T9 = cr[WS(rs, 6)];
|
|
Tc = ci[WS(rs, 6)];
|
|
T8 = W[10];
|
|
Ta = T8 * T9;
|
|
TY = T8 * Tc;
|
|
{
|
|
E T7, TX, Td, TZ, T5, Tb;
|
|
T5 = W[5];
|
|
T7 = FMA(T5, T6, T4);
|
|
TX = FNMS(T5, T3, TW);
|
|
Tb = W[11];
|
|
Td = FMA(Tb, Tc, Ta);
|
|
TZ = FNMS(Tb, T9, TY);
|
|
Te = T7 + Td;
|
|
T1S = Td - T7;
|
|
T10 = TX - TZ;
|
|
T1Q = TX + TZ;
|
|
}
|
|
}
|
|
{
|
|
E Th, Tk, Ti, T12, Tx, T17, Tr, T15, Tg, Tj;
|
|
Th = cr[WS(rs, 1)];
|
|
Tk = ci[WS(rs, 1)];
|
|
Tg = W[0];
|
|
Ti = Tg * Th;
|
|
T12 = Tg * Tk;
|
|
{
|
|
E Tt, Tw, Tu, T16, Ts, Tv;
|
|
Tt = cr[WS(rs, 7)];
|
|
Tw = ci[WS(rs, 7)];
|
|
Ts = W[12];
|
|
Tu = Ts * Tt;
|
|
T16 = Ts * Tw;
|
|
Tv = W[13];
|
|
Tx = FMA(Tv, Tw, Tu);
|
|
T17 = FNMS(Tv, Tt, T16);
|
|
}
|
|
{
|
|
E Tn, Tq, To, T14, Tm, Tp;
|
|
Tn = cr[WS(rs, 4)];
|
|
Tq = ci[WS(rs, 4)];
|
|
Tm = W[6];
|
|
To = Tm * Tn;
|
|
T14 = Tm * Tq;
|
|
Tp = W[7];
|
|
Tr = FMA(Tp, Tq, To);
|
|
T15 = FNMS(Tp, Tn, T14);
|
|
}
|
|
T1a = Tr - Tx;
|
|
T1d = T15 - T17;
|
|
Ty = Tr + Tx;
|
|
T18 = T15 + T17;
|
|
Tj = W[1];
|
|
Tl = FMA(Tj, Tk, Ti);
|
|
T13 = FNMS(Tj, Th, T12);
|
|
T19 = FNMS(KP500000000, T18, T13);
|
|
T1c = FNMS(KP500000000, Ty, Tl);
|
|
}
|
|
{
|
|
E TB, TE, TC, T1n, TR, T1k, TL, T1i, TA, TD;
|
|
TB = cr[WS(rs, 2)];
|
|
TE = ci[WS(rs, 2)];
|
|
TA = W[2];
|
|
TC = TA * TB;
|
|
T1n = TA * TE;
|
|
{
|
|
E TN, TQ, TO, T1j, TM, TP;
|
|
TN = cr[WS(rs, 8)];
|
|
TQ = ci[WS(rs, 8)];
|
|
TM = W[14];
|
|
TO = TM * TN;
|
|
T1j = TM * TQ;
|
|
TP = W[15];
|
|
TR = FMA(TP, TQ, TO);
|
|
T1k = FNMS(TP, TN, T1j);
|
|
}
|
|
{
|
|
E TH, TK, TI, T1h, TG, TJ;
|
|
TH = cr[WS(rs, 5)];
|
|
TK = ci[WS(rs, 5)];
|
|
TG = W[8];
|
|
TI = TG * TH;
|
|
T1h = TG * TK;
|
|
TJ = W[9];
|
|
TL = FMA(TJ, TK, TI);
|
|
T1i = FNMS(TJ, TH, T1h);
|
|
}
|
|
T1l = T1i - T1k;
|
|
T1r = TR - TL;
|
|
TS = TL + TR;
|
|
T1p = T1i + T1k;
|
|
TD = W[3];
|
|
TF = FMA(TD, TE, TC);
|
|
T1o = FNMS(TD, TB, T1n);
|
|
T1g = FNMS(KP500000000, TS, TF);
|
|
T1q = FNMS(KP500000000, T1p, T1o);
|
|
}
|
|
{
|
|
E Tf, T21, TU, T24, T1O, T22, T1L, T23;
|
|
Tf = T1 + Te;
|
|
T21 = T1Q + T1P;
|
|
{
|
|
E Tz, TT, T1M, T1N;
|
|
Tz = Tl + Ty;
|
|
TT = TF + TS;
|
|
TU = Tz + TT;
|
|
T24 = TT - Tz;
|
|
T1M = T13 + T18;
|
|
T1N = T1o + T1p;
|
|
T1O = T1M - T1N;
|
|
T22 = T1M + T1N;
|
|
}
|
|
cr[0] = Tf + TU;
|
|
ci[WS(rs, 8)] = T22 + T21;
|
|
T1L = FNMS(KP500000000, TU, Tf);
|
|
ci[WS(rs, 2)] = FNMS(KP866025403, T1O, T1L);
|
|
cr[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
|
|
T23 = FNMS(KP500000000, T22, T21);
|
|
cr[WS(rs, 6)] = FMS(KP866025403, T24, T23);
|
|
ci[WS(rs, 5)] = FMA(KP866025403, T24, T23);
|
|
}
|
|
{
|
|
E T11, T1z, T1T, T1X, T1f, T1w, T1t, T1x, T1u, T1Y, T1C, T1I, T1F, T1J, T1G;
|
|
E T1U, TV, T1R;
|
|
TV = FNMS(KP500000000, Te, T1);
|
|
T11 = FNMS(KP866025403, T10, TV);
|
|
T1z = FMA(KP866025403, T10, TV);
|
|
T1R = FNMS(KP500000000, T1Q, T1P);
|
|
T1T = FMA(KP866025403, T1S, T1R);
|
|
T1X = FNMS(KP866025403, T1S, T1R);
|
|
{
|
|
E T1b, T1e, T1m, T1s;
|
|
T1b = FMA(KP866025403, T1a, T19);
|
|
T1e = FNMS(KP866025403, T1d, T1c);
|
|
T1f = FMA(KP176326980, T1e, T1b);
|
|
T1w = FNMS(KP176326980, T1b, T1e);
|
|
T1m = FNMS(KP866025403, T1l, T1g);
|
|
T1s = FNMS(KP866025403, T1r, T1q);
|
|
T1t = FNMS(KP363970234, T1s, T1m);
|
|
T1x = FMA(KP363970234, T1m, T1s);
|
|
}
|
|
T1u = FNMS(KP954188894, T1t, T1f);
|
|
T1Y = FMA(KP954188894, T1x, T1w);
|
|
{
|
|
E T1A, T1B, T1D, T1E;
|
|
T1A = FMA(KP866025403, T1r, T1q);
|
|
T1B = FMA(KP866025403, T1l, T1g);
|
|
T1C = FMA(KP176326980, T1B, T1A);
|
|
T1I = FNMS(KP176326980, T1A, T1B);
|
|
T1D = FMA(KP866025403, T1d, T1c);
|
|
T1E = FNMS(KP866025403, T1a, T19);
|
|
T1F = FMA(KP839099631, T1E, T1D);
|
|
T1J = FNMS(KP839099631, T1D, T1E);
|
|
}
|
|
T1G = FMA(KP777861913, T1F, T1C);
|
|
T1U = FNMS(KP777861913, T1J, T1I);
|
|
cr[WS(rs, 2)] = FMA(KP984807753, T1u, T11);
|
|
ci[WS(rs, 7)] = FNMS(KP984807753, T1U, T1T);
|
|
ci[WS(rs, 6)] = FNMS(KP984807753, T1Y, T1X);
|
|
cr[WS(rs, 1)] = FMA(KP984807753, T1G, T1z);
|
|
{
|
|
E T1V, T1W, T1H, T1K;
|
|
T1V = FMA(KP492403876, T1U, T1T);
|
|
T1W = FNMS(KP777861913, T1F, T1C);
|
|
cr[WS(rs, 7)] = FMS(KP852868531, T1W, T1V);
|
|
ci[WS(rs, 4)] = FMA(KP852868531, T1W, T1V);
|
|
T1H = FNMS(KP492403876, T1G, T1z);
|
|
T1K = FMA(KP777861913, T1J, T1I);
|
|
ci[WS(rs, 1)] = FNMS(KP852868531, T1K, T1H);
|
|
cr[WS(rs, 4)] = FMA(KP852868531, T1K, T1H);
|
|
}
|
|
{
|
|
E T1v, T1y, T1Z, T20;
|
|
T1v = FNMS(KP492403876, T1u, T11);
|
|
T1y = FNMS(KP954188894, T1x, T1w);
|
|
ci[WS(rs, 3)] = FNMS(KP852868531, T1y, T1v);
|
|
ci[0] = FMA(KP852868531, T1y, T1v);
|
|
T1Z = FMA(KP492403876, T1Y, T1X);
|
|
T20 = FMA(KP954188894, T1t, T1f);
|
|
cr[WS(rs, 5)] = FMS(KP852868531, T20, T1Z);
|
|
cr[WS(rs, 8)] = -(FMA(KP852868531, T20, T1Z));
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_FULL, 1, 9 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, { 24, 16, 72, 0 } };
|
|
|
|
void X(codelet_hf_9) (planner *p) {
|
|
X(khc2hc_register) (p, hf_9, &desc);
|
|
}
|
|
#else
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include rdft/scalar/hf.h */
|
|
|
|
/*
|
|
* This function contains 96 FP additions, 72 FP multiplications,
|
|
* (or, 60 additions, 36 multiplications, 36 fused multiply/add),
|
|
* 41 stack variables, 8 constants, and 36 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hf.h"
|
|
|
|
static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DK(KP642787609, +0.642787609686539326322643409907263432907559884);
|
|
DK(KP766044443, +0.766044443118978035202392650555416673935832457);
|
|
DK(KP939692620, +0.939692620785908384054109277324731469936208134);
|
|
DK(KP342020143, +0.342020143325668733044099614682259580763083368);
|
|
DK(KP984807753, +0.984807753012208059366743024589523013670643252);
|
|
DK(KP173648177, +0.173648177666930348851716626769314796000375677);
|
|
DK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
|
DK(KP866025403, +0.866025403784438646763723170752936183471402627);
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
|
|
E T1, T1B, TQ, T1A, Tc, TN, T1C, T1D, TL, T1x, T19, T1o, T1c, T1n, Tu;
|
|
E T1w, TW, T1k, T11, T1l;
|
|
{
|
|
E T6, TO, Tb, TP;
|
|
T1 = cr[0];
|
|
T1B = ci[0];
|
|
{
|
|
E T3, T5, T2, T4;
|
|
T3 = cr[WS(rs, 3)];
|
|
T5 = ci[WS(rs, 3)];
|
|
T2 = W[4];
|
|
T4 = W[5];
|
|
T6 = FMA(T2, T3, T4 * T5);
|
|
TO = FNMS(T4, T3, T2 * T5);
|
|
}
|
|
{
|
|
E T8, Ta, T7, T9;
|
|
T8 = cr[WS(rs, 6)];
|
|
Ta = ci[WS(rs, 6)];
|
|
T7 = W[10];
|
|
T9 = W[11];
|
|
Tb = FMA(T7, T8, T9 * Ta);
|
|
TP = FNMS(T9, T8, T7 * Ta);
|
|
}
|
|
TQ = KP866025403 * (TO - TP);
|
|
T1A = KP866025403 * (Tb - T6);
|
|
Tc = T6 + Tb;
|
|
TN = FNMS(KP500000000, Tc, T1);
|
|
T1C = TO + TP;
|
|
T1D = FNMS(KP500000000, T1C, T1B);
|
|
}
|
|
{
|
|
E Tz, T13, TE, T14, TJ, T15, TK, T16;
|
|
{
|
|
E Tw, Ty, Tv, Tx;
|
|
Tw = cr[WS(rs, 2)];
|
|
Ty = ci[WS(rs, 2)];
|
|
Tv = W[2];
|
|
Tx = W[3];
|
|
Tz = FMA(Tv, Tw, Tx * Ty);
|
|
T13 = FNMS(Tx, Tw, Tv * Ty);
|
|
}
|
|
{
|
|
E TB, TD, TA, TC;
|
|
TB = cr[WS(rs, 5)];
|
|
TD = ci[WS(rs, 5)];
|
|
TA = W[8];
|
|
TC = W[9];
|
|
TE = FMA(TA, TB, TC * TD);
|
|
T14 = FNMS(TC, TB, TA * TD);
|
|
}
|
|
{
|
|
E TG, TI, TF, TH;
|
|
TG = cr[WS(rs, 8)];
|
|
TI = ci[WS(rs, 8)];
|
|
TF = W[14];
|
|
TH = W[15];
|
|
TJ = FMA(TF, TG, TH * TI);
|
|
T15 = FNMS(TH, TG, TF * TI);
|
|
}
|
|
TK = TE + TJ;
|
|
T16 = T14 + T15;
|
|
TL = Tz + TK;
|
|
T1x = T13 + T16;
|
|
{
|
|
E T17, T18, T1a, T1b;
|
|
T17 = FNMS(KP500000000, T16, T13);
|
|
T18 = KP866025403 * (TJ - TE);
|
|
T19 = T17 - T18;
|
|
T1o = T18 + T17;
|
|
T1a = FNMS(KP500000000, TK, Tz);
|
|
T1b = KP866025403 * (T14 - T15);
|
|
T1c = T1a - T1b;
|
|
T1n = T1a + T1b;
|
|
}
|
|
}
|
|
{
|
|
E Ti, TX, Tn, TT, Ts, TU, Tt, TY;
|
|
{
|
|
E Tf, Th, Te, Tg;
|
|
Tf = cr[WS(rs, 1)];
|
|
Th = ci[WS(rs, 1)];
|
|
Te = W[0];
|
|
Tg = W[1];
|
|
Ti = FMA(Te, Tf, Tg * Th);
|
|
TX = FNMS(Tg, Tf, Te * Th);
|
|
}
|
|
{
|
|
E Tk, Tm, Tj, Tl;
|
|
Tk = cr[WS(rs, 4)];
|
|
Tm = ci[WS(rs, 4)];
|
|
Tj = W[6];
|
|
Tl = W[7];
|
|
Tn = FMA(Tj, Tk, Tl * Tm);
|
|
TT = FNMS(Tl, Tk, Tj * Tm);
|
|
}
|
|
{
|
|
E Tp, Tr, To, Tq;
|
|
Tp = cr[WS(rs, 7)];
|
|
Tr = ci[WS(rs, 7)];
|
|
To = W[12];
|
|
Tq = W[13];
|
|
Ts = FMA(To, Tp, Tq * Tr);
|
|
TU = FNMS(Tq, Tp, To * Tr);
|
|
}
|
|
Tt = Tn + Ts;
|
|
TY = TT + TU;
|
|
Tu = Ti + Tt;
|
|
T1w = TX + TY;
|
|
{
|
|
E TS, TV, TZ, T10;
|
|
TS = FNMS(KP500000000, Tt, Ti);
|
|
TV = KP866025403 * (TT - TU);
|
|
TW = TS - TV;
|
|
T1k = TS + TV;
|
|
TZ = FNMS(KP500000000, TY, TX);
|
|
T10 = KP866025403 * (Ts - Tn);
|
|
T11 = TZ - T10;
|
|
T1l = T10 + TZ;
|
|
}
|
|
}
|
|
{
|
|
E T1y, Td, TM, T1v;
|
|
T1y = KP866025403 * (T1w - T1x);
|
|
Td = T1 + Tc;
|
|
TM = Tu + TL;
|
|
T1v = FNMS(KP500000000, TM, Td);
|
|
cr[0] = Td + TM;
|
|
cr[WS(rs, 3)] = T1v + T1y;
|
|
ci[WS(rs, 2)] = T1v - T1y;
|
|
}
|
|
{
|
|
E TR, T1I, T1e, T1K, T1i, T1H, T1f, T1J;
|
|
TR = TN - TQ;
|
|
T1I = T1D - T1A;
|
|
{
|
|
E T12, T1d, T1g, T1h;
|
|
T12 = FMA(KP173648177, TW, KP984807753 * T11);
|
|
T1d = FNMS(KP939692620, T1c, KP342020143 * T19);
|
|
T1e = T12 + T1d;
|
|
T1K = KP866025403 * (T1d - T12);
|
|
T1g = FNMS(KP984807753, TW, KP173648177 * T11);
|
|
T1h = FMA(KP342020143, T1c, KP939692620 * T19);
|
|
T1i = KP866025403 * (T1g + T1h);
|
|
T1H = T1g - T1h;
|
|
}
|
|
cr[WS(rs, 2)] = TR + T1e;
|
|
ci[WS(rs, 6)] = T1H + T1I;
|
|
T1f = FNMS(KP500000000, T1e, TR);
|
|
ci[0] = T1f - T1i;
|
|
ci[WS(rs, 3)] = T1f + T1i;
|
|
T1J = FMS(KP500000000, T1H, T1I);
|
|
cr[WS(rs, 5)] = T1J - T1K;
|
|
cr[WS(rs, 8)] = T1K + T1J;
|
|
}
|
|
{
|
|
E T1L, T1M, T1N, T1O;
|
|
T1L = KP866025403 * (TL - Tu);
|
|
T1M = T1C + T1B;
|
|
T1N = T1w + T1x;
|
|
T1O = FNMS(KP500000000, T1N, T1M);
|
|
cr[WS(rs, 6)] = T1L - T1O;
|
|
ci[WS(rs, 8)] = T1N + T1M;
|
|
ci[WS(rs, 5)] = T1L + T1O;
|
|
}
|
|
{
|
|
E T1j, T1E, T1q, T1z, T1u, T1F, T1r, T1G;
|
|
T1j = TN + TQ;
|
|
T1E = T1A + T1D;
|
|
{
|
|
E T1m, T1p, T1s, T1t;
|
|
T1m = FMA(KP766044443, T1k, KP642787609 * T1l);
|
|
T1p = FMA(KP173648177, T1n, KP984807753 * T1o);
|
|
T1q = T1m + T1p;
|
|
T1z = KP866025403 * (T1p - T1m);
|
|
T1s = FNMS(KP642787609, T1k, KP766044443 * T1l);
|
|
T1t = FNMS(KP984807753, T1n, KP173648177 * T1o);
|
|
T1u = KP866025403 * (T1s - T1t);
|
|
T1F = T1s + T1t;
|
|
}
|
|
cr[WS(rs, 1)] = T1j + T1q;
|
|
T1r = FNMS(KP500000000, T1q, T1j);
|
|
ci[WS(rs, 1)] = T1r - T1u;
|
|
cr[WS(rs, 4)] = T1r + T1u;
|
|
ci[WS(rs, 7)] = T1F + T1E;
|
|
T1G = FNMS(KP500000000, T1F, T1E);
|
|
cr[WS(rs, 7)] = T1z - T1G;
|
|
ci[WS(rs, 4)] = T1z + T1G;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_FULL, 1, 9 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, { 60, 36, 36, 0 } };
|
|
|
|
void X(codelet_hf_9) (planner *p) {
|
|
X(khc2hc_register) (p, hf_9, &desc);
|
|
}
|
|
#endif
|