cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:04:13 EDT 2018 */ cannam@167: cannam@167: #include "dft/codelet-dft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -n 7 -name t1_7 -include dft/scalar/t.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 72 FP additions, 66 FP multiplications, cannam@167: * (or, 18 additions, 12 multiplications, 54 fused multiply/add), cannam@167: * 37 stack variables, 6 constants, and 28 memory accesses cannam@167: */ cannam@167: #include "dft/scalar/t.h" cannam@167: cannam@167: static void t1_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@167: DK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@167: DK(KP801937735, +0.801937735804838252472204639014890102331838324); cannam@167: DK(KP554958132, +0.554958132087371191422194871006410481067288862); cannam@167: DK(KP692021471, +0.692021471630095869627814897002069140197260599); cannam@167: DK(KP356895867, +0.356895867892209443894399510021300583399127187); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + (mb * 12); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) { cannam@167: E T1, T1c, Te, T1h, TR, T19, Tr, T1g, TM, T1a, TE, T1i, TW, T1b; cannam@167: T1 = ri[0]; cannam@167: T1c = ii[0]; cannam@167: { cannam@167: E T3, T6, T4, TN, T9, Tc, Ta, TP, T2, T8; cannam@167: T3 = ri[WS(rs, 1)]; cannam@167: T6 = ii[WS(rs, 1)]; cannam@167: T2 = W[0]; cannam@167: T4 = T2 * T3; cannam@167: TN = T2 * T6; cannam@167: T9 = ri[WS(rs, 6)]; cannam@167: Tc = ii[WS(rs, 6)]; cannam@167: T8 = W[10]; cannam@167: Ta = T8 * T9; cannam@167: TP = T8 * Tc; cannam@167: { cannam@167: E T7, TO, Td, TQ, T5, Tb; cannam@167: T5 = W[1]; cannam@167: T7 = FMA(T5, T6, T4); cannam@167: TO = FNMS(T5, T3, TN); cannam@167: Tb = W[11]; cannam@167: Td = FMA(Tb, Tc, Ta); cannam@167: TQ = FNMS(Tb, T9, TP); cannam@167: Te = T7 + Td; cannam@167: T1h = Td - T7; cannam@167: TR = TO - TQ; cannam@167: T19 = TO + TQ; cannam@167: } cannam@167: } cannam@167: { cannam@167: E Tg, Tj, Th, TI, Tm, Tp, Tn, TK, Tf, Tl; cannam@167: Tg = ri[WS(rs, 2)]; cannam@167: Tj = ii[WS(rs, 2)]; cannam@167: Tf = W[2]; cannam@167: Th = Tf * Tg; cannam@167: TI = Tf * Tj; cannam@167: Tm = ri[WS(rs, 5)]; cannam@167: Tp = ii[WS(rs, 5)]; cannam@167: Tl = W[8]; cannam@167: Tn = Tl * Tm; cannam@167: TK = Tl * Tp; cannam@167: { cannam@167: E Tk, TJ, Tq, TL, Ti, To; cannam@167: Ti = W[3]; cannam@167: Tk = FMA(Ti, Tj, Th); cannam@167: TJ = FNMS(Ti, Tg, TI); cannam@167: To = W[9]; cannam@167: Tq = FMA(To, Tp, Tn); cannam@167: TL = FNMS(To, Tm, TK); cannam@167: Tr = Tk + Tq; cannam@167: T1g = Tq - Tk; cannam@167: TM = TJ - TL; cannam@167: T1a = TJ + TL; cannam@167: } cannam@167: } cannam@167: { cannam@167: E Tt, Tw, Tu, TS, Tz, TC, TA, TU, Ts, Ty; cannam@167: Tt = ri[WS(rs, 3)]; cannam@167: Tw = ii[WS(rs, 3)]; cannam@167: Ts = W[4]; cannam@167: Tu = Ts * Tt; cannam@167: TS = Ts * Tw; cannam@167: Tz = ri[WS(rs, 4)]; cannam@167: TC = ii[WS(rs, 4)]; cannam@167: Ty = W[6]; cannam@167: TA = Ty * Tz; cannam@167: TU = Ty * TC; cannam@167: { cannam@167: E Tx, TT, TD, TV, Tv, TB; cannam@167: Tv = W[5]; cannam@167: Tx = FMA(Tv, Tw, Tu); cannam@167: TT = FNMS(Tv, Tt, TS); cannam@167: TB = W[7]; cannam@167: TD = FMA(TB, TC, TA); cannam@167: TV = FNMS(TB, Tz, TU); cannam@167: TE = Tx + TD; cannam@167: T1i = TD - Tx; cannam@167: TW = TT - TV; cannam@167: T1b = TT + TV; cannam@167: } cannam@167: } cannam@167: ri[0] = T1 + Te + Tr + TE; cannam@167: ii[0] = T19 + T1a + T1b + T1c; cannam@167: { cannam@167: E TG, TY, TF, TX, TH; cannam@167: TF = FNMS(KP356895867, Tr, Te); cannam@167: TG = FNMS(KP692021471, TF, TE); cannam@167: TX = FMA(KP554958132, TW, TR); cannam@167: TY = FMA(KP801937735, TX, TM); cannam@167: TH = FNMS(KP900968867, TG, T1); cannam@167: ri[WS(rs, 6)] = FNMS(KP974927912, TY, TH); cannam@167: ri[WS(rs, 1)] = FMA(KP974927912, TY, TH); cannam@167: } cannam@167: { cannam@167: E T1e, T1k, T1d, T1j, T1f; cannam@167: T1d = FNMS(KP356895867, T1a, T19); cannam@167: T1e = FNMS(KP692021471, T1d, T1b); cannam@167: T1j = FMA(KP554958132, T1i, T1h); cannam@167: T1k = FMA(KP801937735, T1j, T1g); cannam@167: T1f = FNMS(KP900968867, T1e, T1c); cannam@167: ii[WS(rs, 1)] = FMA(KP974927912, T1k, T1f); cannam@167: ii[WS(rs, 6)] = FNMS(KP974927912, T1k, T1f); cannam@167: } cannam@167: { cannam@167: E T10, T13, TZ, T12, T11; cannam@167: TZ = FNMS(KP356895867, Te, TE); cannam@167: T10 = FNMS(KP692021471, TZ, Tr); cannam@167: T12 = FMA(KP554958132, TM, TW); cannam@167: T13 = FNMS(KP801937735, T12, TR); cannam@167: T11 = FNMS(KP900968867, T10, T1); cannam@167: ri[WS(rs, 5)] = FNMS(KP974927912, T13, T11); cannam@167: ri[WS(rs, 2)] = FMA(KP974927912, T13, T11); cannam@167: } cannam@167: { cannam@167: E T1m, T1p, T1l, T1o, T1n; cannam@167: T1l = FNMS(KP356895867, T19, T1b); cannam@167: T1m = FNMS(KP692021471, T1l, T1a); cannam@167: T1o = FMA(KP554958132, T1g, T1i); cannam@167: T1p = FNMS(KP801937735, T1o, T1h); cannam@167: T1n = FNMS(KP900968867, T1m, T1c); cannam@167: ii[WS(rs, 2)] = FMA(KP974927912, T1p, T1n); cannam@167: ii[WS(rs, 5)] = FNMS(KP974927912, T1p, T1n); cannam@167: } cannam@167: { cannam@167: E T15, T18, T14, T17, T16; cannam@167: T14 = FNMS(KP356895867, TE, Tr); cannam@167: T15 = FNMS(KP692021471, T14, Te); cannam@167: T17 = FNMS(KP554958132, TR, TM); cannam@167: T18 = FNMS(KP801937735, T17, TW); cannam@167: T16 = FNMS(KP900968867, T15, T1); cannam@167: ri[WS(rs, 4)] = FNMS(KP974927912, T18, T16); cannam@167: ri[WS(rs, 3)] = FMA(KP974927912, T18, T16); cannam@167: } cannam@167: { cannam@167: E T1r, T1u, T1q, T1t, T1s; cannam@167: T1q = FNMS(KP356895867, T1b, T1a); cannam@167: T1r = FNMS(KP692021471, T1q, T19); cannam@167: T1t = FNMS(KP554958132, T1h, T1g); cannam@167: T1u = FNMS(KP801937735, T1t, T1i); cannam@167: T1s = FNMS(KP900968867, T1r, T1c); cannam@167: ii[WS(rs, 3)] = FMA(KP974927912, T1u, T1s); cannam@167: ii[WS(rs, 4)] = FNMS(KP974927912, T1u, T1s); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 0, 7}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const ct_desc desc = { 7, "t1_7", twinstr, &GENUS, {18, 12, 54, 0}, 0, 0, 0 }; cannam@167: cannam@167: void X(codelet_t1_7) (planner *p) { cannam@167: X(kdft_dit_register) (p, t1_7, &desc); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 7 -name t1_7 -include dft/scalar/t.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 72 FP additions, 60 FP multiplications, cannam@167: * (or, 36 additions, 24 multiplications, 36 fused multiply/add), cannam@167: * 29 stack variables, 6 constants, and 28 memory accesses cannam@167: */ cannam@167: #include "dft/scalar/t.h" cannam@167: cannam@167: static void t1_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP222520933, +0.222520933956314404288902564496794759466355569); cannam@167: DK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@167: DK(KP623489801, +0.623489801858733530525004884004239810632274731); cannam@167: DK(KP433883739, +0.433883739117558120475768332848358754609990728); cannam@167: DK(KP781831482, +0.781831482468029808708444526674057750232334519); cannam@167: DK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + (mb * 12); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) { cannam@167: E T1, TR, Tc, TS, TC, TO, Tn, TT, TI, TP, Ty, TU, TF, TQ; cannam@167: T1 = ri[0]; cannam@167: TR = ii[0]; cannam@167: { cannam@167: E T6, TA, Tb, TB; cannam@167: { cannam@167: E T3, T5, T2, T4; cannam@167: T3 = ri[WS(rs, 1)]; cannam@167: T5 = ii[WS(rs, 1)]; cannam@167: T2 = W[0]; cannam@167: T4 = W[1]; cannam@167: T6 = FMA(T2, T3, T4 * T5); cannam@167: TA = FNMS(T4, T3, T2 * T5); cannam@167: } cannam@167: { cannam@167: E T8, Ta, T7, T9; cannam@167: T8 = ri[WS(rs, 6)]; cannam@167: Ta = ii[WS(rs, 6)]; cannam@167: T7 = W[10]; cannam@167: T9 = W[11]; cannam@167: Tb = FMA(T7, T8, T9 * Ta); cannam@167: TB = FNMS(T9, T8, T7 * Ta); cannam@167: } cannam@167: Tc = T6 + Tb; cannam@167: TS = Tb - T6; cannam@167: TC = TA - TB; cannam@167: TO = TA + TB; cannam@167: } cannam@167: { cannam@167: E Th, TG, Tm, TH; cannam@167: { cannam@167: E Te, Tg, Td, Tf; cannam@167: Te = ri[WS(rs, 2)]; cannam@167: Tg = ii[WS(rs, 2)]; cannam@167: Td = W[2]; cannam@167: Tf = W[3]; cannam@167: Th = FMA(Td, Te, Tf * Tg); cannam@167: TG = FNMS(Tf, Te, Td * Tg); cannam@167: } cannam@167: { cannam@167: E Tj, Tl, Ti, Tk; cannam@167: Tj = ri[WS(rs, 5)]; cannam@167: Tl = ii[WS(rs, 5)]; cannam@167: Ti = W[8]; cannam@167: Tk = W[9]; cannam@167: Tm = FMA(Ti, Tj, Tk * Tl); cannam@167: TH = FNMS(Tk, Tj, Ti * Tl); cannam@167: } cannam@167: Tn = Th + Tm; cannam@167: TT = Tm - Th; cannam@167: TI = TG - TH; cannam@167: TP = TG + TH; cannam@167: } cannam@167: { cannam@167: E Ts, TD, Tx, TE; cannam@167: { cannam@167: E Tp, Tr, To, Tq; cannam@167: Tp = ri[WS(rs, 3)]; cannam@167: Tr = ii[WS(rs, 3)]; cannam@167: To = W[4]; cannam@167: Tq = W[5]; cannam@167: Ts = FMA(To, Tp, Tq * Tr); cannam@167: TD = FNMS(Tq, Tp, To * Tr); cannam@167: } cannam@167: { cannam@167: E Tu, Tw, Tt, Tv; cannam@167: Tu = ri[WS(rs, 4)]; cannam@167: Tw = ii[WS(rs, 4)]; cannam@167: Tt = W[6]; cannam@167: Tv = W[7]; cannam@167: Tx = FMA(Tt, Tu, Tv * Tw); cannam@167: TE = FNMS(Tv, Tu, Tt * Tw); cannam@167: } cannam@167: Ty = Ts + Tx; cannam@167: TU = Tx - Ts; cannam@167: TF = TD - TE; cannam@167: TQ = TD + TE; cannam@167: } cannam@167: ri[0] = T1 + Tc + Tn + Ty; cannam@167: ii[0] = TO + TP + TQ + TR; cannam@167: { cannam@167: E TJ, Tz, TX, TY; cannam@167: TJ = FNMS(KP781831482, TF, KP974927912 * TC) - (KP433883739 * TI); cannam@167: Tz = FMA(KP623489801, Ty, T1) + FNMA(KP900968867, Tn, KP222520933 * Tc); cannam@167: ri[WS(rs, 5)] = Tz - TJ; cannam@167: ri[WS(rs, 2)] = Tz + TJ; cannam@167: TX = FNMS(KP781831482, TU, KP974927912 * TS) - (KP433883739 * TT); cannam@167: TY = FMA(KP623489801, TQ, TR) + FNMA(KP900968867, TP, KP222520933 * TO); cannam@167: ii[WS(rs, 2)] = TX + TY; cannam@167: ii[WS(rs, 5)] = TY - TX; cannam@167: } cannam@167: { cannam@167: E TL, TK, TV, TW; cannam@167: TL = FMA(KP781831482, TC, KP974927912 * TI) + (KP433883739 * TF); cannam@167: TK = FMA(KP623489801, Tc, T1) + FNMA(KP900968867, Ty, KP222520933 * Tn); cannam@167: ri[WS(rs, 6)] = TK - TL; cannam@167: ri[WS(rs, 1)] = TK + TL; cannam@167: TV = FMA(KP781831482, TS, KP974927912 * TT) + (KP433883739 * TU); cannam@167: TW = FMA(KP623489801, TO, TR) + FNMA(KP900968867, TQ, KP222520933 * TP); cannam@167: ii[WS(rs, 1)] = TV + TW; cannam@167: ii[WS(rs, 6)] = TW - TV; cannam@167: } cannam@167: { cannam@167: E TN, TM, TZ, T10; cannam@167: TN = FMA(KP433883739, TC, KP974927912 * TF) - (KP781831482 * TI); cannam@167: TM = FMA(KP623489801, Tn, T1) + FNMA(KP222520933, Ty, KP900968867 * Tc); cannam@167: ri[WS(rs, 4)] = TM - TN; cannam@167: ri[WS(rs, 3)] = TM + TN; cannam@167: TZ = FMA(KP433883739, TS, KP974927912 * TU) - (KP781831482 * TT); cannam@167: T10 = FMA(KP623489801, TP, TR) + FNMA(KP222520933, TQ, KP900968867 * TO); cannam@167: ii[WS(rs, 3)] = TZ + T10; cannam@167: ii[WS(rs, 4)] = T10 - TZ; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 0, 7}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const ct_desc desc = { 7, "t1_7", twinstr, &GENUS, {36, 24, 36, 0}, 0, 0, 0 }; cannam@167: cannam@167: void X(codelet_t1_7) (planner *p) { cannam@167: X(kdft_dit_register) (p, t1_7, &desc); cannam@167: } cannam@167: #endif