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 08:07:57 EDT 2018 */ cannam@167: cannam@167: #include "rdft/codelet-rdft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cbdft_6 -include rdft/scalar/hc2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 58 FP additions, 32 FP multiplications, cannam@167: * (or, 36 additions, 10 multiplications, 22 fused multiply/add), cannam@167: * 34 stack variables, 2 constants, and 24 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cb.h" cannam@167: cannam@167: static void hc2cbdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@167: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) { cannam@167: E Tp, TD, Tj, TV, Tq, Tr, TG, TP, T4, Ts, TQ, Tb, Tc, TA, TU; cannam@167: { cannam@167: E Tf, TF, Ti, TE, Td, Te; cannam@167: Td = Ip[WS(rs, 1)]; cannam@167: Te = Im[WS(rs, 1)]; cannam@167: Tf = Td - Te; cannam@167: TF = Te + Td; cannam@167: { cannam@167: E Tn, To, Tg, Th; cannam@167: Tn = Ip[0]; cannam@167: To = Im[WS(rs, 2)]; cannam@167: Tp = Tn - To; cannam@167: TD = Tn + To; cannam@167: Tg = Ip[WS(rs, 2)]; cannam@167: Th = Im[0]; cannam@167: Ti = Tg - Th; cannam@167: TE = Tg + Th; cannam@167: } cannam@167: Tj = Tf - Ti; cannam@167: TV = TF + TE; cannam@167: Tq = Tf + Ti; cannam@167: Tr = FNMS(KP500000000, Tq, Tp); cannam@167: TG = TE - TF; cannam@167: TP = FNMS(KP500000000, TG, TD); cannam@167: } cannam@167: { cannam@167: E Tw, Ta, Ty, T7, Tx, T2, T3, Tz; cannam@167: T2 = Rp[0]; cannam@167: T3 = Rm[WS(rs, 2)]; cannam@167: T4 = T2 + T3; cannam@167: Tw = T2 - T3; cannam@167: { cannam@167: E T8, T9, T5, T6; cannam@167: T8 = Rm[WS(rs, 1)]; cannam@167: T9 = Rp[WS(rs, 1)]; cannam@167: Ta = T8 + T9; cannam@167: Ty = T8 - T9; cannam@167: T5 = Rp[WS(rs, 2)]; cannam@167: T6 = Rm[0]; cannam@167: T7 = T5 + T6; cannam@167: Tx = T5 - T6; cannam@167: } cannam@167: Ts = T7 - Ta; cannam@167: TQ = Tx - Ty; cannam@167: Tb = T7 + Ta; cannam@167: Tc = FNMS(KP500000000, Tb, T4); cannam@167: Tz = Tx + Ty; cannam@167: TA = Tw + Tz; cannam@167: TU = FNMS(KP500000000, Tz, Tw); cannam@167: } cannam@167: { cannam@167: E TN, TY, TR, TW, TS, TZ, TO, TX, T10, TT; cannam@167: TN = T4 + Tb; cannam@167: TY = Tp + Tq; cannam@167: TR = FMA(KP866025403, TQ, TP); cannam@167: TW = FNMS(KP866025403, TV, TU); cannam@167: TO = W[0]; cannam@167: TS = TO * TR; cannam@167: TZ = TO * TW; cannam@167: TT = W[1]; cannam@167: TX = FMA(TT, TW, TS); cannam@167: T10 = FNMS(TT, TR, TZ); cannam@167: Rp[0] = TN - TX; cannam@167: Ip[0] = TY + T10; cannam@167: Rm[0] = TN + TX; cannam@167: Im[0] = T10 - TY; cannam@167: } cannam@167: { cannam@167: E Tt, TH, Tv, TB, TC, TL, T1, Tl, Tm, TJ, Tk; cannam@167: Tt = FNMS(KP866025403, Ts, Tr); cannam@167: TH = TD + TG; cannam@167: Tv = W[4]; cannam@167: TB = Tv * TA; cannam@167: TC = W[5]; cannam@167: TL = TC * TA; cannam@167: Tk = FNMS(KP866025403, Tj, Tc); cannam@167: T1 = W[3]; cannam@167: Tl = T1 * Tk; cannam@167: Tm = W[2]; cannam@167: TJ = Tm * Tk; cannam@167: { cannam@167: E Tu, TI, TK, TM; cannam@167: Tu = FMA(Tm, Tt, Tl); cannam@167: TI = FNMS(TC, TH, TB); cannam@167: Ip[WS(rs, 1)] = Tu + TI; cannam@167: Im[WS(rs, 1)] = TI - Tu; cannam@167: TK = FNMS(T1, Tt, TJ); cannam@167: TM = FMA(Tv, TH, TL); cannam@167: Rp[WS(rs, 1)] = TK - TM; cannam@167: Rm[WS(rs, 1)] = TK + TM; cannam@167: } cannam@167: } cannam@167: { cannam@167: E T15, T11, T13, T14, T1d, T18, T1b, T19, T1f, T12, T17; cannam@167: T15 = FMA(KP866025403, Ts, Tr); cannam@167: T12 = FMA(KP866025403, Tj, Tc); cannam@167: T11 = W[6]; cannam@167: T13 = T11 * T12; cannam@167: T14 = W[7]; cannam@167: T1d = T14 * T12; cannam@167: T18 = FNMS(KP866025403, TQ, TP); cannam@167: T1b = FMA(KP866025403, TV, TU); cannam@167: T17 = W[8]; cannam@167: T19 = T17 * T18; cannam@167: T1f = T17 * T1b; cannam@167: { cannam@167: E T16, T1e, T1c, T1g, T1a; cannam@167: T16 = FNMS(T14, T15, T13); cannam@167: T1e = FMA(T11, T15, T1d); cannam@167: T1a = W[9]; cannam@167: T1c = FMA(T1a, T1b, T19); cannam@167: T1g = FNMS(T1a, T18, T1f); cannam@167: Rp[WS(rs, 2)] = T16 - T1c; cannam@167: Ip[WS(rs, 2)] = T1e + T1g; cannam@167: Rm[WS(rs, 2)] = T16 + T1c; cannam@167: Im[WS(rs, 2)] = T1g - T1e; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 6}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {36, 10, 22, 0} }; cannam@167: cannam@167: void X(codelet_hc2cbdft_6) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cbdft_6 -include rdft/scalar/hc2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 58 FP additions, 28 FP multiplications, cannam@167: * (or, 44 additions, 14 multiplications, 14 fused multiply/add), cannam@167: * 29 stack variables, 2 constants, and 24 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cb.h" cannam@167: cannam@167: static void hc2cbdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@167: DK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) { cannam@167: E T4, Tv, Tr, TL, Tb, Tc, Ty, TP, To, TB, Tj, TQ, Tp, Tq, TE; cannam@167: E TM; cannam@167: { cannam@167: E Ta, Tx, T7, Tw, T2, T3; cannam@167: T2 = Rp[0]; cannam@167: T3 = Rm[WS(rs, 2)]; cannam@167: T4 = T2 + T3; cannam@167: Tv = T2 - T3; cannam@167: { cannam@167: E T8, T9, T5, T6; cannam@167: T8 = Rm[WS(rs, 1)]; cannam@167: T9 = Rp[WS(rs, 1)]; cannam@167: Ta = T8 + T9; cannam@167: Tx = T8 - T9; cannam@167: T5 = Rp[WS(rs, 2)]; cannam@167: T6 = Rm[0]; cannam@167: T7 = T5 + T6; cannam@167: Tw = T5 - T6; cannam@167: } cannam@167: Tr = KP866025403 * (T7 - Ta); cannam@167: TL = KP866025403 * (Tw - Tx); cannam@167: Tb = T7 + Ta; cannam@167: Tc = FNMS(KP500000000, Tb, T4); cannam@167: Ty = Tw + Tx; cannam@167: TP = FNMS(KP500000000, Ty, Tv); cannam@167: } cannam@167: { cannam@167: E Tf, TC, Ti, TD, Td, Te; cannam@167: Td = Ip[WS(rs, 1)]; cannam@167: Te = Im[WS(rs, 1)]; cannam@167: Tf = Td - Te; cannam@167: TC = Te + Td; cannam@167: { cannam@167: E Tm, Tn, Tg, Th; cannam@167: Tm = Ip[0]; cannam@167: Tn = Im[WS(rs, 2)]; cannam@167: To = Tm - Tn; cannam@167: TB = Tm + Tn; cannam@167: Tg = Ip[WS(rs, 2)]; cannam@167: Th = Im[0]; cannam@167: Ti = Tg - Th; cannam@167: TD = Tg + Th; cannam@167: } cannam@167: Tj = KP866025403 * (Tf - Ti); cannam@167: TQ = KP866025403 * (TC + TD); cannam@167: Tp = Tf + Ti; cannam@167: Tq = FNMS(KP500000000, Tp, To); cannam@167: TE = TC - TD; cannam@167: TM = FMA(KP500000000, TE, TB); cannam@167: } cannam@167: { cannam@167: E TJ, TT, TS, TU; cannam@167: TJ = T4 + Tb; cannam@167: TT = To + Tp; cannam@167: { cannam@167: E TN, TR, TK, TO; cannam@167: TN = TL + TM; cannam@167: TR = TP - TQ; cannam@167: TK = W[0]; cannam@167: TO = W[1]; cannam@167: TS = FMA(TK, TN, TO * TR); cannam@167: TU = FNMS(TO, TN, TK * TR); cannam@167: } cannam@167: Rp[0] = TJ - TS; cannam@167: Ip[0] = TT + TU; cannam@167: Rm[0] = TJ + TS; cannam@167: Im[0] = TU - TT; cannam@167: } cannam@167: { cannam@167: E TZ, T15, T14, T16; cannam@167: { cannam@167: E TW, TY, TV, TX; cannam@167: TW = Tc + Tj; cannam@167: TY = Tr + Tq; cannam@167: TV = W[6]; cannam@167: TX = W[7]; cannam@167: TZ = FNMS(TX, TY, TV * TW); cannam@167: T15 = FMA(TX, TW, TV * TY); cannam@167: } cannam@167: { cannam@167: E T11, T13, T10, T12; cannam@167: T11 = TM - TL; cannam@167: T13 = TP + TQ; cannam@167: T10 = W[8]; cannam@167: T12 = W[9]; cannam@167: T14 = FMA(T10, T11, T12 * T13); cannam@167: T16 = FNMS(T12, T11, T10 * T13); cannam@167: } cannam@167: Rp[WS(rs, 2)] = TZ - T14; cannam@167: Ip[WS(rs, 2)] = T15 + T16; cannam@167: Rm[WS(rs, 2)] = TZ + T14; cannam@167: Im[WS(rs, 2)] = T16 - T15; cannam@167: } cannam@167: { cannam@167: E Tt, TH, TG, TI; cannam@167: { cannam@167: E Tk, Ts, T1, Tl; cannam@167: Tk = Tc - Tj; cannam@167: Ts = Tq - Tr; cannam@167: T1 = W[3]; cannam@167: Tl = W[2]; cannam@167: Tt = FMA(T1, Tk, Tl * Ts); cannam@167: TH = FNMS(T1, Ts, Tl * Tk); cannam@167: } cannam@167: { cannam@167: E Tz, TF, Tu, TA; cannam@167: Tz = Tv + Ty; cannam@167: TF = TB - TE; cannam@167: Tu = W[4]; cannam@167: TA = W[5]; cannam@167: TG = FNMS(TA, TF, Tu * Tz); cannam@167: TI = FMA(TA, Tz, Tu * TF); cannam@167: } cannam@167: Ip[WS(rs, 1)] = Tt + TG; cannam@167: Rp[WS(rs, 1)] = TH - TI; cannam@167: Im[WS(rs, 1)] = TG - Tt; cannam@167: Rm[WS(rs, 1)] = TH + TI; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 6}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {44, 14, 14, 0} }; cannam@167: cannam@167: void X(codelet_hc2cbdft_6) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT); cannam@167: } cannam@167: #endif