Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:06:43 EDT 2018 */ Chris@82: Chris@82: #include "rdft/codelet-rdft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cf.native -fma -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include rdft/scalar/r2cfII.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 72 FP additions, 41 FP multiplications, Chris@82: * (or, 38 additions, 7 multiplications, 34 fused multiply/add), Chris@82: * 42 stack variables, 12 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cfII.h" Chris@82: Chris@82: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@82: DK(KP910592997, +0.910592997310029334643087372129977886038870291); Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP690983005, +0.690983005625052575897706582817180941139845410); Chris@82: DK(KP447213595, +0.447213595499957939281834733746255247088123672); Chris@82: DK(KP552786404, +0.552786404500042060718165266253744752911876328); Chris@82: DK(KP809016994, +0.809016994374947424102293417182819058860154590); Chris@82: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: { Chris@82: INT i; Chris@82: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@82: E Ta, Tl, T1, T6, T7, TX, TT, T8, Tg, Th, TM, TZ, Tj, Tz, Tr; Chris@82: E Ts, TP, TY, Tu, TC; Chris@82: Ta = R0[WS(rs, 5)]; Chris@82: Tl = R1[WS(rs, 2)]; Chris@82: { Chris@82: E T2, T5, T3, T4, TR, TS; Chris@82: T1 = R0[0]; Chris@82: T2 = R0[WS(rs, 3)]; Chris@82: T5 = R1[WS(rs, 4)]; Chris@82: T3 = R0[WS(rs, 6)]; Chris@82: T4 = R1[WS(rs, 1)]; Chris@82: TR = T2 + T5; Chris@82: TS = T3 + T4; Chris@82: T6 = T2 + T3 - T4 - T5; Chris@82: T7 = FNMS(KP250000000, T6, T1); Chris@82: TX = FNMS(KP618033988, TR, TS); Chris@82: TT = FMA(KP618033988, TS, TR); Chris@82: T8 = (T3 + T5 - T2) - T4; Chris@82: } Chris@82: { Chris@82: E Tf, TL, TK, Ti, Ty; Chris@82: { Chris@82: E Tb, Tc, Td, Te; Chris@82: Tb = R1[0]; Chris@82: Tg = R0[WS(rs, 2)]; Chris@82: Tc = R1[WS(rs, 3)]; Chris@82: Td = R1[WS(rs, 6)]; Chris@82: Te = Tc + Td; Chris@82: Tf = Tb - Te; Chris@82: TL = Tc - Td; Chris@82: Th = Tb + Te; Chris@82: TK = Tg + Tb; Chris@82: } Chris@82: TM = FMA(KP618033988, TL, TK); Chris@82: TZ = FNMS(KP618033988, TK, TL); Chris@82: Ti = FMA(KP809016994, Th, Tg); Chris@82: Tj = FNMS(KP552786404, Ti, Tf); Chris@82: Ty = FMA(KP447213595, Th, Tf); Chris@82: Tz = FNMS(KP690983005, Ty, Tg); Chris@82: } Chris@82: { Chris@82: E Tq, TO, TN, Tt, TB; Chris@82: { Chris@82: E Tm, Tn, To, Tp; Chris@82: Tm = R0[WS(rs, 7)]; Chris@82: Tr = R1[WS(rs, 5)]; Chris@82: Tn = R0[WS(rs, 1)]; Chris@82: To = R0[WS(rs, 4)]; Chris@82: Tp = Tn + To; Chris@82: Tq = Tm - Tp; Chris@82: TO = To - Tn; Chris@82: Ts = Tm + Tp; Chris@82: TN = Tr + Tm; Chris@82: } Chris@82: TP = FMA(KP618033988, TO, TN); Chris@82: TY = FNMS(KP618033988, TN, TO); Chris@82: Tt = FMA(KP809016994, Ts, Tr); Chris@82: Tu = FNMS(KP552786404, Tt, Tq); Chris@82: TB = FMA(KP447213595, Ts, Tq); Chris@82: TC = FNMS(KP690983005, TB, Tr); Chris@82: } Chris@82: { Chris@82: E TF, TG, TH, TI; Chris@82: TF = T1 + T6; Chris@82: TG = Ts - Tr - Tl; Chris@82: TH = Ta + Tg - Th; Chris@82: TI = TG + TH; Chris@82: Cr[WS(csr, 2)] = FNMS(KP500000000, TI, TF); Chris@82: Ci[WS(csi, 2)] = KP866025403 * (TH - TG); Chris@82: Cr[WS(csr, 7)] = TF + TI; Chris@82: } Chris@82: { Chris@82: E Tx, T14, T10, T11, TE, T12, TA, TD, T13; Chris@82: Tx = FMA(KP559016994, T8, T7); Chris@82: T14 = TZ - TY; Chris@82: T10 = TY + TZ; Chris@82: T11 = FMA(KP500000000, T10, TX); Chris@82: TA = FNMS(KP809016994, Tz, Ta); Chris@82: TD = FNMS(KP809016994, TC, Tl); Chris@82: TE = TA - TD; Chris@82: T12 = TD + TA; Chris@82: Cr[WS(csr, 1)] = Tx + TE; Chris@82: Ci[WS(csi, 1)] = KP951056516 * (T10 - TX); Chris@82: Ci[WS(csi, 3)] = KP951056516 * (FNMS(KP910592997, T12, T11)); Chris@82: Ci[WS(csi, 6)] = -(KP951056516 * (FMA(KP910592997, T12, T11))); Chris@82: T13 = FNMS(KP500000000, TE, Tx); Chris@82: Cr[WS(csr, 3)] = FNMS(KP823639103, T14, T13); Chris@82: Cr[WS(csr, 6)] = FMA(KP823639103, T14, T13); Chris@82: } Chris@82: { Chris@82: E T9, TQ, TU, TV, Tw, TW, Tk, Tv, TJ; Chris@82: T9 = FNMS(KP559016994, T8, T7); Chris@82: TQ = TM - TP; Chris@82: TU = TP + TM; Chris@82: TV = FMA(KP500000000, TU, TT); Chris@82: Tk = FNMS(KP559016994, Tj, Ta); Chris@82: Tv = FNMS(KP559016994, Tu, Tl); Chris@82: Tw = Tk - Tv; Chris@82: TW = Tv + Tk; Chris@82: Cr[WS(csr, 4)] = T9 + Tw; Chris@82: Ci[WS(csi, 4)] = KP951056516 * (TT - TU); Chris@82: Ci[0] = -(KP951056516 * (FMA(KP910592997, TW, TV))); Chris@82: Ci[WS(csi, 5)] = -(KP951056516 * (FNMS(KP910592997, TW, TV))); Chris@82: TJ = FNMS(KP500000000, Tw, T9); Chris@82: Cr[WS(csr, 5)] = FNMS(KP823639103, TQ, TJ); Chris@82: Cr[0] = FMA(KP823639103, TQ, TJ); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cfII_15", {38, 7, 34, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cfII_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cfII_15, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cfII_15 -dft-II -include rdft/scalar/r2cfII.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 72 FP additions, 33 FP multiplications, Chris@82: * (or, 54 additions, 15 multiplications, 18 fused multiply/add), Chris@82: * 37 stack variables, 8 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cfII.h" Chris@82: Chris@82: static void r2cfII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP809016994, +0.809016994374947424102293417182819058860154590); Chris@82: DK(KP309016994, +0.309016994374947424102293417182819058860154590); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: { Chris@82: INT i; Chris@82: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@82: E T1, T2, Tx, TR, TE, T7, TD, Th, Tm, Tr, TQ, TA, TB, Tf, Te; Chris@82: E Tu, TS, Td, TH, TO; Chris@82: T1 = R0[WS(rs, 5)]; Chris@82: { Chris@82: E T3, Tv, T6, Tw, T4, T5; Chris@82: T2 = R0[WS(rs, 2)]; Chris@82: T3 = R1[0]; Chris@82: Tv = T2 + T3; Chris@82: T4 = R1[WS(rs, 3)]; Chris@82: T5 = R1[WS(rs, 6)]; Chris@82: T6 = T4 + T5; Chris@82: Tw = T4 - T5; Chris@82: Tx = FMA(KP951056516, Tv, KP587785252 * Tw); Chris@82: TR = FNMS(KP587785252, Tv, KP951056516 * Tw); Chris@82: TE = KP559016994 * (T3 - T6); Chris@82: T7 = T3 + T6; Chris@82: TD = KP250000000 * T7; Chris@82: } Chris@82: { Chris@82: E Ti, Tl, Tj, Tk, Tp, Tq; Chris@82: Th = R0[0]; Chris@82: Ti = R1[WS(rs, 4)]; Chris@82: Tl = R0[WS(rs, 6)]; Chris@82: Tj = R1[WS(rs, 1)]; Chris@82: Tk = R0[WS(rs, 3)]; Chris@82: Tp = Tk + Ti; Chris@82: Tq = Tl + Tj; Chris@82: Tm = Ti + Tj - (Tk + Tl); Chris@82: Tr = FMA(KP951056516, Tp, KP587785252 * Tq); Chris@82: TQ = FNMS(KP951056516, Tq, KP587785252 * Tp); Chris@82: TA = FMA(KP250000000, Tm, Th); Chris@82: TB = KP559016994 * (Tl + Ti - (Tk + Tj)); Chris@82: } Chris@82: { Chris@82: E T9, Tt, Tc, Ts, Ta, Tb, TG; Chris@82: Tf = R1[WS(rs, 2)]; Chris@82: T9 = R0[WS(rs, 7)]; Chris@82: Te = R1[WS(rs, 5)]; Chris@82: Tt = T9 + Te; Chris@82: Ta = R0[WS(rs, 1)]; Chris@82: Tb = R0[WS(rs, 4)]; Chris@82: Tc = Ta + Tb; Chris@82: Ts = Ta - Tb; Chris@82: Tu = FNMS(KP951056516, Tt, KP587785252 * Ts); Chris@82: TS = FMA(KP951056516, Ts, KP587785252 * Tt); Chris@82: Td = T9 + Tc; Chris@82: TG = KP559016994 * (T9 - Tc); Chris@82: TH = FNMS(KP309016994, Te, TG) + FNMA(KP250000000, Td, Tf); Chris@82: TO = FMS(KP809016994, Te, Tf) + FNMA(KP250000000, Td, TG); Chris@82: } Chris@82: { Chris@82: E Tn, T8, Tg, To; Chris@82: Tn = Th - Tm; Chris@82: T8 = T1 + T2 - T7; Chris@82: Tg = Td - Te - Tf; Chris@82: To = T8 + Tg; Chris@82: Ci[WS(csi, 2)] = KP866025403 * (T8 - Tg); Chris@82: Cr[WS(csr, 2)] = FNMS(KP500000000, To, Tn); Chris@82: Cr[WS(csr, 7)] = Tn + To; Chris@82: } Chris@82: { Chris@82: E TM, TX, TT, TV, TP, TU, TN, TW; Chris@82: TM = TB + TA; Chris@82: TX = KP866025403 * (TR + TS); Chris@82: TT = TR - TS; Chris@82: TV = FMS(KP500000000, TT, TQ); Chris@82: TN = T1 + TE + FNMS(KP809016994, T2, TD); Chris@82: TP = TN + TO; Chris@82: TU = KP866025403 * (TO - TN); Chris@82: Cr[WS(csr, 1)] = TM + TP; Chris@82: Ci[WS(csi, 1)] = TQ + TT; Chris@82: Ci[WS(csi, 6)] = TU - TV; Chris@82: Ci[WS(csi, 3)] = TU + TV; Chris@82: TW = FNMS(KP500000000, TP, TM); Chris@82: Cr[WS(csr, 3)] = TW - TX; Chris@82: Cr[WS(csr, 6)] = TW + TX; Chris@82: } Chris@82: { Chris@82: E Tz, TC, Ty, TK, TI, TL, TF, TJ; Chris@82: Tz = KP866025403 * (Tx + Tu); Chris@82: TC = TA - TB; Chris@82: Ty = Tu - Tx; Chris@82: TK = FMS(KP500000000, Ty, Tr); Chris@82: TF = FMA(KP309016994, T2, T1) + TD - TE; Chris@82: TI = TF + TH; Chris@82: TL = KP866025403 * (TH - TF); Chris@82: Ci[WS(csi, 4)] = Tr + Ty; Chris@82: Cr[WS(csr, 4)] = TC + TI; Chris@82: Ci[WS(csi, 5)] = TK - TL; Chris@82: Ci[0] = TK + TL; Chris@82: TJ = FNMS(KP500000000, TI, TC); Chris@82: Cr[0] = Tz + TJ; Chris@82: Cr[WS(csr, 5)] = TJ - Tz; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cfII_15", {54, 15, 18, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cfII_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cfII_15, &desc); Chris@82: } Chris@82: Chris@82: #endif