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 08:07:28 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_r2cb.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cb_15 -include rdft/scalar/r2cb.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 43 FP multiplications, Chris@82: * (or, 21 additions, 0 multiplications, 43 fused multiply/add), Chris@82: * 46 stack variables, 9 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cb.h" Chris@82: Chris@82: static void r2cb_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(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@82: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@82: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@82: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@82: { Chris@82: INT i; Chris@82: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@82: E T3, Tt, Th, TC, TY, TZ, TD, TH, TI, Tm, Tu, Tr, Tv, T8, Td; Chris@82: E Te; Chris@82: { Chris@82: E Tg, T1, T2, Tf; Chris@82: Tg = Ci[WS(csi, 5)]; Chris@82: T1 = Cr[0]; Chris@82: T2 = Cr[WS(csr, 5)]; Chris@82: Tf = T1 - T2; Chris@82: T3 = FMA(KP2_000000000, T2, T1); Chris@82: Tt = FNMS(KP1_732050807, Tg, Tf); Chris@82: Th = FMA(KP1_732050807, Tg, Tf); Chris@82: } Chris@82: { Chris@82: E T4, TA, T9, TF, T5, T6, T7, Ta, Tb, Tc, Tq, TG, Tl, TB, Ti; Chris@82: E Tn; Chris@82: T4 = Cr[WS(csr, 3)]; Chris@82: TA = Ci[WS(csi, 3)]; Chris@82: T9 = Cr[WS(csr, 6)]; Chris@82: TF = Ci[WS(csi, 6)]; Chris@82: T5 = Cr[WS(csr, 7)]; Chris@82: T6 = Cr[WS(csr, 2)]; Chris@82: T7 = T5 + T6; Chris@82: Ta = Cr[WS(csr, 4)]; Chris@82: Tb = Cr[WS(csr, 1)]; Chris@82: Tc = Ta + Tb; Chris@82: { Chris@82: E To, Tp, Tj, Tk; Chris@82: To = Ci[WS(csi, 4)]; Chris@82: Tp = Ci[WS(csi, 1)]; Chris@82: Tq = To + Tp; Chris@82: TG = Tp - To; Chris@82: Tj = Ci[WS(csi, 7)]; Chris@82: Tk = Ci[WS(csi, 2)]; Chris@82: Tl = Tj - Tk; Chris@82: TB = Tj + Tk; Chris@82: } Chris@82: TC = FMA(KP500000000, TB, TA); Chris@82: TY = TG + TF; Chris@82: TZ = TA - TB; Chris@82: TD = T5 - T6; Chris@82: TH = FNMS(KP500000000, TG, TF); Chris@82: TI = Ta - Tb; Chris@82: Ti = FNMS(KP2_000000000, T4, T7); Chris@82: Tm = FMA(KP1_732050807, Tl, Ti); Chris@82: Tu = FNMS(KP1_732050807, Tl, Ti); Chris@82: Tn = FNMS(KP2_000000000, T9, Tc); Chris@82: Tr = FMA(KP1_732050807, Tq, Tn); Chris@82: Tv = FNMS(KP1_732050807, Tq, Tn); Chris@82: T8 = T4 + T7; Chris@82: Td = T9 + Tc; Chris@82: Te = T8 + Td; Chris@82: } Chris@82: R0[0] = FMA(KP2_000000000, Te, T3); Chris@82: { Chris@82: E T10, T12, TX, T11, TV, TW; Chris@82: T10 = FNMS(KP618033988, TZ, TY); Chris@82: T12 = FMA(KP618033988, TY, TZ); Chris@82: TV = FNMS(KP500000000, Te, T3); Chris@82: TW = T8 - Td; Chris@82: TX = FNMS(KP1_118033988, TW, TV); Chris@82: T11 = FMA(KP1_118033988, TW, TV); Chris@82: R1[WS(rs, 1)] = FNMS(KP1_902113032, T10, TX); Chris@82: R1[WS(rs, 4)] = FMA(KP1_902113032, T12, T11); Chris@82: R0[WS(rs, 6)] = FMA(KP1_902113032, T10, TX); Chris@82: R0[WS(rs, 3)] = FNMS(KP1_902113032, T12, T11); Chris@82: } Chris@82: { Chris@82: E TO, Ts, TN, TS, TU, TQ, TR, TT, TP; Chris@82: TO = Tr - Tm; Chris@82: Ts = Tm + Tr; Chris@82: TN = FMA(KP250000000, Ts, Th); Chris@82: TQ = FNMS(KP866025403, TI, TH); Chris@82: TR = FNMS(KP866025403, TD, TC); Chris@82: TS = FNMS(KP618033988, TR, TQ); Chris@82: TU = FMA(KP618033988, TQ, TR); Chris@82: R1[WS(rs, 2)] = Th - Ts; Chris@82: TT = FMA(KP559016994, TO, TN); Chris@82: R1[WS(rs, 5)] = FNMS(KP1_902113032, TU, TT); Chris@82: R0[WS(rs, 7)] = FMA(KP1_902113032, TU, TT); Chris@82: TP = FNMS(KP559016994, TO, TN); Chris@82: R0[WS(rs, 4)] = FNMS(KP1_902113032, TS, TP); Chris@82: R0[WS(rs, 1)] = FMA(KP1_902113032, TS, TP); Chris@82: } Chris@82: { Chris@82: E Ty, Tw, Tx, TK, TM, TE, TJ, TL, Tz; Chris@82: Ty = Tv - Tu; Chris@82: Tw = Tu + Tv; Chris@82: Tx = FMA(KP250000000, Tw, Tt); Chris@82: TE = FMA(KP866025403, TD, TC); Chris@82: TJ = FMA(KP866025403, TI, TH); Chris@82: TK = FMA(KP618033988, TJ, TE); Chris@82: TM = FNMS(KP618033988, TE, TJ); Chris@82: R0[WS(rs, 5)] = Tt - Tw; Chris@82: TL = FNMS(KP559016994, Ty, Tx); Chris@82: R1[WS(rs, 6)] = FNMS(KP1_902113032, TM, TL); Chris@82: R1[WS(rs, 3)] = FMA(KP1_902113032, TM, TL); Chris@82: Tz = FMA(KP559016994, Ty, Tx); Chris@82: R1[0] = FNMS(KP1_902113032, TK, Tz); Chris@82: R0[WS(rs, 2)] = FMA(KP1_902113032, TK, Tz); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cb_15", {21, 0, 43, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cb_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cb_15, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cb_15 -include rdft/scalar/r2cb.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 31 FP multiplications, Chris@82: * (or, 47 additions, 14 multiplications, 17 fused multiply/add), Chris@82: * 44 stack variables, 7 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cb.h" Chris@82: Chris@82: static void r2cb_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(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@82: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@82: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@82: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@82: { Chris@82: INT i; Chris@82: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@82: E T3, Tu, Ti, TB, TZ, T10, TE, TG, TJ, Tn, Tv, Ts, Tw, T8, Td; Chris@82: E Te; Chris@82: { Chris@82: E Th, T1, T2, Tf, Tg; Chris@82: Tg = Ci[WS(csi, 5)]; Chris@82: Th = KP1_732050807 * Tg; Chris@82: T1 = Cr[0]; Chris@82: T2 = Cr[WS(csr, 5)]; Chris@82: Tf = T1 - T2; Chris@82: T3 = FMA(KP2_000000000, T2, T1); Chris@82: Tu = Tf - Th; Chris@82: Ti = Tf + Th; Chris@82: } Chris@82: { Chris@82: E T4, TD, T9, TI, T5, T6, T7, Ta, Tb, Tc, Tr, TH, Tm, TC, Tj; Chris@82: E To; Chris@82: T4 = Cr[WS(csr, 3)]; Chris@82: TD = Ci[WS(csi, 3)]; Chris@82: T9 = Cr[WS(csr, 6)]; Chris@82: TI = Ci[WS(csi, 6)]; Chris@82: T5 = Cr[WS(csr, 7)]; Chris@82: T6 = Cr[WS(csr, 2)]; Chris@82: T7 = T5 + T6; Chris@82: Ta = Cr[WS(csr, 4)]; Chris@82: Tb = Cr[WS(csr, 1)]; Chris@82: Tc = Ta + Tb; Chris@82: { Chris@82: E Tp, Tq, Tk, Tl; Chris@82: Tp = Ci[WS(csi, 4)]; Chris@82: Tq = Ci[WS(csi, 1)]; Chris@82: Tr = KP866025403 * (Tp + Tq); Chris@82: TH = Tp - Tq; Chris@82: Tk = Ci[WS(csi, 7)]; Chris@82: Tl = Ci[WS(csi, 2)]; Chris@82: Tm = KP866025403 * (Tk - Tl); Chris@82: TC = Tk + Tl; Chris@82: } Chris@82: TB = KP866025403 * (T5 - T6); Chris@82: TZ = TD - TC; Chris@82: T10 = TI - TH; Chris@82: TE = FMA(KP500000000, TC, TD); Chris@82: TG = KP866025403 * (Ta - Tb); Chris@82: TJ = FMA(KP500000000, TH, TI); Chris@82: Tj = FNMS(KP500000000, T7, T4); Chris@82: Tn = Tj - Tm; Chris@82: Tv = Tj + Tm; Chris@82: To = FNMS(KP500000000, Tc, T9); Chris@82: Ts = To - Tr; Chris@82: Tw = To + Tr; Chris@82: T8 = T4 + T7; Chris@82: Td = T9 + Tc; Chris@82: Te = T8 + Td; Chris@82: } Chris@82: R0[0] = FMA(KP2_000000000, Te, T3); Chris@82: { Chris@82: E T11, T13, TY, T12, TW, TX; Chris@82: T11 = FNMS(KP1_902113032, T10, KP1_175570504 * TZ); Chris@82: T13 = FMA(KP1_902113032, TZ, KP1_175570504 * T10); Chris@82: TW = FNMS(KP500000000, Te, T3); Chris@82: TX = KP1_118033988 * (T8 - Td); Chris@82: TY = TW - TX; Chris@82: T12 = TX + TW; Chris@82: R0[WS(rs, 6)] = TY - T11; Chris@82: R1[WS(rs, 4)] = T12 + T13; Chris@82: R1[WS(rs, 1)] = TY + T11; Chris@82: R0[WS(rs, 3)] = T12 - T13; Chris@82: } Chris@82: { Chris@82: E TP, Tt, TO, TT, TV, TR, TS, TU, TQ; Chris@82: TP = KP1_118033988 * (Tn - Ts); Chris@82: Tt = Tn + Ts; Chris@82: TO = FNMS(KP500000000, Tt, Ti); Chris@82: TR = TE - TB; Chris@82: TS = TJ - TG; Chris@82: TT = FNMS(KP1_902113032, TS, KP1_175570504 * TR); Chris@82: TV = FMA(KP1_902113032, TR, KP1_175570504 * TS); Chris@82: R1[WS(rs, 2)] = FMA(KP2_000000000, Tt, Ti); Chris@82: TU = TP + TO; Chris@82: R1[WS(rs, 5)] = TU - TV; Chris@82: R0[WS(rs, 7)] = TU + TV; Chris@82: TQ = TO - TP; Chris@82: R0[WS(rs, 1)] = TQ - TT; Chris@82: R0[WS(rs, 4)] = TQ + TT; Chris@82: } Chris@82: { Chris@82: E Tz, Tx, Ty, TL, TN, TF, TK, TM, TA; Chris@82: Tz = KP1_118033988 * (Tv - Tw); Chris@82: Tx = Tv + Tw; Chris@82: Ty = FNMS(KP500000000, Tx, Tu); Chris@82: TF = TB + TE; Chris@82: TK = TG + TJ; Chris@82: TL = FNMS(KP1_902113032, TK, KP1_175570504 * TF); Chris@82: TN = FMA(KP1_902113032, TF, KP1_175570504 * TK); Chris@82: R0[WS(rs, 5)] = FMA(KP2_000000000, Tx, Tu); Chris@82: TM = Tz + Ty; Chris@82: R1[0] = TM - TN; Chris@82: R0[WS(rs, 2)] = TM + TN; Chris@82: TA = Ty - Tz; Chris@82: R1[WS(rs, 3)] = TA - TL; Chris@82: R1[WS(rs, 6)] = TA + TL; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cb_15", {47, 14, 17, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cb_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cb_15, &desc); Chris@82: } Chris@82: Chris@82: #endif