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 08:06:26 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 r2cf_15 -include rdft/scalar/r2cf.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 35 FP multiplications, Chris@82: * (or, 36 additions, 7 multiplications, 28 fused multiply/add), Chris@82: * 45 stack variables, 8 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cf.h" Chris@82: Chris@82: static void r2cf_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(KP910592997, +0.910592997310029334643087372129977886038870291); Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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 Ti, TR, TF, TM, TN, T7, Te, Tf, TV, TW, TX, Ts, Tv, TH, Tl; Chris@82: E To, TG, TS, TT, TU; Chris@82: { Chris@82: E TD, Tg, Th, TE; Chris@82: TD = R0[0]; Chris@82: Tg = R0[WS(rs, 5)]; Chris@82: Th = R1[WS(rs, 2)]; Chris@82: TE = Th + Tg; Chris@82: Ti = Tg - Th; Chris@82: TR = TD + TE; Chris@82: TF = FNMS(KP500000000, TE, TD); Chris@82: } Chris@82: { Chris@82: E Tj, Tq, Tt, Tm, T3, Tk, Ta, Tr, Td, Tu, T6, Tn; Chris@82: Tj = R1[WS(rs, 1)]; Chris@82: Tq = R0[WS(rs, 3)]; Chris@82: Tt = R1[WS(rs, 4)]; Chris@82: Tm = R0[WS(rs, 6)]; Chris@82: { Chris@82: E T1, T2, T8, T9; Chris@82: T1 = R0[WS(rs, 4)]; Chris@82: T2 = R1[WS(rs, 6)]; Chris@82: T3 = T1 - T2; Chris@82: Tk = T1 + T2; Chris@82: T8 = R1[WS(rs, 5)]; Chris@82: T9 = R1[0]; Chris@82: Ta = T8 - T9; Chris@82: Tr = T8 + T9; Chris@82: } Chris@82: { Chris@82: E Tb, Tc, T4, T5; Chris@82: Tb = R0[WS(rs, 7)]; Chris@82: Tc = R0[WS(rs, 2)]; Chris@82: Td = Tb - Tc; Chris@82: Tu = Tb + Tc; Chris@82: T4 = R0[WS(rs, 1)]; Chris@82: T5 = R1[WS(rs, 3)]; Chris@82: T6 = T4 - T5; Chris@82: Tn = T4 + T5; Chris@82: } Chris@82: TM = T6 - T3; Chris@82: TN = Td - Ta; Chris@82: T7 = T3 + T6; Chris@82: Te = Ta + Td; Chris@82: Tf = T7 + Te; Chris@82: TV = Tq + Tr; Chris@82: TW = Tt + Tu; Chris@82: TX = TV + TW; Chris@82: Ts = FNMS(KP500000000, Tr, Tq); Chris@82: Tv = FNMS(KP500000000, Tu, Tt); Chris@82: TH = Ts + Tv; Chris@82: Tl = FNMS(KP500000000, Tk, Tj); Chris@82: To = FNMS(KP500000000, Tn, Tm); Chris@82: TG = Tl + To; Chris@82: TS = Tj + Tk; Chris@82: TT = Tm + Tn; Chris@82: TU = TS + TT; Chris@82: } Chris@82: Ci[WS(csi, 5)] = KP866025403 * (Tf - Ti); Chris@82: { Chris@82: E TK, TQ, TO, TI, TJ, TP, TL; Chris@82: TK = TG - TH; Chris@82: TQ = FNMS(KP618033988, TM, TN); Chris@82: TO = FMA(KP618033988, TN, TM); Chris@82: TI = TG + TH; Chris@82: TJ = FNMS(KP250000000, TI, TF); Chris@82: Cr[WS(csr, 5)] = TF + TI; Chris@82: TP = FNMS(KP559016994, TK, TJ); Chris@82: Cr[WS(csr, 2)] = FMA(KP823639103, TQ, TP); Chris@82: Cr[WS(csr, 7)] = FNMS(KP823639103, TQ, TP); Chris@82: TL = FMA(KP559016994, TK, TJ); Chris@82: Cr[WS(csr, 1)] = FMA(KP823639103, TO, TL); Chris@82: Cr[WS(csr, 4)] = FNMS(KP823639103, TO, TL); Chris@82: } Chris@82: { Chris@82: E T11, T12, T10, TY, TZ; Chris@82: T11 = TW - TV; Chris@82: T12 = TS - TT; Chris@82: Ci[WS(csi, 3)] = KP951056516 * (FMA(KP618033988, T12, T11)); Chris@82: Ci[WS(csi, 6)] = -(KP951056516 * (FNMS(KP618033988, T11, T12))); Chris@82: T10 = TU - TX; Chris@82: TY = TU + TX; Chris@82: TZ = FNMS(KP250000000, TY, TR); Chris@82: Cr[WS(csr, 3)] = FNMS(KP559016994, T10, TZ); Chris@82: Cr[0] = TR + TY; Chris@82: Cr[WS(csr, 6)] = FMA(KP559016994, T10, TZ); Chris@82: { Chris@82: E Tx, TB, TA, TC; Chris@82: { Chris@82: E Tp, Tw, Ty, Tz; Chris@82: Tp = Tl - To; Chris@82: Tw = Ts - Tv; Chris@82: Tx = FMA(KP618033988, Tw, Tp); Chris@82: TB = FNMS(KP618033988, Tp, Tw); Chris@82: Ty = FMA(KP250000000, Tf, Ti); Chris@82: Tz = Te - T7; Chris@82: TA = FMA(KP559016994, Tz, Ty); Chris@82: TC = FNMS(KP559016994, Tz, Ty); Chris@82: } Chris@82: Ci[WS(csi, 1)] = -(KP951056516 * (FNMS(KP910592997, TA, Tx))); Chris@82: Ci[WS(csi, 7)] = KP951056516 * (FMA(KP910592997, TC, TB)); Chris@82: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP910592997, TA, Tx)); Chris@82: Ci[WS(csi, 2)] = KP951056516 * (FNMS(KP910592997, TC, TB)); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cf_15", {36, 7, 28, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cf_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cf_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 r2cf_15 -include rdft/scalar/r2cf.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 25 FP multiplications, Chris@82: * (or, 50 additions, 11 multiplications, 14 fused multiply/add), Chris@82: * 47 stack variables, 10 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cf.h" Chris@82: Chris@82: static void r2cf_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(KP484122918, +0.484122918275927110647408174972799951354115213); Chris@82: DK(KP216506350, +0.216506350946109661690930792688234045867850657); Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP509036960, +0.509036960455127183450980863393907648510733164); Chris@82: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@82: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); 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 Ti, TR, TL, TD, TE, T7, Te, Tf, TV, TW, TX, Tv, Ty, TH, To; Chris@82: E Tr, TG, TS, TT, TU; Chris@82: { Chris@82: E TJ, Tg, Th, TK; Chris@82: TJ = R0[0]; Chris@82: Tg = R0[WS(rs, 5)]; Chris@82: Th = R1[WS(rs, 2)]; Chris@82: TK = Th + Tg; Chris@82: Ti = Tg - Th; Chris@82: TR = TJ + TK; Chris@82: TL = FNMS(KP500000000, TK, TJ); Chris@82: } Chris@82: { Chris@82: E Tm, Tt, Tw, Tp, T3, Tx, Ta, Tn, Td, Tq, T6, Tu; Chris@82: Tm = R1[WS(rs, 1)]; Chris@82: Tt = R0[WS(rs, 3)]; Chris@82: Tw = R1[WS(rs, 4)]; Chris@82: Tp = R0[WS(rs, 6)]; Chris@82: { Chris@82: E T1, T2, T8, T9; Chris@82: T1 = R0[WS(rs, 7)]; Chris@82: T2 = R0[WS(rs, 2)]; Chris@82: T3 = T1 - T2; Chris@82: Tx = T1 + T2; Chris@82: T8 = R1[WS(rs, 6)]; Chris@82: T9 = R0[WS(rs, 4)]; Chris@82: Ta = T8 - T9; Chris@82: Tn = T9 + T8; Chris@82: } Chris@82: { Chris@82: E Tb, Tc, T4, T5; Chris@82: Tb = R1[WS(rs, 3)]; Chris@82: Tc = R0[WS(rs, 1)]; Chris@82: Td = Tb - Tc; Chris@82: Tq = Tc + Tb; Chris@82: T4 = R1[0]; Chris@82: T5 = R1[WS(rs, 5)]; Chris@82: T6 = T4 - T5; Chris@82: Tu = T5 + T4; Chris@82: } Chris@82: TD = Ta - Td; Chris@82: TE = T6 + T3; Chris@82: T7 = T3 - T6; Chris@82: Te = Ta + Td; Chris@82: Tf = T7 - Te; Chris@82: TV = Tt + Tu; Chris@82: TW = Tw + Tx; Chris@82: TX = TV + TW; Chris@82: Tv = FNMS(KP500000000, Tu, Tt); Chris@82: Ty = FNMS(KP500000000, Tx, Tw); Chris@82: TH = Tv + Ty; Chris@82: To = FNMS(KP500000000, Tn, Tm); Chris@82: Tr = FNMS(KP500000000, Tq, Tp); Chris@82: TG = To + Tr; Chris@82: TS = Tm + Tn; Chris@82: TT = Tp + Tq; Chris@82: TU = TS + TT; Chris@82: } Chris@82: Ci[WS(csi, 5)] = KP866025403 * (Tf - Ti); Chris@82: { Chris@82: E TF, TP, TI, TM, TN, TQ, TO; Chris@82: TF = FMA(KP823639103, TD, KP509036960 * TE); Chris@82: TP = FNMS(KP509036960, TD, KP823639103 * TE); Chris@82: TI = KP559016994 * (TG - TH); Chris@82: TM = TG + TH; Chris@82: TN = FNMS(KP250000000, TM, TL); Chris@82: Cr[WS(csr, 5)] = TL + TM; Chris@82: TQ = TN - TI; Chris@82: Cr[WS(csr, 2)] = TP + TQ; Chris@82: Cr[WS(csr, 7)] = TQ - TP; Chris@82: TO = TI + TN; Chris@82: Cr[WS(csr, 1)] = TF + TO; Chris@82: Cr[WS(csr, 4)] = TO - TF; Chris@82: } Chris@82: { Chris@82: E T11, T12, T10, TY, TZ; Chris@82: T11 = TS - TT; Chris@82: T12 = TW - TV; Chris@82: Ci[WS(csi, 3)] = FMA(KP587785252, T11, KP951056516 * T12); Chris@82: Ci[WS(csi, 6)] = FNMS(KP951056516, T11, KP587785252 * T12); Chris@82: T10 = KP559016994 * (TU - TX); Chris@82: TY = TU + TX; Chris@82: TZ = FNMS(KP250000000, TY, TR); Chris@82: Cr[WS(csr, 3)] = TZ - T10; Chris@82: Cr[0] = TR + TY; Chris@82: Cr[WS(csr, 6)] = T10 + TZ; Chris@82: { Chris@82: E Tl, TB, TA, TC; Chris@82: { Chris@82: E Tj, Tk, Ts, Tz; Chris@82: Tj = FMA(KP866025403, Ti, KP216506350 * Tf); Chris@82: Tk = KP484122918 * (Te + T7); Chris@82: Tl = Tj + Tk; Chris@82: TB = Tk - Tj; Chris@82: Ts = To - Tr; Chris@82: Tz = Tv - Ty; Chris@82: TA = FMA(KP951056516, Ts, KP587785252 * Tz); Chris@82: TC = FNMS(KP587785252, Ts, KP951056516 * Tz); Chris@82: } Chris@82: Ci[WS(csi, 1)] = Tl - TA; Chris@82: Ci[WS(csi, 7)] = TC - TB; Chris@82: Ci[WS(csi, 4)] = Tl + TA; Chris@82: Ci[WS(csi, 2)] = TB + TC; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cf_15", {50, 11, 14, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cf_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cf_15, &desc); Chris@82: } Chris@82: Chris@82: #endif