Chris@10: /* Chris@10: * Copyright (c) 2003, 2007-11 Matteo Frigo Chris@10: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology Chris@10: * Chris@10: * This program is free software; you can redistribute it and/or modify Chris@10: * it under the terms of the GNU General Public License as published by Chris@10: * the Free Software Foundation; either version 2 of the License, or Chris@10: * (at your option) any later version. Chris@10: * Chris@10: * This program is distributed in the hope that it will be useful, Chris@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@10: * GNU General Public License for more details. Chris@10: * Chris@10: * You should have received a copy of the GNU General Public License Chris@10: * along with this program; if not, write to the Free Software Chris@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@10: * Chris@10: */ Chris@10: Chris@10: /* This file was automatically generated --- DO NOT EDIT */ Chris@10: /* Generated on Sun Nov 25 07:39:46 EST 2012 */ Chris@10: Chris@10: #include "codelet-rdft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cf_15 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 35 FP multiplications, Chris@10: * (or, 36 additions, 7 multiplications, 28 fused multiply/add), Chris@10: * 50 stack variables, 8 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP910592997, +0.910592997310029334643087372129977886038870291); Chris@10: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E Tw, Tz, Tp, Ty; Chris@10: { Chris@10: E Ti, TF, TR, TN, TX, T11, TM, TS, Tl, TH, Tf, To, TT, TD, Tg; Chris@10: E Th; Chris@10: TD = R0[0]; Chris@10: Tg = R0[WS(rs, 5)]; Chris@10: Th = R1[WS(rs, 2)]; Chris@10: { Chris@10: E Tj, Tq, Tt, Tm, T3, Tk, T4, Ta, Tr, Td, Tu, T5, TE; Chris@10: Tj = R1[WS(rs, 1)]; Chris@10: Tq = R0[WS(rs, 3)]; Chris@10: Tt = R1[WS(rs, 4)]; Chris@10: TE = Th + Tg; Chris@10: Ti = Tg - Th; Chris@10: Tm = R0[WS(rs, 6)]; Chris@10: { Chris@10: E T8, T9, T1, T2, Tb, Tc; Chris@10: T1 = R0[WS(rs, 4)]; Chris@10: T2 = R1[WS(rs, 6)]; Chris@10: TF = FNMS(KP500000000, TE, TD); Chris@10: TR = TD + TE; Chris@10: T8 = R1[WS(rs, 5)]; Chris@10: T3 = T1 - T2; Chris@10: Tk = T1 + T2; Chris@10: T9 = R1[0]; Chris@10: Tb = R0[WS(rs, 7)]; Chris@10: Tc = R0[WS(rs, 2)]; Chris@10: T4 = R0[WS(rs, 1)]; Chris@10: Ta = T8 - T9; Chris@10: Tr = T8 + T9; Chris@10: Td = Tb - Tc; Chris@10: Tu = Tb + Tc; Chris@10: T5 = R1[WS(rs, 3)]; Chris@10: } Chris@10: { Chris@10: E Ts, Tv, Te, Tn, T7, T6, TV, TW; Chris@10: TV = Tq + Tr; Chris@10: Ts = FNMS(KP500000000, Tr, Tq); Chris@10: Tv = FNMS(KP500000000, Tu, Tt); Chris@10: TW = Tt + Tu; Chris@10: Te = Ta + Td; Chris@10: TN = Td - Ta; Chris@10: Tn = T4 + T5; Chris@10: T6 = T4 - T5; Chris@10: TX = TV + TW; Chris@10: T11 = TW - TV; Chris@10: TM = T6 - T3; Chris@10: T7 = T3 + T6; Chris@10: TS = Tj + Tk; Chris@10: Tl = FNMS(KP500000000, Tk, Tj); Chris@10: TH = Ts + Tv; Chris@10: Tw = Ts - Tv; Chris@10: Tz = Te - T7; Chris@10: Tf = T7 + Te; Chris@10: To = FNMS(KP500000000, Tn, Tm); Chris@10: TT = Tm + Tn; Chris@10: } Chris@10: } Chris@10: { Chris@10: E TO, TQ, TU, T12, TK, TI, TG; Chris@10: Ci[WS(csi, 5)] = KP866025403 * (Tf - Ti); Chris@10: TG = Tl + To; Chris@10: Tp = Tl - To; Chris@10: TO = FMA(KP618033988, TN, TM); Chris@10: TQ = FNMS(KP618033988, TM, TN); Chris@10: TU = TS + TT; Chris@10: T12 = TS - TT; Chris@10: TK = TG - TH; Chris@10: TI = TG + TH; Chris@10: { Chris@10: E T10, TY, TL, TP, TJ, TZ; Chris@10: T10 = TU - TX; Chris@10: TY = TU + TX; Chris@10: Cr[WS(csr, 5)] = TF + TI; Chris@10: TJ = FNMS(KP250000000, TI, TF); Chris@10: Ci[WS(csi, 6)] = -(KP951056516 * (FNMS(KP618033988, T11, T12))); Chris@10: Ci[WS(csi, 3)] = KP951056516 * (FMA(KP618033988, T12, T11)); Chris@10: TL = FMA(KP559016994, TK, TJ); Chris@10: TP = FNMS(KP559016994, TK, TJ); Chris@10: Cr[0] = TR + TY; Chris@10: TZ = FNMS(KP250000000, TY, TR); Chris@10: Cr[WS(csr, 4)] = FNMS(KP823639103, TO, TL); Chris@10: Cr[WS(csr, 1)] = FMA(KP823639103, TO, TL); Chris@10: Cr[WS(csr, 7)] = FNMS(KP823639103, TQ, TP); Chris@10: Cr[WS(csr, 2)] = FMA(KP823639103, TQ, TP); Chris@10: Cr[WS(csr, 6)] = FMA(KP559016994, T10, TZ); Chris@10: Cr[WS(csr, 3)] = FNMS(KP559016994, T10, TZ); Chris@10: Ty = FMA(KP250000000, Tf, Ti); Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E TB, Tx, TC, TA; Chris@10: TB = FNMS(KP618033988, Tp, Tw); Chris@10: Tx = FMA(KP618033988, Tw, Tp); Chris@10: TC = FNMS(KP559016994, Tz, Ty); Chris@10: TA = FMA(KP559016994, Tz, Ty); Chris@10: Ci[WS(csi, 2)] = KP951056516 * (FNMS(KP910592997, TC, TB)); Chris@10: Ci[WS(csi, 7)] = KP951056516 * (FMA(KP910592997, TC, TB)); Chris@10: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP910592997, TA, Tx)); Chris@10: Ci[WS(csi, 1)] = -(KP951056516 * (FNMS(KP910592997, TA, Tx))); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cf_15", {36, 7, 28, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_15, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 15 -name r2cf_15 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 25 FP multiplications, Chris@10: * (or, 50 additions, 11 multiplications, 14 fused multiply/add), Chris@10: * 47 stack variables, 10 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP484122918, +0.484122918275927110647408174972799951354115213); Chris@10: DK(KP216506350, +0.216506350946109661690930792688234045867850657); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP509036960, +0.509036960455127183450980863393907648510733164); Chris@10: DK(KP823639103, +0.823639103546331925877420039278190003029660514); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E Ti, TR, TL, TD, TE, T7, Te, Tf, TV, TW, TX, Tv, Ty, TH, To; Chris@10: E Tr, TG, TS, TT, TU; Chris@10: { Chris@10: E TJ, Tg, Th, TK; Chris@10: TJ = R0[0]; Chris@10: Tg = R0[WS(rs, 5)]; Chris@10: Th = R1[WS(rs, 2)]; Chris@10: TK = Th + Tg; Chris@10: Ti = Tg - Th; Chris@10: TR = TJ + TK; Chris@10: TL = FNMS(KP500000000, TK, TJ); Chris@10: } Chris@10: { Chris@10: E Tm, Tt, Tw, Tp, T3, Tx, Ta, Tn, Td, Tq, T6, Tu; Chris@10: Tm = R1[WS(rs, 1)]; Chris@10: Tt = R0[WS(rs, 3)]; Chris@10: Tw = R1[WS(rs, 4)]; Chris@10: Tp = R0[WS(rs, 6)]; Chris@10: { Chris@10: E T1, T2, T8, T9; Chris@10: T1 = R0[WS(rs, 7)]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = T1 - T2; Chris@10: Tx = T1 + T2; Chris@10: T8 = R1[WS(rs, 6)]; Chris@10: T9 = R0[WS(rs, 4)]; Chris@10: Ta = T8 - T9; Chris@10: Tn = T9 + T8; Chris@10: } Chris@10: { Chris@10: E Tb, Tc, T4, T5; Chris@10: Tb = R1[WS(rs, 3)]; Chris@10: Tc = R0[WS(rs, 1)]; Chris@10: Td = Tb - Tc; Chris@10: Tq = Tc + Tb; Chris@10: T4 = R1[0]; Chris@10: T5 = R1[WS(rs, 5)]; Chris@10: T6 = T4 - T5; Chris@10: Tu = T5 + T4; Chris@10: } Chris@10: TD = Ta - Td; Chris@10: TE = T6 + T3; Chris@10: T7 = T3 - T6; Chris@10: Te = Ta + Td; Chris@10: Tf = T7 - Te; Chris@10: TV = Tt + Tu; Chris@10: TW = Tw + Tx; Chris@10: TX = TV + TW; Chris@10: Tv = FNMS(KP500000000, Tu, Tt); Chris@10: Ty = FNMS(KP500000000, Tx, Tw); Chris@10: TH = Tv + Ty; Chris@10: To = FNMS(KP500000000, Tn, Tm); Chris@10: Tr = FNMS(KP500000000, Tq, Tp); Chris@10: TG = To + Tr; Chris@10: TS = Tm + Tn; Chris@10: TT = Tp + Tq; Chris@10: TU = TS + TT; Chris@10: } Chris@10: Ci[WS(csi, 5)] = KP866025403 * (Tf - Ti); Chris@10: { Chris@10: E TF, TP, TI, TM, TN, TQ, TO; Chris@10: TF = FMA(KP823639103, TD, KP509036960 * TE); Chris@10: TP = FNMS(KP509036960, TD, KP823639103 * TE); Chris@10: TI = KP559016994 * (TG - TH); Chris@10: TM = TG + TH; Chris@10: TN = FNMS(KP250000000, TM, TL); Chris@10: Cr[WS(csr, 5)] = TL + TM; Chris@10: TQ = TN - TI; Chris@10: Cr[WS(csr, 2)] = TP + TQ; Chris@10: Cr[WS(csr, 7)] = TQ - TP; Chris@10: TO = TI + TN; Chris@10: Cr[WS(csr, 1)] = TF + TO; Chris@10: Cr[WS(csr, 4)] = TO - TF; Chris@10: } Chris@10: { Chris@10: E T11, T12, T10, TY, TZ; Chris@10: T11 = TS - TT; Chris@10: T12 = TW - TV; Chris@10: Ci[WS(csi, 3)] = FMA(KP587785252, T11, KP951056516 * T12); Chris@10: Ci[WS(csi, 6)] = FNMS(KP951056516, T11, KP587785252 * T12); Chris@10: T10 = KP559016994 * (TU - TX); Chris@10: TY = TU + TX; Chris@10: TZ = FNMS(KP250000000, TY, TR); Chris@10: Cr[WS(csr, 3)] = TZ - T10; Chris@10: Cr[0] = TR + TY; Chris@10: Cr[WS(csr, 6)] = T10 + TZ; Chris@10: { Chris@10: E Tl, TB, TA, TC; Chris@10: { Chris@10: E Tj, Tk, Ts, Tz; Chris@10: Tj = FMA(KP866025403, Ti, KP216506350 * Tf); Chris@10: Tk = KP484122918 * (Te + T7); Chris@10: Tl = Tj + Tk; Chris@10: TB = Tk - Tj; Chris@10: Ts = To - Tr; Chris@10: Tz = Tv - Ty; Chris@10: TA = FMA(KP951056516, Ts, KP587785252 * Tz); Chris@10: TC = FNMS(KP587785252, Ts, KP951056516 * Tz); Chris@10: } Chris@10: Ci[WS(csi, 1)] = Tl - TA; Chris@10: Ci[WS(csi, 7)] = TC - TB; Chris@10: Ci[WS(csi, 4)] = Tl + TA; Chris@10: Ci[WS(csi, 2)] = TB + TC; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cf_15", {50, 11, 14, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_15, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */