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 07:35:43 EST 2012 */ Chris@10: Chris@10: #include "codelet-dft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 10 -name n1_10 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 84 FP additions, 36 FP multiplications, Chris@10: * (or, 48 additions, 0 multiplications, 36 fused multiply/add), Chris@10: * 59 stack variables, 4 constants, and 40 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) { Chris@10: E T1g, T1a, T18, T1m, T1k, T1f, T19, T11, T1h, T1l; Chris@10: { Chris@10: E Tj, T3, T1b, TN, T1j, TU, T1i, TV, Tq, T10, Ti, Ts, Tw, T15, Tx; Chris@10: E T13, TG, Ty, TB, TC; Chris@10: { Chris@10: E T1, T2, TL, TM; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 5)]; Chris@10: TL = ii[0]; Chris@10: TM = ii[WS(is, 5)]; Chris@10: { Chris@10: E T7, Tk, T6, To, Tg, T8, Tb, Tc; Chris@10: { Chris@10: E T4, T5, Te, Tf; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: Tj = T1 + T2; Chris@10: T3 = T1 - T2; Chris@10: T1b = TL + TM; Chris@10: TN = TL - TM; Chris@10: T5 = ri[WS(is, 7)]; Chris@10: Te = ri[WS(is, 6)]; Chris@10: Tf = ri[WS(is, 1)]; Chris@10: T7 = ri[WS(is, 8)]; Chris@10: Tk = T4 + T5; Chris@10: T6 = T4 - T5; Chris@10: To = Te + Tf; Chris@10: Tg = Te - Tf; Chris@10: T8 = ri[WS(is, 3)]; Chris@10: Tb = ri[WS(is, 4)]; Chris@10: Tc = ri[WS(is, 9)]; Chris@10: } Chris@10: { Chris@10: E TE, TF, Tu, Tv; Chris@10: { Chris@10: E Ta, Th, Tl, T9; Chris@10: Tu = ii[WS(is, 2)]; Chris@10: Tl = T7 + T8; Chris@10: T9 = T7 - T8; Chris@10: { Chris@10: E Tn, Td, Tm, Tp; Chris@10: Tn = Tb + Tc; Chris@10: Td = Tb - Tc; Chris@10: Tm = Tk + Tl; Chris@10: T1j = Tk - Tl; Chris@10: Ta = T6 + T9; Chris@10: TU = T6 - T9; Chris@10: Tp = Tn + To; Chris@10: T1i = Tn - To; Chris@10: Th = Td + Tg; Chris@10: TV = Td - Tg; Chris@10: Tq = Tm + Tp; Chris@10: T10 = Tm - Tp; Chris@10: Tv = ii[WS(is, 7)]; Chris@10: } Chris@10: Ti = Ta + Th; Chris@10: Ts = Ta - Th; Chris@10: } Chris@10: TE = ii[WS(is, 6)]; Chris@10: TF = ii[WS(is, 1)]; Chris@10: Tw = Tu - Tv; Chris@10: T15 = Tu + Tv; Chris@10: Tx = ii[WS(is, 8)]; Chris@10: T13 = TE + TF; Chris@10: TG = TE - TF; Chris@10: Ty = ii[WS(is, 3)]; Chris@10: TB = ii[WS(is, 4)]; Chris@10: TC = ii[WS(is, 9)]; Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E T17, TA, T14, TH, T1e, TQ, TS; Chris@10: { Chris@10: E TO, TP, T16, Tz; Chris@10: ro[WS(os, 5)] = T3 + Ti; Chris@10: T16 = Tx + Ty; Chris@10: Tz = Tx - Ty; Chris@10: { Chris@10: E T12, TD, T1c, T1d; Chris@10: T12 = TB + TC; Chris@10: TD = TB - TC; Chris@10: T1c = T15 + T16; Chris@10: T17 = T15 - T16; Chris@10: TO = Tw + Tz; Chris@10: TA = Tw - Tz; Chris@10: T1d = T12 + T13; Chris@10: T14 = T12 - T13; Chris@10: TP = TD + TG; Chris@10: TH = TD - TG; Chris@10: T1e = T1c + T1d; Chris@10: T1g = T1c - T1d; Chris@10: } Chris@10: ro[0] = Tj + Tq; Chris@10: TQ = TO + TP; Chris@10: TS = TO - TP; Chris@10: } Chris@10: { Chris@10: E TK, TI, TY, TW, TR, TJ, Tt, Tr, TZ, TX, TT; Chris@10: TK = FNMS(KP618033988, TA, TH); Chris@10: TI = FMA(KP618033988, TH, TA); Chris@10: io[0] = T1b + T1e; Chris@10: io[WS(os, 5)] = TN + TQ; Chris@10: Tr = FNMS(KP250000000, Ti, T3); Chris@10: TY = FNMS(KP618033988, TU, TV); Chris@10: TW = FMA(KP618033988, TV, TU); Chris@10: TR = FNMS(KP250000000, TQ, TN); Chris@10: TJ = FNMS(KP559016994, Ts, Tr); Chris@10: Tt = FMA(KP559016994, Ts, Tr); Chris@10: T1a = FMA(KP618033988, T14, T17); Chris@10: T18 = FNMS(KP618033988, T17, T14); Chris@10: ro[WS(os, 7)] = FNMS(KP951056516, TK, TJ); Chris@10: ro[WS(os, 3)] = FMA(KP951056516, TK, TJ); Chris@10: ro[WS(os, 1)] = FMA(KP951056516, TI, Tt); Chris@10: ro[WS(os, 9)] = FNMS(KP951056516, TI, Tt); Chris@10: TX = FNMS(KP559016994, TS, TR); Chris@10: TT = FMA(KP559016994, TS, TR); Chris@10: TZ = FNMS(KP250000000, Tq, Tj); Chris@10: io[WS(os, 3)] = FNMS(KP951056516, TY, TX); Chris@10: io[WS(os, 7)] = FMA(KP951056516, TY, TX); Chris@10: io[WS(os, 9)] = FMA(KP951056516, TW, TT); Chris@10: io[WS(os, 1)] = FNMS(KP951056516, TW, TT); Chris@10: T1m = FMA(KP618033988, T1i, T1j); Chris@10: T1k = FNMS(KP618033988, T1j, T1i); Chris@10: T1f = FNMS(KP250000000, T1e, T1b); Chris@10: T19 = FMA(KP559016994, T10, TZ); Chris@10: T11 = FNMS(KP559016994, T10, TZ); Chris@10: } Chris@10: } Chris@10: } Chris@10: ro[WS(os, 4)] = FNMS(KP951056516, T1a, T19); Chris@10: ro[WS(os, 6)] = FMA(KP951056516, T1a, T19); Chris@10: ro[WS(os, 8)] = FMA(KP951056516, T18, T11); Chris@10: ro[WS(os, 2)] = FNMS(KP951056516, T18, T11); Chris@10: T1h = FNMS(KP559016994, T1g, T1f); Chris@10: T1l = FMA(KP559016994, T1g, T1f); Chris@10: io[WS(os, 4)] = FMA(KP951056516, T1m, T1l); Chris@10: io[WS(os, 6)] = FNMS(KP951056516, T1m, T1l); Chris@10: io[WS(os, 8)] = FNMS(KP951056516, T1k, T1h); Chris@10: io[WS(os, 2)] = FMA(KP951056516, T1k, T1h); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 10, "n1_10", {48, 0, 36, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_10) (planner *p) { Chris@10: X(kdft_register) (p, n1_10, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 10 -name n1_10 -include n.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 84 FP additions, 24 FP multiplications, Chris@10: * (or, 72 additions, 12 multiplications, 12 fused multiply/add), Chris@10: * 41 stack variables, 4 constants, and 40 memory accesses Chris@10: */ Chris@10: #include "n.h" Chris@10: Chris@10: static void n1_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) { Chris@10: E T3, Tj, TQ, T1e, TU, TV, T1c, T1b, Tm, Tp, Tq, Ta, Th, Ti, TA; Chris@10: E TH, T17, T14, T1f, T1g, T1h, TL, TM, TR; Chris@10: { Chris@10: E T1, T2, TO, TP; Chris@10: T1 = ri[0]; Chris@10: T2 = ri[WS(is, 5)]; Chris@10: T3 = T1 - T2; Chris@10: Tj = T1 + T2; Chris@10: TO = ii[0]; Chris@10: TP = ii[WS(is, 5)]; Chris@10: TQ = TO - TP; Chris@10: T1e = TO + TP; Chris@10: } Chris@10: { Chris@10: E T6, Tk, Tg, To, T9, Tl, Td, Tn; Chris@10: { Chris@10: E T4, T5, Te, Tf; Chris@10: T4 = ri[WS(is, 2)]; Chris@10: T5 = ri[WS(is, 7)]; Chris@10: T6 = T4 - T5; Chris@10: Tk = T4 + T5; Chris@10: Te = ri[WS(is, 6)]; Chris@10: Tf = ri[WS(is, 1)]; Chris@10: Tg = Te - Tf; Chris@10: To = Te + Tf; Chris@10: } Chris@10: { Chris@10: E T7, T8, Tb, Tc; Chris@10: T7 = ri[WS(is, 8)]; Chris@10: T8 = ri[WS(is, 3)]; Chris@10: T9 = T7 - T8; Chris@10: Tl = T7 + T8; Chris@10: Tb = ri[WS(is, 4)]; Chris@10: Tc = ri[WS(is, 9)]; Chris@10: Td = Tb - Tc; Chris@10: Tn = Tb + Tc; Chris@10: } Chris@10: TU = T6 - T9; Chris@10: TV = Td - Tg; Chris@10: T1c = Tk - Tl; Chris@10: T1b = Tn - To; Chris@10: Tm = Tk + Tl; Chris@10: Tp = Tn + To; Chris@10: Tq = Tm + Tp; Chris@10: Ta = T6 + T9; Chris@10: Th = Td + Tg; Chris@10: Ti = Ta + Th; Chris@10: } Chris@10: { Chris@10: E Tw, T15, TG, T13, Tz, T16, TD, T12; Chris@10: { Chris@10: E Tu, Tv, TE, TF; Chris@10: Tu = ii[WS(is, 2)]; Chris@10: Tv = ii[WS(is, 7)]; Chris@10: Tw = Tu - Tv; Chris@10: T15 = Tu + Tv; Chris@10: TE = ii[WS(is, 6)]; Chris@10: TF = ii[WS(is, 1)]; Chris@10: TG = TE - TF; Chris@10: T13 = TE + TF; Chris@10: } Chris@10: { Chris@10: E Tx, Ty, TB, TC; Chris@10: Tx = ii[WS(is, 8)]; Chris@10: Ty = ii[WS(is, 3)]; Chris@10: Tz = Tx - Ty; Chris@10: T16 = Tx + Ty; Chris@10: TB = ii[WS(is, 4)]; Chris@10: TC = ii[WS(is, 9)]; Chris@10: TD = TB - TC; Chris@10: T12 = TB + TC; Chris@10: } Chris@10: TA = Tw - Tz; Chris@10: TH = TD - TG; Chris@10: T17 = T15 - T16; Chris@10: T14 = T12 - T13; Chris@10: T1f = T15 + T16; Chris@10: T1g = T12 + T13; Chris@10: T1h = T1f + T1g; Chris@10: TL = Tw + Tz; Chris@10: TM = TD + TG; Chris@10: TR = TL + TM; Chris@10: } Chris@10: ro[WS(os, 5)] = T3 + Ti; Chris@10: io[WS(os, 5)] = TQ + TR; Chris@10: ro[0] = Tj + Tq; Chris@10: io[0] = T1e + T1h; Chris@10: { Chris@10: E TI, TK, Tt, TJ, Tr, Ts; Chris@10: TI = FMA(KP951056516, TA, KP587785252 * TH); Chris@10: TK = FNMS(KP587785252, TA, KP951056516 * TH); Chris@10: Tr = KP559016994 * (Ta - Th); Chris@10: Ts = FNMS(KP250000000, Ti, T3); Chris@10: Tt = Tr + Ts; Chris@10: TJ = Ts - Tr; Chris@10: ro[WS(os, 9)] = Tt - TI; Chris@10: ro[WS(os, 3)] = TJ + TK; Chris@10: ro[WS(os, 1)] = Tt + TI; Chris@10: ro[WS(os, 7)] = TJ - TK; Chris@10: } Chris@10: { Chris@10: E TW, TY, TT, TX, TN, TS; Chris@10: TW = FMA(KP951056516, TU, KP587785252 * TV); Chris@10: TY = FNMS(KP587785252, TU, KP951056516 * TV); Chris@10: TN = KP559016994 * (TL - TM); Chris@10: TS = FNMS(KP250000000, TR, TQ); Chris@10: TT = TN + TS; Chris@10: TX = TS - TN; Chris@10: io[WS(os, 1)] = TT - TW; Chris@10: io[WS(os, 7)] = TY + TX; Chris@10: io[WS(os, 9)] = TW + TT; Chris@10: io[WS(os, 3)] = TX - TY; Chris@10: } Chris@10: { Chris@10: E T18, T1a, T11, T19, TZ, T10; Chris@10: T18 = FNMS(KP587785252, T17, KP951056516 * T14); Chris@10: T1a = FMA(KP951056516, T17, KP587785252 * T14); Chris@10: TZ = FNMS(KP250000000, Tq, Tj); Chris@10: T10 = KP559016994 * (Tm - Tp); Chris@10: T11 = TZ - T10; Chris@10: T19 = T10 + TZ; Chris@10: ro[WS(os, 2)] = T11 - T18; Chris@10: ro[WS(os, 6)] = T19 + T1a; Chris@10: ro[WS(os, 8)] = T11 + T18; Chris@10: ro[WS(os, 4)] = T19 - T1a; Chris@10: } Chris@10: { Chris@10: E T1d, T1l, T1k, T1m, T1i, T1j; Chris@10: T1d = FNMS(KP587785252, T1c, KP951056516 * T1b); Chris@10: T1l = FMA(KP951056516, T1c, KP587785252 * T1b); Chris@10: T1i = FNMS(KP250000000, T1h, T1e); Chris@10: T1j = KP559016994 * (T1f - T1g); Chris@10: T1k = T1i - T1j; Chris@10: T1m = T1j + T1i; Chris@10: io[WS(os, 2)] = T1d + T1k; Chris@10: io[WS(os, 6)] = T1m - T1l; Chris@10: io[WS(os, 8)] = T1k - T1d; Chris@10: io[WS(os, 4)] = T1l + T1m; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 10, "n1_10", {72, 12, 12, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void X(codelet_n1_10) (planner *p) { Chris@10: X(kdft_register) (p, n1_10, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */