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 07:41:08 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_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cb_15 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 43 FP multiplications, Chris@10: * (or, 21 additions, 0 multiplications, 43 fused multiply/add), Chris@10: * 54 stack variables, 9 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E TL, Tz, TM, TK; Chris@10: { Chris@10: E T3, Th, Tt, TD, TI, TH, TY, TC, TZ, Tu, Tm, Tv, Tr, Te, TW; Chris@10: E Tg, T1, T2, T12, T10, TV; Chris@10: Tg = Ci[WS(csi, 5)]; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 5)]; Chris@10: { Chris@10: E T4, TA, T9, TF, T7, Tj, Tc, Tk, TG, Tq, Tf, Tl, TB; Chris@10: T4 = Cr[WS(csr, 3)]; Chris@10: TA = Ci[WS(csi, 3)]; Chris@10: T9 = Cr[WS(csr, 6)]; Chris@10: Tf = T1 - T2; Chris@10: T3 = FMA(KP2_000000000, T2, T1); Chris@10: TF = Ci[WS(csi, 6)]; Chris@10: { Chris@10: E Ta, Tb, T5, T6, To, Tp; Chris@10: T5 = Cr[WS(csr, 7)]; Chris@10: T6 = Cr[WS(csr, 2)]; Chris@10: Th = FMA(KP1_732050807, Tg, Tf); Chris@10: Tt = FNMS(KP1_732050807, Tg, Tf); Chris@10: Ta = Cr[WS(csr, 4)]; Chris@10: TD = T5 - T6; Chris@10: T7 = T5 + T6; Chris@10: Tb = Cr[WS(csr, 1)]; Chris@10: To = Ci[WS(csi, 4)]; Chris@10: Tp = Ci[WS(csi, 1)]; Chris@10: Tj = Ci[WS(csi, 7)]; Chris@10: Tc = Ta + Tb; Chris@10: TI = Ta - Tb; Chris@10: Tk = Ci[WS(csi, 2)]; Chris@10: TG = Tp - To; Chris@10: Tq = To + Tp; Chris@10: } Chris@10: Tl = Tj - Tk; Chris@10: TB = Tj + Tk; Chris@10: TH = FNMS(KP500000000, TG, TF); Chris@10: TY = TG + TF; Chris@10: TC = FMA(KP500000000, TB, TA); Chris@10: TZ = TA - TB; Chris@10: { Chris@10: E Ti, T8, Td, Tn; Chris@10: Ti = FNMS(KP2_000000000, T4, T7); Chris@10: T8 = T4 + T7; Chris@10: Td = T9 + Tc; Chris@10: Tn = FNMS(KP2_000000000, T9, Tc); Chris@10: Tu = FNMS(KP1_732050807, Tl, Ti); Chris@10: Tm = FMA(KP1_732050807, Tl, Ti); Chris@10: Tv = FNMS(KP1_732050807, Tq, Tn); Chris@10: Tr = FMA(KP1_732050807, Tq, Tn); Chris@10: Te = T8 + Td; Chris@10: TW = T8 - Td; Chris@10: } Chris@10: } Chris@10: T12 = FMA(KP618033988, TY, TZ); Chris@10: T10 = FNMS(KP618033988, TZ, TY); Chris@10: TV = FNMS(KP500000000, Te, T3); Chris@10: R0[0] = FMA(KP2_000000000, Te, T3); Chris@10: { Chris@10: E TJ, TE, TT, TP, TU, TS, Ty, Tw, Tx; Chris@10: { Chris@10: E TO, Ts, TQ, TN, TR, T11, TX; Chris@10: TO = Tr - Tm; Chris@10: Ts = Tm + Tr; Chris@10: T11 = FMA(KP1_118033988, TW, TV); Chris@10: TX = FNMS(KP1_118033988, TW, TV); Chris@10: TQ = FNMS(KP866025403, TI, TH); Chris@10: TJ = FMA(KP866025403, TI, TH); Chris@10: TN = FMA(KP250000000, Ts, Th); Chris@10: R0[WS(rs, 3)] = FNMS(KP1_902113032, T12, T11); Chris@10: R1[WS(rs, 4)] = FMA(KP1_902113032, T12, T11); Chris@10: R0[WS(rs, 6)] = FMA(KP1_902113032, T10, TX); Chris@10: R1[WS(rs, 1)] = FNMS(KP1_902113032, T10, TX); Chris@10: TR = FNMS(KP866025403, TD, TC); Chris@10: TE = FMA(KP866025403, TD, TC); Chris@10: R1[WS(rs, 2)] = Th - Ts; Chris@10: TT = FMA(KP559016994, TO, TN); Chris@10: TP = FNMS(KP559016994, TO, TN); Chris@10: TU = FMA(KP618033988, TQ, TR); Chris@10: TS = FNMS(KP618033988, TR, TQ); Chris@10: } Chris@10: Ty = Tv - Tu; Chris@10: Tw = Tu + Tv; Chris@10: R0[WS(rs, 7)] = FMA(KP1_902113032, TU, TT); Chris@10: R1[WS(rs, 5)] = FNMS(KP1_902113032, TU, TT); Chris@10: R0[WS(rs, 1)] = FMA(KP1_902113032, TS, TP); Chris@10: R0[WS(rs, 4)] = FNMS(KP1_902113032, TS, TP); Chris@10: Tx = FMA(KP250000000, Tw, Tt); Chris@10: R0[WS(rs, 5)] = Tt - Tw; Chris@10: TL = FNMS(KP559016994, Ty, Tx); Chris@10: Tz = FMA(KP559016994, Ty, Tx); Chris@10: TM = FNMS(KP618033988, TE, TJ); Chris@10: TK = FMA(KP618033988, TJ, TE); Chris@10: } Chris@10: } Chris@10: R1[WS(rs, 3)] = FMA(KP1_902113032, TM, TL); Chris@10: R1[WS(rs, 6)] = FNMS(KP1_902113032, TM, TL); Chris@10: R0[WS(rs, 2)] = FMA(KP1_902113032, TK, Tz); Chris@10: R1[0] = FNMS(KP1_902113032, TK, Tz); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cb_15", {21, 0, 43, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_15, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cb_15 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 31 FP multiplications, Chris@10: * (or, 47 additions, 14 multiplications, 17 fused multiply/add), Chris@10: * 44 stack variables, 7 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: 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@10: { Chris@10: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@10: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@10: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: { Chris@10: INT i; Chris@10: 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@10: E T3, Tu, Ti, TB, TZ, T10, TE, TG, TJ, Tn, Tv, Ts, Tw, T8, Td; Chris@10: E Te; Chris@10: { Chris@10: E Th, T1, T2, Tf, Tg; Chris@10: Tg = Ci[WS(csi, 5)]; Chris@10: Th = KP1_732050807 * Tg; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 5)]; Chris@10: Tf = T1 - T2; Chris@10: T3 = FMA(KP2_000000000, T2, T1); Chris@10: Tu = Tf - Th; Chris@10: Ti = Tf + Th; Chris@10: } Chris@10: { Chris@10: E T4, TD, T9, TI, T5, T6, T7, Ta, Tb, Tc, Tr, TH, Tm, TC, Tj; Chris@10: E To; Chris@10: T4 = Cr[WS(csr, 3)]; Chris@10: TD = Ci[WS(csi, 3)]; Chris@10: T9 = Cr[WS(csr, 6)]; Chris@10: TI = Ci[WS(csi, 6)]; Chris@10: T5 = Cr[WS(csr, 7)]; Chris@10: T6 = Cr[WS(csr, 2)]; Chris@10: T7 = T5 + T6; Chris@10: Ta = Cr[WS(csr, 4)]; Chris@10: Tb = Cr[WS(csr, 1)]; Chris@10: Tc = Ta + Tb; Chris@10: { Chris@10: E Tp, Tq, Tk, Tl; Chris@10: Tp = Ci[WS(csi, 4)]; Chris@10: Tq = Ci[WS(csi, 1)]; Chris@10: Tr = KP866025403 * (Tp + Tq); Chris@10: TH = Tp - Tq; Chris@10: Tk = Ci[WS(csi, 7)]; Chris@10: Tl = Ci[WS(csi, 2)]; Chris@10: Tm = KP866025403 * (Tk - Tl); Chris@10: TC = Tk + Tl; Chris@10: } Chris@10: TB = KP866025403 * (T5 - T6); Chris@10: TZ = TD - TC; Chris@10: T10 = TI - TH; Chris@10: TE = FMA(KP500000000, TC, TD); Chris@10: TG = KP866025403 * (Ta - Tb); Chris@10: TJ = FMA(KP500000000, TH, TI); Chris@10: Tj = FNMS(KP500000000, T7, T4); Chris@10: Tn = Tj - Tm; Chris@10: Tv = Tj + Tm; Chris@10: To = FNMS(KP500000000, Tc, T9); Chris@10: Ts = To - Tr; Chris@10: Tw = To + Tr; Chris@10: T8 = T4 + T7; Chris@10: Td = T9 + Tc; Chris@10: Te = T8 + Td; Chris@10: } Chris@10: R0[0] = FMA(KP2_000000000, Te, T3); Chris@10: { Chris@10: E T11, T13, TY, T12, TW, TX; Chris@10: T11 = FNMS(KP1_902113032, T10, KP1_175570504 * TZ); Chris@10: T13 = FMA(KP1_902113032, TZ, KP1_175570504 * T10); Chris@10: TW = FNMS(KP500000000, Te, T3); Chris@10: TX = KP1_118033988 * (T8 - Td); Chris@10: TY = TW - TX; Chris@10: T12 = TX + TW; Chris@10: R0[WS(rs, 6)] = TY - T11; Chris@10: R1[WS(rs, 4)] = T12 + T13; Chris@10: R1[WS(rs, 1)] = TY + T11; Chris@10: R0[WS(rs, 3)] = T12 - T13; Chris@10: } Chris@10: { Chris@10: E TP, Tt, TO, TT, TV, TR, TS, TU, TQ; Chris@10: TP = KP1_118033988 * (Tn - Ts); Chris@10: Tt = Tn + Ts; Chris@10: TO = FNMS(KP500000000, Tt, Ti); Chris@10: TR = TE - TB; Chris@10: TS = TJ - TG; Chris@10: TT = FNMS(KP1_902113032, TS, KP1_175570504 * TR); Chris@10: TV = FMA(KP1_902113032, TR, KP1_175570504 * TS); Chris@10: R1[WS(rs, 2)] = FMA(KP2_000000000, Tt, Ti); Chris@10: TU = TP + TO; Chris@10: R1[WS(rs, 5)] = TU - TV; Chris@10: R0[WS(rs, 7)] = TU + TV; Chris@10: TQ = TO - TP; Chris@10: R0[WS(rs, 1)] = TQ - TT; Chris@10: R0[WS(rs, 4)] = TQ + TT; Chris@10: } Chris@10: { Chris@10: E Tz, Tx, Ty, TL, TN, TF, TK, TM, TA; Chris@10: Tz = KP1_118033988 * (Tv - Tw); Chris@10: Tx = Tv + Tw; Chris@10: Ty = FNMS(KP500000000, Tx, Tu); Chris@10: TF = TB + TE; Chris@10: TK = TG + TJ; Chris@10: TL = FNMS(KP1_902113032, TK, KP1_175570504 * TF); Chris@10: TN = FMA(KP1_902113032, TF, KP1_175570504 * TK); Chris@10: R0[WS(rs, 5)] = FMA(KP2_000000000, Tx, Tu); Chris@10: TM = Tz + Ty; Chris@10: R1[0] = TM - TN; Chris@10: R0[WS(rs, 2)] = TM + TN; Chris@10: TA = Ty - Tz; Chris@10: R1[WS(rs, 3)] = TA - TL; Chris@10: R1[WS(rs, 6)] = TA + TL; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cb_15", {47, 14, 17, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_15, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */