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 07:40:15 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 16 -name r2cfII_16 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 66 FP additions, 48 FP multiplications, Chris@10: * (or, 18 additions, 0 multiplications, 48 fused multiply/add), Chris@10: * 54 stack variables, 7 constants, and 32 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_16(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(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@10: DK(KP198912367, +0.198912367379658006911597622644676228597850501); Chris@10: DK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@10: DK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@10: DK(KP668178637, +0.668178637919298919997757686523080761552472251); Chris@10: DK(KP414213562, +0.414213562373095048801688724209698078569671875); Chris@10: DK(KP707106781, +0.707106781186547524400844362104849039284835938); 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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { Chris@10: E TN, TF, TX, TV, TO, TP, TY, TM, TQ, TW; Chris@10: { Chris@10: E TT, TZ, TB, T5, Tu, TK, TJ, Tr, T9, TC, T8, Tl, TH, TG, Ti; Chris@10: E Ta; Chris@10: { Chris@10: E T1, TR, Tn, Ts, To, TS, T4, Tp, T2, T3; Chris@10: T1 = R0[0]; Chris@10: TR = R0[WS(rs, 4)]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 6)]; Chris@10: Tn = R1[WS(rs, 7)]; Chris@10: Ts = R1[WS(rs, 3)]; Chris@10: To = R1[WS(rs, 1)]; Chris@10: TS = T2 + T3; Chris@10: T4 = T2 - T3; Chris@10: Tp = R1[WS(rs, 5)]; Chris@10: { Chris@10: E Te, Tj, Tf, Tg, Tt, Tq; Chris@10: Te = R1[0]; Chris@10: TT = FMA(KP707106781, TS, TR); Chris@10: TZ = FNMS(KP707106781, TS, TR); Chris@10: TB = FMA(KP707106781, T4, T1); Chris@10: T5 = FNMS(KP707106781, T4, T1); Chris@10: Tt = To + Tp; Chris@10: Tq = To - Tp; Chris@10: Tj = R1[WS(rs, 4)]; Chris@10: Tf = R1[WS(rs, 2)]; Chris@10: Tu = FNMS(KP707106781, Tt, Ts); Chris@10: TK = FMA(KP707106781, Tt, Ts); Chris@10: TJ = FMS(KP707106781, Tq, Tn); Chris@10: Tr = FMA(KP707106781, Tq, Tn); Chris@10: Tg = R1[WS(rs, 6)]; Chris@10: { Chris@10: E T6, T7, Tk, Th; Chris@10: T6 = R0[WS(rs, 5)]; Chris@10: T7 = R0[WS(rs, 1)]; Chris@10: T9 = R0[WS(rs, 3)]; Chris@10: Tk = Tf + Tg; Chris@10: Th = Tf - Tg; Chris@10: TC = FNMS(KP414213562, T6, T7); Chris@10: T8 = FMA(KP414213562, T7, T6); Chris@10: Tl = FNMS(KP707106781, Tk, Tj); Chris@10: TH = FMA(KP707106781, Tk, Tj); Chris@10: TG = FMA(KP707106781, Th, Te); Chris@10: Ti = FNMS(KP707106781, Th, Te); Chris@10: Ta = R0[WS(rs, 7)]; Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E TE, TU, Ty, Tv, TI, TL; Chris@10: Ty = FNMS(KP668178637, Tr, Tu); Chris@10: Tv = FMA(KP668178637, Tu, Tr); Chris@10: { Chris@10: E Tw, T14, T12, TA, T11, T13, Tx, Td; Chris@10: { Chris@10: E Tz, Tm, TD, Tb, T10, Tc; Chris@10: Tz = FNMS(KP668178637, Ti, Tl); Chris@10: Tm = FMA(KP668178637, Tl, Ti); Chris@10: TD = FMS(KP414213562, T9, Ta); Chris@10: Tb = FMA(KP414213562, Ta, T9); Chris@10: Tw = Tm - Tv; Chris@10: T14 = Tm + Tv; Chris@10: T10 = TD - TC; Chris@10: TE = TC + TD; Chris@10: Tc = T8 - Tb; Chris@10: TU = T8 + Tb; Chris@10: T12 = Tz + Ty; Chris@10: TA = Ty - Tz; Chris@10: T11 = FMA(KP923879532, T10, TZ); Chris@10: T13 = FNMS(KP923879532, T10, TZ); Chris@10: Tx = FNMS(KP923879532, Tc, T5); Chris@10: Td = FMA(KP923879532, Tc, T5); Chris@10: } Chris@10: Ci[WS(csi, 2)] = -(FMA(KP831469612, T14, T13)); Chris@10: Ci[WS(csi, 5)] = FNMS(KP831469612, T14, T13); Chris@10: Cr[WS(csr, 1)] = FMA(KP831469612, Tw, Td); Chris@10: Cr[WS(csr, 6)] = FNMS(KP831469612, Tw, Td); Chris@10: Cr[WS(csr, 5)] = FNMS(KP831469612, TA, Tx); Chris@10: Ci[WS(csi, 1)] = FMA(KP831469612, T12, T11); Chris@10: Cr[WS(csr, 2)] = FMA(KP831469612, TA, Tx); Chris@10: Ci[WS(csi, 6)] = FMS(KP831469612, T12, T11); Chris@10: } Chris@10: TN = FNMS(KP923879532, TE, TB); Chris@10: TF = FMA(KP923879532, TE, TB); Chris@10: TX = FNMS(KP923879532, TU, TT); Chris@10: TV = FMA(KP923879532, TU, TT); Chris@10: TO = FMA(KP198912367, TG, TH); Chris@10: TI = FNMS(KP198912367, TH, TG); Chris@10: TL = FMA(KP198912367, TK, TJ); Chris@10: TP = FNMS(KP198912367, TJ, TK); Chris@10: TY = TL - TI; Chris@10: TM = TI + TL; Chris@10: } Chris@10: } Chris@10: Ci[WS(csi, 4)] = FMS(KP980785280, TY, TX); Chris@10: Ci[WS(csi, 3)] = FMA(KP980785280, TY, TX); Chris@10: Cr[0] = FMA(KP980785280, TM, TF); Chris@10: Cr[WS(csr, 7)] = FNMS(KP980785280, TM, TF); Chris@10: TQ = TO - TP; Chris@10: TW = TO + TP; Chris@10: Ci[0] = -(FMA(KP980785280, TW, TV)); Chris@10: Ci[WS(csi, 7)] = FNMS(KP980785280, TW, TV); Chris@10: Cr[WS(csr, 3)] = FMA(KP980785280, TQ, TN); Chris@10: Cr[WS(csr, 4)] = FNMS(KP980785280, TQ, TN); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 16, "r2cfII_16", {18, 0, 48, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_16) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_16, &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 16 -name r2cfII_16 -dft-II -include r2cfII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 66 FP additions, 30 FP multiplications, Chris@10: * (or, 54 additions, 18 multiplications, 12 fused multiply/add), Chris@10: * 32 stack variables, 7 constants, and 32 memory accesses Chris@10: */ Chris@10: #include "r2cfII.h" Chris@10: Chris@10: static void r2cfII_16(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(KP555570233, +0.555570233019602224742830813948532874374937191); Chris@10: DK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@10: DK(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@10: DK(KP195090322, +0.195090322016128267848284868477022240927691618); Chris@10: DK(KP382683432, +0.382683432365089771728459984030398866761344562); Chris@10: DK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@10: DK(KP707106781, +0.707106781186547524400844362104849039284835938); 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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { Chris@10: E T5, T11, TB, TV, Tr, TK, Tu, TJ, Ti, TH, Tl, TG, Tc, T10, TE; Chris@10: E TS; Chris@10: { Chris@10: E T1, TU, T4, TT, T2, T3; Chris@10: T1 = R0[0]; Chris@10: TU = R0[WS(rs, 4)]; Chris@10: T2 = R0[WS(rs, 2)]; Chris@10: T3 = R0[WS(rs, 6)]; Chris@10: T4 = KP707106781 * (T2 - T3); Chris@10: TT = KP707106781 * (T2 + T3); Chris@10: T5 = T1 + T4; Chris@10: T11 = TU - TT; Chris@10: TB = T1 - T4; Chris@10: TV = TT + TU; Chris@10: } Chris@10: { Chris@10: E Tq, Tt, Tp, Ts, Tn, To; Chris@10: Tq = R1[WS(rs, 7)]; Chris@10: Tt = R1[WS(rs, 3)]; Chris@10: Tn = R1[WS(rs, 1)]; Chris@10: To = R1[WS(rs, 5)]; Chris@10: Tp = KP707106781 * (Tn - To); Chris@10: Ts = KP707106781 * (Tn + To); Chris@10: Tr = Tp - Tq; Chris@10: TK = Tt - Ts; Chris@10: Tu = Ts + Tt; Chris@10: TJ = Tp + Tq; Chris@10: } Chris@10: { Chris@10: E Te, Tk, Th, Tj, Tf, Tg; Chris@10: Te = R1[0]; Chris@10: Tk = R1[WS(rs, 4)]; Chris@10: Tf = R1[WS(rs, 2)]; Chris@10: Tg = R1[WS(rs, 6)]; Chris@10: Th = KP707106781 * (Tf - Tg); Chris@10: Tj = KP707106781 * (Tf + Tg); Chris@10: Ti = Te + Th; Chris@10: TH = Tk - Tj; Chris@10: Tl = Tj + Tk; Chris@10: TG = Te - Th; Chris@10: } Chris@10: { Chris@10: E T8, TC, Tb, TD; Chris@10: { Chris@10: E T6, T7, T9, Ta; Chris@10: T6 = R0[WS(rs, 1)]; Chris@10: T7 = R0[WS(rs, 5)]; Chris@10: T8 = FNMS(KP382683432, T7, KP923879532 * T6); Chris@10: TC = FMA(KP382683432, T6, KP923879532 * T7); Chris@10: T9 = R0[WS(rs, 3)]; Chris@10: Ta = R0[WS(rs, 7)]; Chris@10: Tb = FNMS(KP923879532, Ta, KP382683432 * T9); Chris@10: TD = FMA(KP923879532, T9, KP382683432 * Ta); Chris@10: } Chris@10: Tc = T8 + Tb; Chris@10: T10 = Tb - T8; Chris@10: TE = TC - TD; Chris@10: TS = TC + TD; Chris@10: } Chris@10: { Chris@10: E Td, TW, Tw, TR, Tm, Tv; Chris@10: Td = T5 - Tc; Chris@10: TW = TS + TV; Chris@10: Tm = FMA(KP195090322, Ti, KP980785280 * Tl); Chris@10: Tv = FNMS(KP980785280, Tu, KP195090322 * Tr); Chris@10: Tw = Tm + Tv; Chris@10: TR = Tv - Tm; Chris@10: Cr[WS(csr, 4)] = Td - Tw; Chris@10: Ci[WS(csi, 7)] = TR + TW; Chris@10: Cr[WS(csr, 3)] = Td + Tw; Chris@10: Ci[0] = TR - TW; Chris@10: } Chris@10: { Chris@10: E Tx, TY, TA, TX, Ty, Tz; Chris@10: Tx = T5 + Tc; Chris@10: TY = TV - TS; Chris@10: Ty = FNMS(KP195090322, Tl, KP980785280 * Ti); Chris@10: Tz = FMA(KP980785280, Tr, KP195090322 * Tu); Chris@10: TA = Ty + Tz; Chris@10: TX = Tz - Ty; Chris@10: Cr[WS(csr, 7)] = Tx - TA; Chris@10: Ci[WS(csi, 3)] = TX + TY; Chris@10: Cr[0] = Tx + TA; Chris@10: Ci[WS(csi, 4)] = TX - TY; Chris@10: } Chris@10: { Chris@10: E TF, T12, TM, TZ, TI, TL; Chris@10: TF = TB + TE; Chris@10: T12 = T10 - T11; Chris@10: TI = FMA(KP831469612, TG, KP555570233 * TH); Chris@10: TL = FMA(KP831469612, TJ, KP555570233 * TK); Chris@10: TM = TI - TL; Chris@10: TZ = TI + TL; Chris@10: Cr[WS(csr, 6)] = TF - TM; Chris@10: Ci[WS(csi, 2)] = T12 - TZ; Chris@10: Cr[WS(csr, 1)] = TF + TM; Chris@10: Ci[WS(csi, 5)] = -(TZ + T12); Chris@10: } Chris@10: { Chris@10: E TN, T14, TQ, T13, TO, TP; Chris@10: TN = TB - TE; Chris@10: T14 = T10 + T11; Chris@10: TO = FNMS(KP555570233, TJ, KP831469612 * TK); Chris@10: TP = FNMS(KP555570233, TG, KP831469612 * TH); Chris@10: TQ = TO - TP; Chris@10: T13 = TP + TO; Chris@10: Cr[WS(csr, 5)] = TN - TQ; Chris@10: Ci[WS(csi, 1)] = T13 + T14; Chris@10: Cr[WS(csr, 2)] = TN + TQ; Chris@10: Ci[WS(csi, 6)] = T13 - T14; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 16, "r2cfII_16", {54, 18, 12, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cfII_16) (planner *p) { Chris@10: X(kr2c_register) (p, r2cfII_16, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */