cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:39:48 EST 2012 */ cannam@95: cannam@95: #include "codelet-rdft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cf_20 -include r2cf.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 86 FP additions, 32 FP multiplications, cannam@95: * (or, 58 additions, 4 multiplications, 28 fused multiply/add), cannam@95: * 70 stack variables, 4 constants, and 40 memory accesses cannam@95: */ cannam@95: #include "r2cf.h" cannam@95: cannam@95: static void r2cf_20(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(80, rs), MAKE_VOLATILE_STRIDE(80, csr), MAKE_VOLATILE_STRIDE(80, csi)) { cannam@95: E T1i, T1c, T1a, T1o, T1m, T1h, T1b, T13, T1j, T1n; cannam@95: { cannam@95: E T3, T1d, TJ, TV, T1k, T16, T19, T1l, Ty, Ti, T12, TD, T1g, TR, TX; cannam@95: E TK, Tt, TU, TW, TL, TE; cannam@95: { cannam@95: E T1, T2, TG, TH; cannam@95: T1 = R0[0]; cannam@95: T2 = R0[WS(rs, 5)]; cannam@95: TG = R1[WS(rs, 2)]; cannam@95: TH = R1[WS(rs, 7)]; cannam@95: { cannam@95: E T6, To, T17, Tx, T18, TC, Tj, T9, Tp, Tu, Td, T15, Tm, Tq, Te; cannam@95: E Tf; cannam@95: { cannam@95: E TA, TB, T7, T8; cannam@95: { cannam@95: E T4, TF, TI, T5, Tv, Tw; cannam@95: T4 = R0[WS(rs, 2)]; cannam@95: T3 = T1 - T2; cannam@95: TF = T1 + T2; cannam@95: T1d = TG - TH; cannam@95: TI = TG + TH; cannam@95: T5 = R0[WS(rs, 7)]; cannam@95: Tv = R1[WS(rs, 6)]; cannam@95: Tw = R1[WS(rs, 1)]; cannam@95: TJ = TF - TI; cannam@95: TV = TF + TI; cannam@95: T6 = T4 - T5; cannam@95: To = T4 + T5; cannam@95: T17 = Tw - Tv; cannam@95: Tx = Tv + Tw; cannam@95: } cannam@95: TA = R1[WS(rs, 8)]; cannam@95: TB = R1[WS(rs, 3)]; cannam@95: T7 = R0[WS(rs, 8)]; cannam@95: T8 = R0[WS(rs, 3)]; cannam@95: { cannam@95: E Tb, Tc, Tk, Tl; cannam@95: Tb = R0[WS(rs, 4)]; cannam@95: T18 = TB - TA; cannam@95: TC = TA + TB; cannam@95: Tj = T7 + T8; cannam@95: T9 = T7 - T8; cannam@95: Tc = R0[WS(rs, 9)]; cannam@95: Tk = R1[0]; cannam@95: Tl = R1[WS(rs, 5)]; cannam@95: Tp = R1[WS(rs, 4)]; cannam@95: Tu = Tb + Tc; cannam@95: Td = Tb - Tc; cannam@95: T15 = Tl - Tk; cannam@95: Tm = Tk + Tl; cannam@95: Tq = R1[WS(rs, 9)]; cannam@95: Te = R0[WS(rs, 6)]; cannam@95: Tf = R0[WS(rs, 1)]; cannam@95: } cannam@95: } cannam@95: { cannam@95: E Ta, Tr, Tz, T1e, T1f, Th, T14, Tg, TP, TQ; cannam@95: Ta = T6 + T9; cannam@95: T1k = T6 - T9; cannam@95: T14 = Tq - Tp; cannam@95: Tr = Tp + Tq; cannam@95: Tz = Te + Tf; cannam@95: Tg = Te - Tf; cannam@95: T16 = T14 - T15; cannam@95: T1e = T14 + T15; cannam@95: T1f = T17 + T18; cannam@95: T19 = T17 - T18; cannam@95: Th = Td + Tg; cannam@95: T1l = Td - Tg; cannam@95: Ty = Tu - Tx; cannam@95: TP = Tu + Tx; cannam@95: Ti = Ta + Th; cannam@95: T12 = Ta - Th; cannam@95: TD = Tz - TC; cannam@95: TQ = Tz + TC; cannam@95: T1g = T1e + T1f; cannam@95: T1i = T1e - T1f; cannam@95: { cannam@95: E TT, Tn, Ts, TS; cannam@95: TT = Tj + Tm; cannam@95: Tn = Tj - Tm; cannam@95: Ts = To - Tr; cannam@95: TS = To + Tr; cannam@95: TR = TP - TQ; cannam@95: TX = TP + TQ; cannam@95: TK = Ts + Tn; cannam@95: Tt = Tn - Ts; cannam@95: TU = TS - TT; cannam@95: TW = TS + TT; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: Cr[WS(csr, 5)] = T3 + Ti; cannam@95: Ci[WS(csi, 5)] = T1g - T1d; cannam@95: TL = Ty + TD; cannam@95: TE = Ty - TD; cannam@95: { cannam@95: E TY, T10, TM, TO, T11, TZ, TN; cannam@95: TY = TW + TX; cannam@95: T10 = TW - TX; cannam@95: Ci[WS(csi, 2)] = KP951056516 * (FMA(KP618033988, Tt, TE)); cannam@95: Ci[WS(csi, 6)] = KP951056516 * (FNMS(KP618033988, TE, Tt)); cannam@95: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP618033988, TR, TU)); cannam@95: Ci[WS(csi, 8)] = -(KP951056516 * (FNMS(KP618033988, TU, TR))); cannam@95: TM = TK + TL; cannam@95: TO = TK - TL; cannam@95: T1c = FNMS(KP618033988, T16, T19); cannam@95: T1a = FMA(KP618033988, T19, T16); cannam@95: Cr[0] = TV + TY; cannam@95: TZ = FNMS(KP250000000, TY, TV); cannam@95: Cr[WS(csr, 10)] = TJ + TM; cannam@95: TN = FNMS(KP250000000, TM, TJ); cannam@95: Cr[WS(csr, 8)] = FNMS(KP559016994, T10, TZ); cannam@95: Cr[WS(csr, 4)] = FMA(KP559016994, T10, TZ); cannam@95: Cr[WS(csr, 6)] = FMA(KP559016994, TO, TN); cannam@95: Cr[WS(csr, 2)] = FNMS(KP559016994, TO, TN); cannam@95: T11 = FNMS(KP250000000, Ti, T3); cannam@95: T1o = FNMS(KP618033988, T1k, T1l); cannam@95: T1m = FMA(KP618033988, T1l, T1k); cannam@95: T1h = FMA(KP250000000, T1g, T1d); cannam@95: T1b = FNMS(KP559016994, T12, T11); cannam@95: T13 = FMA(KP559016994, T12, T11); cannam@95: } cannam@95: } cannam@95: Cr[WS(csr, 3)] = FNMS(KP951056516, T1c, T1b); cannam@95: Cr[WS(csr, 7)] = FMA(KP951056516, T1c, T1b); cannam@95: Cr[WS(csr, 1)] = FMA(KP951056516, T1a, T13); cannam@95: Cr[WS(csr, 9)] = FNMS(KP951056516, T1a, T13); cannam@95: T1j = FNMS(KP559016994, T1i, T1h); cannam@95: T1n = FMA(KP559016994, T1i, T1h); cannam@95: Ci[WS(csi, 3)] = FNMS(KP951056516, T1o, T1n); cannam@95: Ci[WS(csi, 7)] = FMA(KP951056516, T1o, T1n); cannam@95: Ci[WS(csi, 9)] = FMS(KP951056516, T1m, T1j); cannam@95: Ci[WS(csi, 1)] = -(FMA(KP951056516, T1m, T1j)); cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 20, "r2cf_20", {58, 4, 28, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cf_20) (planner *p) { cannam@95: X(kr2c_register) (p, r2cf_20, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cf_20 -include r2cf.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 86 FP additions, 24 FP multiplications, cannam@95: * (or, 74 additions, 12 multiplications, 12 fused multiply/add), cannam@95: * 51 stack variables, 4 constants, and 40 memory accesses cannam@95: */ cannam@95: #include "r2cf.h" cannam@95: cannam@95: static void r2cf_20(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT i; cannam@95: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(80, rs), MAKE_VOLATILE_STRIDE(80, csr), MAKE_VOLATILE_STRIDE(80, csi)) { cannam@95: E T3, T1m, TF, T17, Ts, TM, TN, Tz, Ta, Th, Ti, T1g, T1h, T1k, T10; cannam@95: E T13, T19, TG, TH, TI, T1d, T1e, T1j, TT, TW, T18; cannam@95: { cannam@95: E T1, T2, T15, TD, TE, T16; cannam@95: T1 = R0[0]; cannam@95: T2 = R0[WS(rs, 5)]; cannam@95: T15 = T1 + T2; cannam@95: TD = R1[WS(rs, 7)]; cannam@95: TE = R1[WS(rs, 2)]; cannam@95: T16 = TE + TD; cannam@95: T3 = T1 - T2; cannam@95: T1m = T15 + T16; cannam@95: TF = TD - TE; cannam@95: T17 = T15 - T16; cannam@95: } cannam@95: { cannam@95: E T6, TU, Tv, T12, Ty, TZ, T9, TR, Td, TY, To, TS, Tr, TV, Tg; cannam@95: E T11; cannam@95: { cannam@95: E T4, T5, Tt, Tu; cannam@95: T4 = R0[WS(rs, 2)]; cannam@95: T5 = R0[WS(rs, 7)]; cannam@95: T6 = T4 - T5; cannam@95: TU = T4 + T5; cannam@95: Tt = R1[WS(rs, 8)]; cannam@95: Tu = R1[WS(rs, 3)]; cannam@95: Tv = Tt - Tu; cannam@95: T12 = Tt + Tu; cannam@95: } cannam@95: { cannam@95: E Tw, Tx, T7, T8; cannam@95: Tw = R1[WS(rs, 6)]; cannam@95: Tx = R1[WS(rs, 1)]; cannam@95: Ty = Tw - Tx; cannam@95: TZ = Tw + Tx; cannam@95: T7 = R0[WS(rs, 8)]; cannam@95: T8 = R0[WS(rs, 3)]; cannam@95: T9 = T7 - T8; cannam@95: TR = T7 + T8; cannam@95: } cannam@95: { cannam@95: E Tb, Tc, Tm, Tn; cannam@95: Tb = R0[WS(rs, 4)]; cannam@95: Tc = R0[WS(rs, 9)]; cannam@95: Td = Tb - Tc; cannam@95: TY = Tb + Tc; cannam@95: Tm = R1[0]; cannam@95: Tn = R1[WS(rs, 5)]; cannam@95: To = Tm - Tn; cannam@95: TS = Tm + Tn; cannam@95: } cannam@95: { cannam@95: E Tp, Tq, Te, Tf; cannam@95: Tp = R1[WS(rs, 4)]; cannam@95: Tq = R1[WS(rs, 9)]; cannam@95: Tr = Tp - Tq; cannam@95: TV = Tp + Tq; cannam@95: Te = R0[WS(rs, 6)]; cannam@95: Tf = R0[WS(rs, 1)]; cannam@95: Tg = Te - Tf; cannam@95: T11 = Te + Tf; cannam@95: } cannam@95: Ts = To - Tr; cannam@95: TM = T6 - T9; cannam@95: TN = Td - Tg; cannam@95: Tz = Tv - Ty; cannam@95: Ta = T6 + T9; cannam@95: Th = Td + Tg; cannam@95: Ti = Ta + Th; cannam@95: T1g = TY + TZ; cannam@95: T1h = T11 + T12; cannam@95: T1k = T1g + T1h; cannam@95: T10 = TY - TZ; cannam@95: T13 = T11 - T12; cannam@95: T19 = T10 + T13; cannam@95: TG = Tr + To; cannam@95: TH = Ty + Tv; cannam@95: TI = TG + TH; cannam@95: T1d = TU + TV; cannam@95: T1e = TR + TS; cannam@95: T1j = T1d + T1e; cannam@95: TT = TR - TS; cannam@95: TW = TU - TV; cannam@95: T18 = TW + TT; cannam@95: } cannam@95: Cr[WS(csr, 5)] = T3 + Ti; cannam@95: Ci[WS(csi, 5)] = TF - TI; cannam@95: { cannam@95: E TX, T14, T1f, T1i; cannam@95: TX = TT - TW; cannam@95: T14 = T10 - T13; cannam@95: Ci[WS(csi, 6)] = FNMS(KP587785252, T14, KP951056516 * TX); cannam@95: Ci[WS(csi, 2)] = FMA(KP587785252, TX, KP951056516 * T14); cannam@95: T1f = T1d - T1e; cannam@95: T1i = T1g - T1h; cannam@95: Ci[WS(csi, 8)] = FNMS(KP951056516, T1i, KP587785252 * T1f); cannam@95: Ci[WS(csi, 4)] = FMA(KP951056516, T1f, KP587785252 * T1i); cannam@95: } cannam@95: { cannam@95: E T1l, T1n, T1o, T1c, T1a, T1b; cannam@95: T1l = KP559016994 * (T1j - T1k); cannam@95: T1n = T1j + T1k; cannam@95: T1o = FNMS(KP250000000, T1n, T1m); cannam@95: Cr[WS(csr, 4)] = T1l + T1o; cannam@95: Cr[0] = T1m + T1n; cannam@95: Cr[WS(csr, 8)] = T1o - T1l; cannam@95: T1c = KP559016994 * (T18 - T19); cannam@95: T1a = T18 + T19; cannam@95: T1b = FNMS(KP250000000, T1a, T17); cannam@95: Cr[WS(csr, 2)] = T1b - T1c; cannam@95: Cr[WS(csr, 10)] = T17 + T1a; cannam@95: Cr[WS(csr, 6)] = T1c + T1b; cannam@95: } cannam@95: { cannam@95: E TA, TC, Tl, TB, Tj, Tk; cannam@95: TA = FMA(KP951056516, Ts, KP587785252 * Tz); cannam@95: TC = FNMS(KP587785252, Ts, KP951056516 * Tz); cannam@95: Tj = KP559016994 * (Ta - Th); cannam@95: Tk = FNMS(KP250000000, Ti, T3); cannam@95: Tl = Tj + Tk; cannam@95: TB = Tk - Tj; cannam@95: Cr[WS(csr, 9)] = Tl - TA; cannam@95: Cr[WS(csr, 7)] = TB + TC; cannam@95: Cr[WS(csr, 1)] = Tl + TA; cannam@95: Cr[WS(csr, 3)] = TB - TC; cannam@95: } cannam@95: { cannam@95: E TO, TQ, TL, TP, TJ, TK; cannam@95: TO = FMA(KP951056516, TM, KP587785252 * TN); cannam@95: TQ = FNMS(KP587785252, TM, KP951056516 * TN); cannam@95: TJ = FMA(KP250000000, TI, TF); cannam@95: TK = KP559016994 * (TH - TG); cannam@95: TL = TJ + TK; cannam@95: TP = TK - TJ; cannam@95: Ci[WS(csi, 1)] = TL - TO; cannam@95: Ci[WS(csi, 7)] = TQ + TP; cannam@95: Ci[WS(csi, 9)] = TO + TL; cannam@95: Ci[WS(csi, 3)] = TP - TQ; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 20, "r2cf_20", {74, 12, 12, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cf_20) (planner *p) { cannam@95: X(kr2c_register) (p, r2cf_20, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */