cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:06:27 EDT 2018 */ cannam@167: cannam@167: #include "rdft/codelet-rdft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_r2cf.native -fma -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cf_20 -include rdft/scalar/r2cf.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 86 FP additions, 32 FP multiplications, cannam@167: * (or, 58 additions, 4 multiplications, 28 fused multiply/add), cannam@167: * 51 stack variables, 4 constants, and 40 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cf.h" cannam@167: cannam@167: 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@167: { cannam@167: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@167: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: { cannam@167: INT i; cannam@167: 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@167: E T3, T1d, TJ, TV, T16, T1k, T1l, T19, Ta, Th, Ti, T1e, T1f, T1g, TP; cannam@167: E TQ, TX, Tn, Ts, TK, TS, TT, TW, Ty, TD, TL; cannam@167: { cannam@167: E T1, T2, TF, TG, TH, TI; cannam@167: T1 = R0[0]; cannam@167: T2 = R0[WS(rs, 5)]; cannam@167: TF = T1 + T2; cannam@167: TG = R1[WS(rs, 2)]; cannam@167: TH = R1[WS(rs, 7)]; cannam@167: TI = TG + TH; cannam@167: T3 = T1 - T2; cannam@167: T1d = TG - TH; cannam@167: TJ = TF - TI; cannam@167: TV = TF + TI; cannam@167: } cannam@167: { cannam@167: E T6, To, Tx, T17, TC, T18, T9, Tj, Td, Tu, Tm, T15, Tr, T14, Tg; cannam@167: E Tz; cannam@167: { cannam@167: E T4, T5, Tv, Tw; cannam@167: T4 = R0[WS(rs, 2)]; cannam@167: T5 = R0[WS(rs, 7)]; cannam@167: T6 = T4 - T5; cannam@167: To = T4 + T5; cannam@167: Tv = R1[WS(rs, 6)]; cannam@167: Tw = R1[WS(rs, 1)]; cannam@167: Tx = Tv + Tw; cannam@167: T17 = Tw - Tv; cannam@167: } cannam@167: { cannam@167: E TA, TB, T7, T8; cannam@167: TA = R1[WS(rs, 8)]; cannam@167: TB = R1[WS(rs, 3)]; cannam@167: TC = TA + TB; cannam@167: T18 = TB - TA; cannam@167: T7 = R0[WS(rs, 8)]; cannam@167: T8 = R0[WS(rs, 3)]; cannam@167: T9 = T7 - T8; cannam@167: Tj = T7 + T8; cannam@167: } cannam@167: { cannam@167: E Tb, Tc, Tk, Tl; cannam@167: Tb = R0[WS(rs, 4)]; cannam@167: Tc = R0[WS(rs, 9)]; cannam@167: Td = Tb - Tc; cannam@167: Tu = Tb + Tc; cannam@167: Tk = R1[0]; cannam@167: Tl = R1[WS(rs, 5)]; cannam@167: Tm = Tk + Tl; cannam@167: T15 = Tl - Tk; cannam@167: } cannam@167: { cannam@167: E Tp, Tq, Te, Tf; cannam@167: Tp = R1[WS(rs, 4)]; cannam@167: Tq = R1[WS(rs, 9)]; cannam@167: Tr = Tp + Tq; cannam@167: T14 = Tq - Tp; cannam@167: Te = R0[WS(rs, 6)]; cannam@167: Tf = R0[WS(rs, 1)]; cannam@167: Tg = Te - Tf; cannam@167: Tz = Te + Tf; cannam@167: } cannam@167: T16 = T14 - T15; cannam@167: T1k = T6 - T9; cannam@167: T1l = Td - Tg; cannam@167: T19 = T17 - T18; cannam@167: Ta = T6 + T9; cannam@167: Th = Td + Tg; cannam@167: Ti = Ta + Th; cannam@167: T1e = T14 + T15; cannam@167: T1f = T17 + T18; cannam@167: T1g = T1e + T1f; cannam@167: TP = Tu + Tx; cannam@167: TQ = Tz + TC; cannam@167: TX = TP + TQ; cannam@167: Tn = Tj - Tm; cannam@167: Ts = To - Tr; cannam@167: TK = Ts + Tn; cannam@167: TS = To + Tr; cannam@167: TT = Tj + Tm; cannam@167: TW = TS + TT; cannam@167: Ty = Tu - Tx; cannam@167: TD = Tz - TC; cannam@167: TL = Ty + TD; cannam@167: } cannam@167: Cr[WS(csr, 5)] = T3 + Ti; cannam@167: Ci[WS(csi, 5)] = T1g - T1d; cannam@167: { cannam@167: E Tt, TE, TR, TU; cannam@167: Tt = Tn - Ts; cannam@167: TE = Ty - TD; cannam@167: Ci[WS(csi, 6)] = KP951056516 * (FNMS(KP618033988, TE, Tt)); cannam@167: Ci[WS(csi, 2)] = KP951056516 * (FMA(KP618033988, Tt, TE)); cannam@167: TR = TP - TQ; cannam@167: TU = TS - TT; cannam@167: Ci[WS(csi, 8)] = -(KP951056516 * (FNMS(KP618033988, TU, TR))); cannam@167: Ci[WS(csi, 4)] = KP951056516 * (FMA(KP618033988, TR, TU)); cannam@167: } cannam@167: { cannam@167: E T10, TY, TZ, TO, TM, TN; cannam@167: T10 = TW - TX; cannam@167: TY = TW + TX; cannam@167: TZ = FNMS(KP250000000, TY, TV); cannam@167: Cr[WS(csr, 4)] = FMA(KP559016994, T10, TZ); cannam@167: Cr[0] = TV + TY; cannam@167: Cr[WS(csr, 8)] = FNMS(KP559016994, T10, TZ); cannam@167: TO = TK - TL; cannam@167: TM = TK + TL; cannam@167: TN = FNMS(KP250000000, TM, TJ); cannam@167: Cr[WS(csr, 2)] = FNMS(KP559016994, TO, TN); cannam@167: Cr[WS(csr, 10)] = TJ + TM; cannam@167: Cr[WS(csr, 6)] = FMA(KP559016994, TO, TN); cannam@167: } cannam@167: { cannam@167: E T1a, T1c, T13, T1b, T11, T12; cannam@167: T1a = FMA(KP618033988, T19, T16); cannam@167: T1c = FNMS(KP618033988, T16, T19); cannam@167: T11 = FNMS(KP250000000, Ti, T3); cannam@167: T12 = Ta - Th; cannam@167: T13 = FMA(KP559016994, T12, T11); cannam@167: T1b = FNMS(KP559016994, T12, T11); cannam@167: Cr[WS(csr, 9)] = FNMS(KP951056516, T1a, T13); cannam@167: Cr[WS(csr, 7)] = FMA(KP951056516, T1c, T1b); cannam@167: Cr[WS(csr, 1)] = FMA(KP951056516, T1a, T13); cannam@167: Cr[WS(csr, 3)] = FNMS(KP951056516, T1c, T1b); cannam@167: } cannam@167: { cannam@167: E T1m, T1o, T1j, T1n, T1h, T1i; cannam@167: T1m = FMA(KP618033988, T1l, T1k); cannam@167: T1o = FNMS(KP618033988, T1k, T1l); cannam@167: T1h = FMA(KP250000000, T1g, T1d); cannam@167: T1i = T1e - T1f; cannam@167: T1j = FNMS(KP559016994, T1i, T1h); cannam@167: T1n = FMA(KP559016994, T1i, T1h); cannam@167: Ci[WS(csi, 1)] = -(FMA(KP951056516, T1m, T1j)); cannam@167: Ci[WS(csi, 7)] = FMA(KP951056516, T1o, T1n); cannam@167: Ci[WS(csi, 9)] = FMS(KP951056516, T1m, T1j); cannam@167: Ci[WS(csi, 3)] = FNMS(KP951056516, T1o, T1n); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 20, "r2cf_20", {58, 4, 28, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cf_20) (planner *p) { cannam@167: X(kr2c_register) (p, r2cf_20, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 20 -name r2cf_20 -include rdft/scalar/r2cf.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 86 FP additions, 24 FP multiplications, cannam@167: * (or, 74 additions, 12 multiplications, 12 fused multiply/add), cannam@167: * 51 stack variables, 4 constants, and 40 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cf.h" cannam@167: cannam@167: 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@167: { cannam@167: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@167: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: { cannam@167: INT i; cannam@167: 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@167: E T3, T1m, TF, T17, Ts, TM, TN, Tz, Ta, Th, Ti, T1g, T1h, T1k, T10; cannam@167: E T13, T19, TG, TH, TI, T1d, T1e, T1j, TT, TW, T18; cannam@167: { cannam@167: E T1, T2, T15, TD, TE, T16; cannam@167: T1 = R0[0]; cannam@167: T2 = R0[WS(rs, 5)]; cannam@167: T15 = T1 + T2; cannam@167: TD = R1[WS(rs, 7)]; cannam@167: TE = R1[WS(rs, 2)]; cannam@167: T16 = TE + TD; cannam@167: T3 = T1 - T2; cannam@167: T1m = T15 + T16; cannam@167: TF = TD - TE; cannam@167: T17 = T15 - T16; cannam@167: } cannam@167: { cannam@167: E T6, TU, Tv, T12, Ty, TZ, T9, TR, Td, TY, To, TS, Tr, TV, Tg; cannam@167: E T11; cannam@167: { cannam@167: E T4, T5, Tt, Tu; cannam@167: T4 = R0[WS(rs, 2)]; cannam@167: T5 = R0[WS(rs, 7)]; cannam@167: T6 = T4 - T5; cannam@167: TU = T4 + T5; cannam@167: Tt = R1[WS(rs, 8)]; cannam@167: Tu = R1[WS(rs, 3)]; cannam@167: Tv = Tt - Tu; cannam@167: T12 = Tt + Tu; cannam@167: } cannam@167: { cannam@167: E Tw, Tx, T7, T8; cannam@167: Tw = R1[WS(rs, 6)]; cannam@167: Tx = R1[WS(rs, 1)]; cannam@167: Ty = Tw - Tx; cannam@167: TZ = Tw + Tx; cannam@167: T7 = R0[WS(rs, 8)]; cannam@167: T8 = R0[WS(rs, 3)]; cannam@167: T9 = T7 - T8; cannam@167: TR = T7 + T8; cannam@167: } cannam@167: { cannam@167: E Tb, Tc, Tm, Tn; cannam@167: Tb = R0[WS(rs, 4)]; cannam@167: Tc = R0[WS(rs, 9)]; cannam@167: Td = Tb - Tc; cannam@167: TY = Tb + Tc; cannam@167: Tm = R1[0]; cannam@167: Tn = R1[WS(rs, 5)]; cannam@167: To = Tm - Tn; cannam@167: TS = Tm + Tn; cannam@167: } cannam@167: { cannam@167: E Tp, Tq, Te, Tf; cannam@167: Tp = R1[WS(rs, 4)]; cannam@167: Tq = R1[WS(rs, 9)]; cannam@167: Tr = Tp - Tq; cannam@167: TV = Tp + Tq; cannam@167: Te = R0[WS(rs, 6)]; cannam@167: Tf = R0[WS(rs, 1)]; cannam@167: Tg = Te - Tf; cannam@167: T11 = Te + Tf; cannam@167: } cannam@167: Ts = To - Tr; cannam@167: TM = T6 - T9; cannam@167: TN = Td - Tg; cannam@167: Tz = Tv - Ty; cannam@167: Ta = T6 + T9; cannam@167: Th = Td + Tg; cannam@167: Ti = Ta + Th; cannam@167: T1g = TY + TZ; cannam@167: T1h = T11 + T12; cannam@167: T1k = T1g + T1h; cannam@167: T10 = TY - TZ; cannam@167: T13 = T11 - T12; cannam@167: T19 = T10 + T13; cannam@167: TG = Tr + To; cannam@167: TH = Ty + Tv; cannam@167: TI = TG + TH; cannam@167: T1d = TU + TV; cannam@167: T1e = TR + TS; cannam@167: T1j = T1d + T1e; cannam@167: TT = TR - TS; cannam@167: TW = TU - TV; cannam@167: T18 = TW + TT; cannam@167: } cannam@167: Cr[WS(csr, 5)] = T3 + Ti; cannam@167: Ci[WS(csi, 5)] = TF - TI; cannam@167: { cannam@167: E TX, T14, T1f, T1i; cannam@167: TX = TT - TW; cannam@167: T14 = T10 - T13; cannam@167: Ci[WS(csi, 6)] = FNMS(KP587785252, T14, KP951056516 * TX); cannam@167: Ci[WS(csi, 2)] = FMA(KP587785252, TX, KP951056516 * T14); cannam@167: T1f = T1d - T1e; cannam@167: T1i = T1g - T1h; cannam@167: Ci[WS(csi, 8)] = FNMS(KP951056516, T1i, KP587785252 * T1f); cannam@167: Ci[WS(csi, 4)] = FMA(KP951056516, T1f, KP587785252 * T1i); cannam@167: } cannam@167: { cannam@167: E T1l, T1n, T1o, T1c, T1a, T1b; cannam@167: T1l = KP559016994 * (T1j - T1k); cannam@167: T1n = T1j + T1k; cannam@167: T1o = FNMS(KP250000000, T1n, T1m); cannam@167: Cr[WS(csr, 4)] = T1l + T1o; cannam@167: Cr[0] = T1m + T1n; cannam@167: Cr[WS(csr, 8)] = T1o - T1l; cannam@167: T1c = KP559016994 * (T18 - T19); cannam@167: T1a = T18 + T19; cannam@167: T1b = FNMS(KP250000000, T1a, T17); cannam@167: Cr[WS(csr, 2)] = T1b - T1c; cannam@167: Cr[WS(csr, 10)] = T17 + T1a; cannam@167: Cr[WS(csr, 6)] = T1c + T1b; cannam@167: } cannam@167: { cannam@167: E TA, TC, Tl, TB, Tj, Tk; cannam@167: TA = FMA(KP951056516, Ts, KP587785252 * Tz); cannam@167: TC = FNMS(KP587785252, Ts, KP951056516 * Tz); cannam@167: Tj = KP559016994 * (Ta - Th); cannam@167: Tk = FNMS(KP250000000, Ti, T3); cannam@167: Tl = Tj + Tk; cannam@167: TB = Tk - Tj; cannam@167: Cr[WS(csr, 9)] = Tl - TA; cannam@167: Cr[WS(csr, 7)] = TB + TC; cannam@167: Cr[WS(csr, 1)] = Tl + TA; cannam@167: Cr[WS(csr, 3)] = TB - TC; cannam@167: } cannam@167: { cannam@167: E TO, TQ, TL, TP, TJ, TK; cannam@167: TO = FMA(KP951056516, TM, KP587785252 * TN); cannam@167: TQ = FNMS(KP587785252, TM, KP951056516 * TN); cannam@167: TJ = FMA(KP250000000, TI, TF); cannam@167: TK = KP559016994 * (TH - TG); cannam@167: TL = TJ + TK; cannam@167: TP = TK - TJ; cannam@167: Ci[WS(csi, 1)] = TL - TO; cannam@167: Ci[WS(csi, 7)] = TQ + TP; cannam@167: Ci[WS(csi, 9)] = TO + TL; cannam@167: Ci[WS(csi, 3)] = TP - TQ; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 20, "r2cf_20", {74, 12, 12, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cf_20) (planner *p) { cannam@167: X(kr2c_register) (p, r2cf_20, &desc); cannam@167: } cannam@167: cannam@167: #endif