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 08:07:29 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_r2cb.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 58 FP additions, 32 FP multiplications, cannam@167: * (or, 26 additions, 0 multiplications, 32 fused multiply/add), cannam@167: * 31 stack variables, 4 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cb.h" cannam@167: cannam@167: static void r2cb_16(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(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@167: DK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@167: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@167: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@167: E T5, TL, Tj, TD, T8, TM, To, TE, Tc, TP, Tf, TQ, Tu, Tz, TR; cannam@167: E TO, TH, TG; cannam@167: { cannam@167: E T4, Ti, T3, Th, T1, T2; cannam@167: T4 = Cr[WS(csr, 4)]; cannam@167: Ti = Ci[WS(csi, 4)]; cannam@167: T1 = Cr[0]; cannam@167: T2 = Cr[WS(csr, 8)]; cannam@167: T3 = T1 + T2; cannam@167: Th = T1 - T2; cannam@167: T5 = FMA(KP2_000000000, T4, T3); cannam@167: TL = FNMS(KP2_000000000, T4, T3); cannam@167: Tj = FNMS(KP2_000000000, Ti, Th); cannam@167: TD = FMA(KP2_000000000, Ti, Th); cannam@167: } cannam@167: { cannam@167: E T6, T7, Tk, Tl, Tm, Tn; cannam@167: T6 = Cr[WS(csr, 2)]; cannam@167: T7 = Cr[WS(csr, 6)]; cannam@167: Tk = T6 - T7; cannam@167: Tl = Ci[WS(csi, 2)]; cannam@167: Tm = Ci[WS(csi, 6)]; cannam@167: Tn = Tl + Tm; cannam@167: T8 = T6 + T7; cannam@167: TM = Tl - Tm; cannam@167: To = Tk - Tn; cannam@167: TE = Tk + Tn; cannam@167: } cannam@167: { cannam@167: E Tq, Ty, Tv, Tt; cannam@167: { cannam@167: E Ta, Tb, Tw, Tx; cannam@167: Ta = Cr[WS(csr, 1)]; cannam@167: Tb = Cr[WS(csr, 7)]; cannam@167: Tc = Ta + Tb; cannam@167: Tq = Ta - Tb; cannam@167: Tw = Ci[WS(csi, 1)]; cannam@167: Tx = Ci[WS(csi, 7)]; cannam@167: Ty = Tw + Tx; cannam@167: TP = Tw - Tx; cannam@167: } cannam@167: { cannam@167: E Td, Te, Tr, Ts; cannam@167: Td = Cr[WS(csr, 5)]; cannam@167: Te = Cr[WS(csr, 3)]; cannam@167: Tf = Td + Te; cannam@167: Tv = Td - Te; cannam@167: Tr = Ci[WS(csi, 5)]; cannam@167: Ts = Ci[WS(csi, 3)]; cannam@167: Tt = Tr + Ts; cannam@167: TQ = Tr - Ts; cannam@167: } cannam@167: Tu = Tq - Tt; cannam@167: Tz = Tv + Ty; cannam@167: TR = TP - TQ; cannam@167: TO = Tc - Tf; cannam@167: TH = Tq + Tt; cannam@167: TG = Ty - Tv; cannam@167: } cannam@167: { cannam@167: E T9, Tg, TT, TU; cannam@167: T9 = FMA(KP2_000000000, T8, T5); cannam@167: Tg = Tc + Tf; cannam@167: R0[WS(rs, 4)] = FNMS(KP2_000000000, Tg, T9); cannam@167: R0[0] = FMA(KP2_000000000, Tg, T9); cannam@167: TT = FMA(KP2_000000000, TM, TL); cannam@167: TU = TO + TR; cannam@167: R0[WS(rs, 3)] = FNMS(KP1_414213562, TU, TT); cannam@167: R0[WS(rs, 7)] = FMA(KP1_414213562, TU, TT); cannam@167: } cannam@167: { cannam@167: E TV, TW, Tp, TA; cannam@167: TV = FNMS(KP2_000000000, T8, T5); cannam@167: TW = TQ + TP; cannam@167: R0[WS(rs, 2)] = FNMS(KP2_000000000, TW, TV); cannam@167: R0[WS(rs, 6)] = FMA(KP2_000000000, TW, TV); cannam@167: Tp = FMA(KP1_414213562, To, Tj); cannam@167: TA = FNMS(KP414213562, Tz, Tu); cannam@167: R1[WS(rs, 4)] = FNMS(KP1_847759065, TA, Tp); cannam@167: R1[0] = FMA(KP1_847759065, TA, Tp); cannam@167: } cannam@167: { cannam@167: E TB, TC, TJ, TK; cannam@167: TB = FNMS(KP1_414213562, To, Tj); cannam@167: TC = FMA(KP414213562, Tu, Tz); cannam@167: R1[WS(rs, 2)] = FNMS(KP1_847759065, TC, TB); cannam@167: R1[WS(rs, 6)] = FMA(KP1_847759065, TC, TB); cannam@167: TJ = FMA(KP1_414213562, TE, TD); cannam@167: TK = FMA(KP414213562, TG, TH); cannam@167: R1[WS(rs, 3)] = FNMS(KP1_847759065, TK, TJ); cannam@167: R1[WS(rs, 7)] = FMA(KP1_847759065, TK, TJ); cannam@167: } cannam@167: { cannam@167: E TN, TS, TF, TI; cannam@167: TN = FNMS(KP2_000000000, TM, TL); cannam@167: TS = TO - TR; cannam@167: R0[WS(rs, 5)] = FNMS(KP1_414213562, TS, TN); cannam@167: R0[WS(rs, 1)] = FMA(KP1_414213562, TS, TN); cannam@167: TF = FNMS(KP1_414213562, TE, TD); cannam@167: TI = FNMS(KP414213562, TH, TG); cannam@167: R1[WS(rs, 1)] = FNMS(KP1_847759065, TI, TF); cannam@167: R1[WS(rs, 5)] = FMA(KP1_847759065, TI, TF); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 16, "r2cb_16", {26, 0, 32, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cb_16) (planner *p) { cannam@167: X(kr2c_register) (p, r2cb_16, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 58 FP additions, 18 FP multiplications, cannam@167: * (or, 54 additions, 14 multiplications, 4 fused multiply/add), cannam@167: * 31 stack variables, 4 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cb.h" cannam@167: cannam@167: static void r2cb_16(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(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@167: DK(KP765366864, +0.765366864730179543456919968060797733522689125); cannam@167: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@167: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@167: E T9, TS, Tl, TG, T6, TR, Ti, TD, Td, Tq, Tg, Tt, Tn, Tu, TV; cannam@167: E TU, TN, TK; cannam@167: { cannam@167: E T7, T8, TE, Tj, Tk, TF; cannam@167: T7 = Cr[WS(csr, 2)]; cannam@167: T8 = Cr[WS(csr, 6)]; cannam@167: TE = T7 - T8; cannam@167: Tj = Ci[WS(csi, 2)]; cannam@167: Tk = Ci[WS(csi, 6)]; cannam@167: TF = Tj + Tk; cannam@167: T9 = KP2_000000000 * (T7 + T8); cannam@167: TS = KP1_414213562 * (TE + TF); cannam@167: Tl = KP2_000000000 * (Tj - Tk); cannam@167: TG = KP1_414213562 * (TE - TF); cannam@167: } cannam@167: { cannam@167: E T5, TC, T3, TA; cannam@167: { cannam@167: E T4, TB, T1, T2; cannam@167: T4 = Cr[WS(csr, 4)]; cannam@167: T5 = KP2_000000000 * T4; cannam@167: TB = Ci[WS(csi, 4)]; cannam@167: TC = KP2_000000000 * TB; cannam@167: T1 = Cr[0]; cannam@167: T2 = Cr[WS(csr, 8)]; cannam@167: T3 = T1 + T2; cannam@167: TA = T1 - T2; cannam@167: } cannam@167: T6 = T3 + T5; cannam@167: TR = TA + TC; cannam@167: Ti = T3 - T5; cannam@167: TD = TA - TC; cannam@167: } cannam@167: { cannam@167: E TI, TM, TL, TJ; cannam@167: { cannam@167: E Tb, Tc, To, Tp; cannam@167: Tb = Cr[WS(csr, 1)]; cannam@167: Tc = Cr[WS(csr, 7)]; cannam@167: Td = Tb + Tc; cannam@167: TI = Tb - Tc; cannam@167: To = Ci[WS(csi, 1)]; cannam@167: Tp = Ci[WS(csi, 7)]; cannam@167: Tq = To - Tp; cannam@167: TM = To + Tp; cannam@167: } cannam@167: { cannam@167: E Te, Tf, Tr, Ts; cannam@167: Te = Cr[WS(csr, 5)]; cannam@167: Tf = Cr[WS(csr, 3)]; cannam@167: Tg = Te + Tf; cannam@167: TL = Te - Tf; cannam@167: Tr = Ci[WS(csi, 5)]; cannam@167: Ts = Ci[WS(csi, 3)]; cannam@167: Tt = Tr - Ts; cannam@167: TJ = Tr + Ts; cannam@167: } cannam@167: Tn = Td - Tg; cannam@167: Tu = Tq - Tt; cannam@167: TV = TM - TL; cannam@167: TU = TI + TJ; cannam@167: TN = TL + TM; cannam@167: TK = TI - TJ; cannam@167: } cannam@167: { cannam@167: E Ta, Th, TT, TW; cannam@167: Ta = T6 + T9; cannam@167: Th = KP2_000000000 * (Td + Tg); cannam@167: R0[WS(rs, 4)] = Ta - Th; cannam@167: R0[0] = Ta + Th; cannam@167: TT = TR - TS; cannam@167: TW = FNMS(KP1_847759065, TV, KP765366864 * TU); cannam@167: R1[WS(rs, 5)] = TT - TW; cannam@167: R1[WS(rs, 1)] = TT + TW; cannam@167: } cannam@167: { cannam@167: E TX, TY, Tm, Tv; cannam@167: TX = TR + TS; cannam@167: TY = FMA(KP1_847759065, TU, KP765366864 * TV); cannam@167: R1[WS(rs, 3)] = TX - TY; cannam@167: R1[WS(rs, 7)] = TX + TY; cannam@167: Tm = Ti - Tl; cannam@167: Tv = KP1_414213562 * (Tn - Tu); cannam@167: R0[WS(rs, 5)] = Tm - Tv; cannam@167: R0[WS(rs, 1)] = Tm + Tv; cannam@167: } cannam@167: { cannam@167: E Tw, Tx, TH, TO; cannam@167: Tw = Ti + Tl; cannam@167: Tx = KP1_414213562 * (Tn + Tu); cannam@167: R0[WS(rs, 3)] = Tw - Tx; cannam@167: R0[WS(rs, 7)] = Tw + Tx; cannam@167: TH = TD + TG; cannam@167: TO = FNMS(KP765366864, TN, KP1_847759065 * TK); cannam@167: R1[WS(rs, 4)] = TH - TO; cannam@167: R1[0] = TH + TO; cannam@167: } cannam@167: { cannam@167: E TP, TQ, Ty, Tz; cannam@167: TP = TD - TG; cannam@167: TQ = FMA(KP765366864, TK, KP1_847759065 * TN); cannam@167: R1[WS(rs, 2)] = TP - TQ; cannam@167: R1[WS(rs, 6)] = TP + TQ; cannam@167: Ty = T6 - T9; cannam@167: Tz = KP2_000000000 * (Tt + Tq); cannam@167: R0[WS(rs, 2)] = Ty - Tz; cannam@167: R0[WS(rs, 6)] = Ty + Tz; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 16, "r2cb_16", {54, 14, 4, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cb_16) (planner *p) { cannam@167: X(kr2c_register) (p, r2cb_16, &desc); cannam@167: } cannam@167: cannam@167: #endif