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 08:06:55 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_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cf_4 -include rdft/scalar/hc2cf.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 22 FP additions, 12 FP multiplications, cannam@167: * (or, 16 additions, 6 multiplications, 6 fused multiply/add), cannam@167: * 15 stack variables, 0 constants, and 16 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cf.h" cannam@167: cannam@167: static void hc2cf_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) { cannam@167: E T1, Tv, T7, Tu, Te, To, Tk, Tq; cannam@167: T1 = Rp[0]; cannam@167: Tv = Rm[0]; cannam@167: { cannam@167: E T3, T6, T4, Tt, T2, T5; cannam@167: T3 = Rp[WS(rs, 1)]; cannam@167: T6 = Rm[WS(rs, 1)]; cannam@167: T2 = W[2]; cannam@167: T4 = T2 * T3; cannam@167: Tt = T2 * T6; cannam@167: T5 = W[3]; cannam@167: T7 = FMA(T5, T6, T4); cannam@167: Tu = FNMS(T5, T3, Tt); cannam@167: } cannam@167: { cannam@167: E Ta, Td, Tb, Tn, T9, Tc; cannam@167: Ta = Ip[0]; cannam@167: Td = Im[0]; cannam@167: T9 = W[0]; cannam@167: Tb = T9 * Ta; cannam@167: Tn = T9 * Td; cannam@167: Tc = W[1]; cannam@167: Te = FMA(Tc, Td, Tb); cannam@167: To = FNMS(Tc, Ta, Tn); cannam@167: } cannam@167: { cannam@167: E Tg, Tj, Th, Tp, Tf, Ti; cannam@167: Tg = Ip[WS(rs, 1)]; cannam@167: Tj = Im[WS(rs, 1)]; cannam@167: Tf = W[4]; cannam@167: Th = Tf * Tg; cannam@167: Tp = Tf * Tj; cannam@167: Ti = W[5]; cannam@167: Tk = FMA(Ti, Tj, Th); cannam@167: Tq = FNMS(Ti, Tg, Tp); cannam@167: } cannam@167: { cannam@167: E T8, Tl, Ts, Tw; cannam@167: T8 = T1 + T7; cannam@167: Tl = Te + Tk; cannam@167: Rm[WS(rs, 1)] = T8 - Tl; cannam@167: Rp[0] = T8 + Tl; cannam@167: Ts = To + Tq; cannam@167: Tw = Tu + Tv; cannam@167: Im[WS(rs, 1)] = Ts - Tw; cannam@167: Ip[0] = Ts + Tw; cannam@167: } cannam@167: { cannam@167: E Tm, Tr, Tx, Ty; cannam@167: Tm = T1 - T7; cannam@167: Tr = To - Tq; cannam@167: Rm[0] = Tm - Tr; cannam@167: Rp[WS(rs, 1)] = Tm + Tr; cannam@167: Tx = Tk - Te; cannam@167: Ty = Tv - Tu; cannam@167: Im[0] = Tx - Ty; cannam@167: Ip[WS(rs, 1)] = Tx + Ty; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 4}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 4, "hc2cf_4", twinstr, &GENUS, {16, 6, 6, 0} }; cannam@167: cannam@167: void X(codelet_hc2cf_4) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cf_4, &desc, HC2C_VIA_RDFT); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cf_4 -include rdft/scalar/hc2cf.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 22 FP additions, 12 FP multiplications, cannam@167: * (or, 16 additions, 6 multiplications, 6 fused multiply/add), cannam@167: * 13 stack variables, 0 constants, and 16 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cf.h" cannam@167: cannam@167: static void hc2cf_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) { cannam@167: E T1, Tp, T6, To, Tc, Tk, Th, Tl; cannam@167: T1 = Rp[0]; cannam@167: Tp = Rm[0]; cannam@167: { cannam@167: E T3, T5, T2, T4; cannam@167: T3 = Rp[WS(rs, 1)]; cannam@167: T5 = Rm[WS(rs, 1)]; cannam@167: T2 = W[2]; cannam@167: T4 = W[3]; cannam@167: T6 = FMA(T2, T3, T4 * T5); cannam@167: To = FNMS(T4, T3, T2 * T5); cannam@167: } cannam@167: { cannam@167: E T9, Tb, T8, Ta; cannam@167: T9 = Ip[0]; cannam@167: Tb = Im[0]; cannam@167: T8 = W[0]; cannam@167: Ta = W[1]; cannam@167: Tc = FMA(T8, T9, Ta * Tb); cannam@167: Tk = FNMS(Ta, T9, T8 * Tb); cannam@167: } cannam@167: { cannam@167: E Te, Tg, Td, Tf; cannam@167: Te = Ip[WS(rs, 1)]; cannam@167: Tg = Im[WS(rs, 1)]; cannam@167: Td = W[4]; cannam@167: Tf = W[5]; cannam@167: Th = FMA(Td, Te, Tf * Tg); cannam@167: Tl = FNMS(Tf, Te, Td * Tg); cannam@167: } cannam@167: { cannam@167: E T7, Ti, Tn, Tq; cannam@167: T7 = T1 + T6; cannam@167: Ti = Tc + Th; cannam@167: Rm[WS(rs, 1)] = T7 - Ti; cannam@167: Rp[0] = T7 + Ti; cannam@167: Tn = Tk + Tl; cannam@167: Tq = To + Tp; cannam@167: Im[WS(rs, 1)] = Tn - Tq; cannam@167: Ip[0] = Tn + Tq; cannam@167: } cannam@167: { cannam@167: E Tj, Tm, Tr, Ts; cannam@167: Tj = T1 - T6; cannam@167: Tm = Tk - Tl; cannam@167: Rm[0] = Tj - Tm; cannam@167: Rp[WS(rs, 1)] = Tj + Tm; cannam@167: Tr = Th - Tc; cannam@167: Ts = Tp - To; cannam@167: Im[0] = Tr - Ts; cannam@167: Ip[WS(rs, 1)] = Tr + Ts; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 4}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 4, "hc2cf_4", twinstr, &GENUS, {16, 6, 6, 0} }; cannam@167: cannam@167: void X(codelet_hc2cf_4) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cf_4, &desc, HC2C_VIA_RDFT); cannam@167: } cannam@167: #endif