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 08:07:51 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 -sign 1 -n 4 -dif -name hc2cb_4 -include rdft/scalar/hc2cb.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: * 22 stack variables, 0 constants, and 16 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cb.h" cannam@167: cannam@167: static void hc2cb_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 T3, T6, T8, Td, Tx, Tu, Tm, Tg, Tr; cannam@167: { cannam@167: E Tb, Tc, Tq, Tk, Te, Tf, Tl, Tp; cannam@167: { cannam@167: E T1, T2, T4, T5; cannam@167: Tb = Ip[0]; cannam@167: Tc = Im[WS(rs, 1)]; cannam@167: Tq = Tb + Tc; cannam@167: T1 = Rp[0]; cannam@167: T2 = Rm[WS(rs, 1)]; cannam@167: T3 = T1 + T2; cannam@167: Tk = T1 - T2; cannam@167: Te = Ip[WS(rs, 1)]; cannam@167: Tf = Im[0]; cannam@167: Tl = Te + Tf; cannam@167: T4 = Rp[WS(rs, 1)]; cannam@167: T5 = Rm[0]; cannam@167: T6 = T4 + T5; cannam@167: Tp = T4 - T5; cannam@167: } cannam@167: T8 = T3 - T6; cannam@167: Td = Tb - Tc; cannam@167: Tx = Tq - Tp; cannam@167: Tu = Tk + Tl; cannam@167: Tm = Tk - Tl; cannam@167: Tg = Te - Tf; cannam@167: Tr = Tp + Tq; cannam@167: } cannam@167: Rp[0] = T3 + T6; cannam@167: Rm[0] = Td + Tg; cannam@167: { cannam@167: E Tn, Ts, Tj, To; cannam@167: Tj = W[0]; cannam@167: Tn = Tj * Tm; cannam@167: Ts = Tj * Tr; cannam@167: To = W[1]; cannam@167: Ip[0] = FNMS(To, Tr, Tn); cannam@167: Im[0] = FMA(To, Tm, Ts); cannam@167: } cannam@167: { cannam@167: E Tv, Ty, Tt, Tw; cannam@167: Tt = W[4]; cannam@167: Tv = Tt * Tu; cannam@167: Ty = Tt * Tx; cannam@167: Tw = W[5]; cannam@167: Ip[WS(rs, 1)] = FNMS(Tw, Tx, Tv); cannam@167: Im[WS(rs, 1)] = FMA(Tw, Tu, Ty); cannam@167: } cannam@167: { cannam@167: E Th, Ta, Ti, T7, T9; cannam@167: Th = Td - Tg; cannam@167: Ta = W[3]; cannam@167: Ti = Ta * T8; cannam@167: T7 = W[2]; cannam@167: T9 = T7 * T8; cannam@167: Rp[WS(rs, 1)] = FNMS(Ta, Th, T9); cannam@167: Rm[WS(rs, 1)] = FMA(T7, Th, Ti); 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, "hc2cb_4", twinstr, &GENUS, {16, 6, 6, 0} }; cannam@167: cannam@167: void X(codelet_hc2cb_4) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cb_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 -sign 1 -n 4 -dif -name hc2cb_4 -include rdft/scalar/hc2cb.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/hc2cb.h" cannam@167: cannam@167: static void hc2cb_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 T3, Ti, Tc, Tn, T6, Tm, Tf, Tj; cannam@167: { cannam@167: E T1, T2, Ta, Tb; cannam@167: T1 = Rp[0]; cannam@167: T2 = Rm[WS(rs, 1)]; cannam@167: T3 = T1 + T2; cannam@167: Ti = T1 - T2; cannam@167: Ta = Ip[0]; cannam@167: Tb = Im[WS(rs, 1)]; cannam@167: Tc = Ta - Tb; cannam@167: Tn = Ta + Tb; cannam@167: } cannam@167: { cannam@167: E T4, T5, Td, Te; cannam@167: T4 = Rp[WS(rs, 1)]; cannam@167: T5 = Rm[0]; cannam@167: T6 = T4 + T5; cannam@167: Tm = T4 - T5; cannam@167: Td = Ip[WS(rs, 1)]; cannam@167: Te = Im[0]; cannam@167: Tf = Td - Te; cannam@167: Tj = Td + Te; cannam@167: } cannam@167: Rp[0] = T3 + T6; cannam@167: Rm[0] = Tc + Tf; cannam@167: { cannam@167: E T8, Tg, T7, T9; cannam@167: T8 = T3 - T6; cannam@167: Tg = Tc - Tf; cannam@167: T7 = W[2]; cannam@167: T9 = W[3]; cannam@167: Rp[WS(rs, 1)] = FNMS(T9, Tg, T7 * T8); cannam@167: Rm[WS(rs, 1)] = FMA(T9, T8, T7 * Tg); cannam@167: } cannam@167: { cannam@167: E Tk, To, Th, Tl; cannam@167: Tk = Ti - Tj; cannam@167: To = Tm + Tn; cannam@167: Th = W[0]; cannam@167: Tl = W[1]; cannam@167: Ip[0] = FNMS(Tl, To, Th * Tk); cannam@167: Im[0] = FMA(Th, To, Tl * Tk); cannam@167: } cannam@167: { cannam@167: E Tq, Ts, Tp, Tr; cannam@167: Tq = Ti + Tj; cannam@167: Ts = Tn - Tm; cannam@167: Tp = W[4]; cannam@167: Tr = W[5]; cannam@167: Ip[WS(rs, 1)] = FNMS(Tr, Ts, Tp * Tq); cannam@167: Im[WS(rs, 1)] = FMA(Tp, Ts, Tr * Tq); 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, "hc2cb_4", twinstr, &GENUS, {16, 6, 6, 0} }; cannam@167: cannam@167: void X(codelet_hc2cb_4) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cb_4, &desc, HC2C_VIA_RDFT); cannam@167: } cannam@167: #endif