Chris@42: /* Chris@42: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@42: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@42: * Chris@42: * This program is free software; you can redistribute it and/or modify Chris@42: * it under the terms of the GNU General Public License as published by Chris@42: * the Free Software Foundation; either version 2 of the License, or Chris@42: * (at your option) any later version. Chris@42: * Chris@42: * This program is distributed in the hope that it will be useful, Chris@42: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@42: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@42: * GNU General Public License for more details. Chris@42: * Chris@42: * You should have received a copy of the GNU General Public License Chris@42: * along with this program; if not, write to the Free Software Chris@42: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@42: * Chris@42: */ Chris@42: Chris@42: /* This file was automatically generated --- DO NOT EDIT */ Chris@42: /* Generated on Sat Jul 30 16:49:26 EDT 2016 */ Chris@42: Chris@42: #include "codelet-rdft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cb_10 -include r2cb.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 34 FP additions, 20 FP multiplications, Chris@42: * (or, 14 additions, 0 multiplications, 20 fused multiply/add), Chris@42: * 30 stack variables, 5 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "r2cb.h" Chris@42: Chris@42: static void r2cb_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@42: { Chris@42: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@42: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@42: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@42: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@42: { Chris@42: INT i; Chris@42: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@42: E Tb, T3, Tc, T6, Tq, To, Ty, Tw, Td, T9; Chris@42: { Chris@42: E Tu, Tn, T7, Tv, Tk, T8; Chris@42: { Chris@42: E T1, T2, Tl, Tm; Chris@42: T1 = Cr[0]; Chris@42: T2 = Cr[WS(csr, 5)]; Chris@42: Tl = Ci[WS(csi, 2)]; Chris@42: Tm = Ci[WS(csi, 3)]; Chris@42: { Chris@42: E Ti, Tj, T4, T5; Chris@42: Ti = Ci[WS(csi, 4)]; Chris@42: Tb = T1 + T2; Chris@42: T3 = T1 - T2; Chris@42: Tu = Tl + Tm; Chris@42: Tn = Tl - Tm; Chris@42: Tj = Ci[WS(csi, 1)]; Chris@42: T4 = Cr[WS(csr, 2)]; Chris@42: T5 = Cr[WS(csr, 3)]; Chris@42: T7 = Cr[WS(csr, 4)]; Chris@42: Tv = Ti + Tj; Chris@42: Tk = Ti - Tj; Chris@42: Tc = T4 + T5; Chris@42: T6 = T4 - T5; Chris@42: T8 = Cr[WS(csr, 1)]; Chris@42: } Chris@42: } Chris@42: Tq = FMA(KP618033988, Tk, Tn); Chris@42: To = FNMS(KP618033988, Tn, Tk); Chris@42: Ty = FNMS(KP618033988, Tu, Tv); Chris@42: Tw = FMA(KP618033988, Tv, Tu); Chris@42: Td = T7 + T8; Chris@42: T9 = T7 - T8; Chris@42: } Chris@42: { Chris@42: E Te, Tg, Ta, Ts, Tf, Tr; Chris@42: Te = Tc + Td; Chris@42: Tg = Tc - Td; Chris@42: Ta = T6 + T9; Chris@42: Ts = T6 - T9; Chris@42: Tf = FNMS(KP500000000, Te, Tb); Chris@42: R0[0] = FMA(KP2_000000000, Te, Tb); Chris@42: Tr = FNMS(KP500000000, Ta, T3); Chris@42: R1[WS(rs, 2)] = FMA(KP2_000000000, Ta, T3); Chris@42: { Chris@42: E Th, Tp, Tt, Tx; Chris@42: Th = FNMS(KP1_118033988, Tg, Tf); Chris@42: Tp = FMA(KP1_118033988, Tg, Tf); Chris@42: Tt = FMA(KP1_118033988, Ts, Tr); Chris@42: Tx = FNMS(KP1_118033988, Ts, Tr); Chris@42: R0[WS(rs, 3)] = FNMS(KP1_902113032, Tq, Tp); Chris@42: R0[WS(rs, 2)] = FMA(KP1_902113032, Tq, Tp); Chris@42: R0[WS(rs, 1)] = FMA(KP1_902113032, To, Th); Chris@42: R0[WS(rs, 4)] = FNMS(KP1_902113032, To, Th); Chris@42: R1[WS(rs, 1)] = FNMS(KP1_902113032, Ty, Tx); Chris@42: R1[WS(rs, 3)] = FMA(KP1_902113032, Ty, Tx); Chris@42: R1[WS(rs, 4)] = FMA(KP1_902113032, Tw, Tt); Chris@42: R1[0] = FNMS(KP1_902113032, Tw, Tt); Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const kr2c_desc desc = { 10, "r2cb_10", {14, 0, 20, 0}, &GENUS }; Chris@42: Chris@42: void X(codelet_r2cb_10) (planner *p) { Chris@42: X(kr2c_register) (p, r2cb_10, &desc); Chris@42: } Chris@42: Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cb_10 -include r2cb.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 34 FP additions, 14 FP multiplications, Chris@42: * (or, 26 additions, 6 multiplications, 8 fused multiply/add), Chris@42: * 26 stack variables, 5 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "r2cb.h" Chris@42: Chris@42: static void r2cb_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@42: { Chris@42: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@42: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@42: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@42: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@42: { Chris@42: INT i; Chris@42: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(40, rs), MAKE_VOLATILE_STRIDE(40, csr), MAKE_VOLATILE_STRIDE(40, csi)) { Chris@42: E T3, Tb, Tn, Tv, Tk, Tu, Ta, Ts, Te, Tg, Ti, Tj; Chris@42: { Chris@42: E T1, T2, Tl, Tm; Chris@42: T1 = Cr[0]; Chris@42: T2 = Cr[WS(csr, 5)]; Chris@42: T3 = T1 - T2; Chris@42: Tb = T1 + T2; Chris@42: Tl = Ci[WS(csi, 4)]; Chris@42: Tm = Ci[WS(csi, 1)]; Chris@42: Tn = Tl - Tm; Chris@42: Tv = Tl + Tm; Chris@42: } Chris@42: Ti = Ci[WS(csi, 2)]; Chris@42: Tj = Ci[WS(csi, 3)]; Chris@42: Tk = Ti - Tj; Chris@42: Tu = Ti + Tj; Chris@42: { Chris@42: E T6, Tc, T9, Td; Chris@42: { Chris@42: E T4, T5, T7, T8; Chris@42: T4 = Cr[WS(csr, 2)]; Chris@42: T5 = Cr[WS(csr, 3)]; Chris@42: T6 = T4 - T5; Chris@42: Tc = T4 + T5; Chris@42: T7 = Cr[WS(csr, 4)]; Chris@42: T8 = Cr[WS(csr, 1)]; Chris@42: T9 = T7 - T8; Chris@42: Td = T7 + T8; Chris@42: } Chris@42: Ta = T6 + T9; Chris@42: Ts = KP1_118033988 * (T6 - T9); Chris@42: Te = Tc + Td; Chris@42: Tg = KP1_118033988 * (Tc - Td); Chris@42: } Chris@42: R1[WS(rs, 2)] = FMA(KP2_000000000, Ta, T3); Chris@42: R0[0] = FMA(KP2_000000000, Te, Tb); Chris@42: { Chris@42: E To, Tq, Th, Tp, Tf; Chris@42: To = FNMS(KP1_902113032, Tn, KP1_175570504 * Tk); Chris@42: Tq = FMA(KP1_902113032, Tk, KP1_175570504 * Tn); Chris@42: Tf = FNMS(KP500000000, Te, Tb); Chris@42: Th = Tf - Tg; Chris@42: Tp = Tg + Tf; Chris@42: R0[WS(rs, 1)] = Th - To; Chris@42: R0[WS(rs, 2)] = Tp + Tq; Chris@42: R0[WS(rs, 4)] = Th + To; Chris@42: R0[WS(rs, 3)] = Tp - Tq; Chris@42: } Chris@42: { Chris@42: E Tw, Ty, Tt, Tx, Tr; Chris@42: Tw = FNMS(KP1_902113032, Tv, KP1_175570504 * Tu); Chris@42: Ty = FMA(KP1_902113032, Tu, KP1_175570504 * Tv); Chris@42: Tr = FNMS(KP500000000, Ta, T3); Chris@42: Tt = Tr - Ts; Chris@42: Tx = Ts + Tr; Chris@42: R1[WS(rs, 3)] = Tt - Tw; Chris@42: R1[WS(rs, 4)] = Tx + Ty; Chris@42: R1[WS(rs, 1)] = Tt + Tw; Chris@42: R1[0] = Tx - Ty; Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: Chris@42: static const kr2c_desc desc = { 10, "r2cb_10", {26, 6, 8, 0}, &GENUS }; Chris@42: Chris@42: void X(codelet_r2cb_10) (planner *p) { Chris@42: X(kr2c_register) (p, r2cb_10, &desc); Chris@42: } Chris@42: Chris@42: #endif /* HAVE_FMA */