Chris@10: /* Chris@10: * Copyright (c) 2003, 2007-11 Matteo Frigo Chris@10: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology Chris@10: * Chris@10: * This program is free software; you can redistribute it and/or modify Chris@10: * it under the terms of the GNU General Public License as published by Chris@10: * the Free Software Foundation; either version 2 of the License, or Chris@10: * (at your option) any later version. Chris@10: * Chris@10: * This program is distributed in the hope that it will be useful, Chris@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@10: * GNU General Public License for more details. Chris@10: * Chris@10: * You should have received a copy of the GNU General Public License Chris@10: * along with this program; if not, write to the Free Software Chris@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@10: * Chris@10: */ Chris@10: Chris@10: /* This file was automatically generated --- DO NOT EDIT */ Chris@10: /* Generated on Sun Nov 25 07:41:07 EST 2012 */ Chris@10: Chris@10: #include "codelet-rdft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -name r2cb_12 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 16 FP multiplications, Chris@10: * (or, 22 additions, 0 multiplications, 16 fused multiply/add), Chris@10: * 31 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: static void r2cb_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { Chris@10: E Ts, Tr; Chris@10: { Chris@10: E Tz, Te, Tn, Tk, Tc, Tw, Ty, Th, T4, T3, Td, T5; Chris@10: { Chris@10: E T8, Tu, Tl, Tm, Tb, T9, Ta, T1, T2, Tv; Chris@10: T8 = Cr[WS(csr, 3)]; Chris@10: T9 = Cr[WS(csr, 5)]; Chris@10: Ta = Cr[WS(csr, 1)]; Chris@10: Tu = Ci[WS(csi, 3)]; Chris@10: Tl = Ci[WS(csi, 5)]; Chris@10: Tm = Ci[WS(csi, 1)]; Chris@10: Tb = T9 + Ta; Chris@10: Tz = T9 - Ta; Chris@10: Te = Ci[WS(csi, 4)]; Chris@10: Tn = Tl - Tm; Chris@10: Tv = Tl + Tm; Chris@10: Tk = FNMS(KP2_000000000, T8, Tb); Chris@10: Tc = T8 + Tb; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: Tw = Tu - Tv; Chris@10: Ty = FMA(KP2_000000000, Tu, Tv); Chris@10: Th = Ci[WS(csi, 2)]; Chris@10: T4 = Cr[WS(csr, 6)]; Chris@10: T3 = FMA(KP2_000000000, T2, T1); Chris@10: Td = T1 - T2; Chris@10: T5 = Cr[WS(csr, 2)]; Chris@10: } Chris@10: { Chris@10: E To, Tp, Tf, Tg, T6, TA, TC; Chris@10: To = FMA(KP1_732050807, Tn, Tk); Chris@10: Ts = FNMS(KP1_732050807, Tn, Tk); Chris@10: Tp = FNMS(KP1_732050807, Te, Td); Chris@10: Tf = FMA(KP1_732050807, Te, Td); Chris@10: Tg = T4 - T5; Chris@10: T6 = FMA(KP2_000000000, T5, T4); Chris@10: TA = FMA(KP1_732050807, Tz, Ty); Chris@10: TC = FNMS(KP1_732050807, Tz, Ty); Chris@10: { Chris@10: E Tt, T7, Ti, Tq, Tj, TB, Tx; Chris@10: Tt = T3 - T6; Chris@10: T7 = T3 + T6; Chris@10: Ti = FNMS(KP1_732050807, Th, Tg); Chris@10: Tq = FMA(KP1_732050807, Th, Tg); Chris@10: R0[0] = FMA(KP2_000000000, Tc, T7); Chris@10: R0[WS(rs, 3)] = FNMS(KP2_000000000, Tc, T7); Chris@10: Tj = Tf + Ti; Chris@10: TB = Tf - Ti; Chris@10: Tr = Tp + Tq; Chris@10: Tx = Tp - Tq; Chris@10: R1[WS(rs, 5)] = TB + TC; Chris@10: R1[WS(rs, 2)] = TB - TC; Chris@10: R0[WS(rs, 4)] = Tj - To; Chris@10: R0[WS(rs, 1)] = Tj + To; Chris@10: R1[WS(rs, 3)] = Tx + TA; Chris@10: R1[0] = Tx - TA; Chris@10: R1[WS(rs, 4)] = FNMS(KP2_000000000, Tw, Tt); Chris@10: R1[WS(rs, 1)] = FMA(KP2_000000000, Tw, Tt); Chris@10: } Chris@10: } Chris@10: } Chris@10: R0[WS(rs, 2)] = Tr - Ts; Chris@10: R0[WS(rs, 5)] = Tr + Ts; Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 12, "r2cb_12", {22, 0, 16, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_12) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_12, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -name r2cb_12 -include r2cb.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 10 FP multiplications, Chris@10: * (or, 34 additions, 6 multiplications, 4 fused multiply/add), Chris@10: * 25 stack variables, 2 constants, and 24 memory accesses Chris@10: */ Chris@10: #include "r2cb.h" Chris@10: Chris@10: static void r2cb_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) { Chris@10: E T8, Tb, Tm, TA, Tw, Tx, Tp, TB, T3, Tr, Tg, T6, Ts, Tk; Chris@10: { Chris@10: E T9, Ta, Tn, To; Chris@10: T8 = Cr[WS(csr, 3)]; Chris@10: T9 = Cr[WS(csr, 5)]; Chris@10: Ta = Cr[WS(csr, 1)]; Chris@10: Tb = T9 + Ta; Chris@10: Tm = FMS(KP2_000000000, T8, Tb); Chris@10: TA = KP1_732050807 * (T9 - Ta); Chris@10: Tw = Ci[WS(csi, 3)]; Chris@10: Tn = Ci[WS(csi, 5)]; Chris@10: To = Ci[WS(csi, 1)]; Chris@10: Tx = Tn + To; Chris@10: Tp = KP1_732050807 * (Tn - To); Chris@10: TB = FMA(KP2_000000000, Tw, Tx); Chris@10: } Chris@10: { Chris@10: E Tf, T1, T2, Td, Te; Chris@10: Te = Ci[WS(csi, 4)]; Chris@10: Tf = KP1_732050807 * Te; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: Td = T1 - T2; Chris@10: T3 = FMA(KP2_000000000, T2, T1); Chris@10: Tr = Td - Tf; Chris@10: Tg = Td + Tf; Chris@10: } Chris@10: { Chris@10: E Tj, T4, T5, Th, Ti; Chris@10: Ti = Ci[WS(csi, 2)]; Chris@10: Tj = KP1_732050807 * Ti; Chris@10: T4 = Cr[WS(csr, 6)]; Chris@10: T5 = Cr[WS(csr, 2)]; Chris@10: Th = T4 - T5; Chris@10: T6 = FMA(KP2_000000000, T5, T4); Chris@10: Ts = Th + Tj; Chris@10: Tk = Th - Tj; Chris@10: } Chris@10: { Chris@10: E T7, Tc, Tz, TC; Chris@10: T7 = T3 + T6; Chris@10: Tc = KP2_000000000 * (T8 + Tb); Chris@10: R0[WS(rs, 3)] = T7 - Tc; Chris@10: R0[0] = T7 + Tc; Chris@10: { Chris@10: E Tl, Tq, TD, TE; Chris@10: Tl = Tg + Tk; Chris@10: Tq = Tm - Tp; Chris@10: R0[WS(rs, 1)] = Tl - Tq; Chris@10: R0[WS(rs, 4)] = Tl + Tq; Chris@10: TD = Tg - Tk; Chris@10: TE = TB - TA; Chris@10: R1[WS(rs, 2)] = TD - TE; Chris@10: R1[WS(rs, 5)] = TD + TE; Chris@10: } Chris@10: Tz = Tr - Ts; Chris@10: TC = TA + TB; Chris@10: R1[0] = Tz - TC; Chris@10: R1[WS(rs, 3)] = Tz + TC; Chris@10: { Chris@10: E Tv, Ty, Tt, Tu; Chris@10: Tv = T3 - T6; Chris@10: Ty = KP2_000000000 * (Tw - Tx); Chris@10: R1[WS(rs, 4)] = Tv - Ty; Chris@10: R1[WS(rs, 1)] = Tv + Ty; Chris@10: Tt = Tr + Ts; Chris@10: Tu = Tm + Tp; Chris@10: R0[WS(rs, 5)] = Tt - Tu; Chris@10: R0[WS(rs, 2)] = Tt + Tu; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 12, "r2cb_12", {34, 6, 4, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cb_12) (planner *p) { Chris@10: X(kr2c_register) (p, r2cb_12, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */