cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:41:07 EST 2012 */ cannam@95: cannam@95: #include "codelet-rdft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* 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 */ cannam@95: cannam@95: /* cannam@95: * This function contains 34 FP additions, 20 FP multiplications, cannam@95: * (or, 14 additions, 0 multiplications, 20 fused multiply/add), cannam@95: * 30 stack variables, 5 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); cannam@95: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E Tb, T3, Tc, T6, Tq, To, Ty, Tw, Td, T9; cannam@95: { cannam@95: E Tu, Tn, T7, Tv, Tk, T8; cannam@95: { cannam@95: E T1, T2, Tl, Tm; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 5)]; cannam@95: Tl = Ci[WS(csi, 2)]; cannam@95: Tm = Ci[WS(csi, 3)]; cannam@95: { cannam@95: E Ti, Tj, T4, T5; cannam@95: Ti = Ci[WS(csi, 4)]; cannam@95: Tb = T1 + T2; cannam@95: T3 = T1 - T2; cannam@95: Tu = Tl + Tm; cannam@95: Tn = Tl - Tm; cannam@95: Tj = Ci[WS(csi, 1)]; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: T7 = Cr[WS(csr, 4)]; cannam@95: Tv = Ti + Tj; cannam@95: Tk = Ti - Tj; cannam@95: Tc = T4 + T5; cannam@95: T6 = T4 - T5; cannam@95: T8 = Cr[WS(csr, 1)]; cannam@95: } cannam@95: } cannam@95: Tq = FMA(KP618033988, Tk, Tn); cannam@95: To = FNMS(KP618033988, Tn, Tk); cannam@95: Ty = FNMS(KP618033988, Tu, Tv); cannam@95: Tw = FMA(KP618033988, Tv, Tu); cannam@95: Td = T7 + T8; cannam@95: T9 = T7 - T8; cannam@95: } cannam@95: { cannam@95: E Te, Tg, Ta, Ts, Tf, Tr; cannam@95: Te = Tc + Td; cannam@95: Tg = Tc - Td; cannam@95: Ta = T6 + T9; cannam@95: Ts = T6 - T9; cannam@95: Tf = FNMS(KP500000000, Te, Tb); cannam@95: R0[0] = FMA(KP2_000000000, Te, Tb); cannam@95: Tr = FNMS(KP500000000, Ta, T3); cannam@95: R1[WS(rs, 2)] = FMA(KP2_000000000, Ta, T3); cannam@95: { cannam@95: E Th, Tp, Tt, Tx; cannam@95: Th = FNMS(KP1_118033988, Tg, Tf); cannam@95: Tp = FMA(KP1_118033988, Tg, Tf); cannam@95: Tt = FMA(KP1_118033988, Ts, Tr); cannam@95: Tx = FNMS(KP1_118033988, Ts, Tr); cannam@95: R0[WS(rs, 3)] = FNMS(KP1_902113032, Tq, Tp); cannam@95: R0[WS(rs, 2)] = FMA(KP1_902113032, Tq, Tp); cannam@95: R0[WS(rs, 1)] = FMA(KP1_902113032, To, Th); cannam@95: R0[WS(rs, 4)] = FNMS(KP1_902113032, To, Th); cannam@95: R1[WS(rs, 1)] = FNMS(KP1_902113032, Ty, Tx); cannam@95: R1[WS(rs, 3)] = FMA(KP1_902113032, Ty, Tx); cannam@95: R1[WS(rs, 4)] = FMA(KP1_902113032, Tw, Tt); cannam@95: R1[0] = FNMS(KP1_902113032, Tw, Tt); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cb_10", {14, 0, 20, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_10, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -name r2cb_10 -include r2cb.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 34 FP additions, 14 FP multiplications, cannam@95: * (or, 26 additions, 6 multiplications, 8 fused multiply/add), cannam@95: * 26 stack variables, 5 constants, and 20 memory accesses cannam@95: */ cannam@95: #include "r2cb.h" cannam@95: cannam@95: static void r2cb_10(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); cannam@95: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@95: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); cannam@95: { cannam@95: INT i; cannam@95: 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)) { cannam@95: E T3, Tb, Tn, Tv, Tk, Tu, Ta, Ts, Te, Tg, Ti, Tj; cannam@95: { cannam@95: E T1, T2, Tl, Tm; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 5)]; cannam@95: T3 = T1 - T2; cannam@95: Tb = T1 + T2; cannam@95: Tl = Ci[WS(csi, 4)]; cannam@95: Tm = Ci[WS(csi, 1)]; cannam@95: Tn = Tl - Tm; cannam@95: Tv = Tl + Tm; cannam@95: } cannam@95: Ti = Ci[WS(csi, 2)]; cannam@95: Tj = Ci[WS(csi, 3)]; cannam@95: Tk = Ti - Tj; cannam@95: Tu = Ti + Tj; cannam@95: { cannam@95: E T6, Tc, T9, Td; cannam@95: { cannam@95: E T4, T5, T7, T8; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: T5 = Cr[WS(csr, 3)]; cannam@95: T6 = T4 - T5; cannam@95: Tc = T4 + T5; cannam@95: T7 = Cr[WS(csr, 4)]; cannam@95: T8 = Cr[WS(csr, 1)]; cannam@95: T9 = T7 - T8; cannam@95: Td = T7 + T8; cannam@95: } cannam@95: Ta = T6 + T9; cannam@95: Ts = KP1_118033988 * (T6 - T9); cannam@95: Te = Tc + Td; cannam@95: Tg = KP1_118033988 * (Tc - Td); cannam@95: } cannam@95: R1[WS(rs, 2)] = FMA(KP2_000000000, Ta, T3); cannam@95: R0[0] = FMA(KP2_000000000, Te, Tb); cannam@95: { cannam@95: E To, Tq, Th, Tp, Tf; cannam@95: To = FNMS(KP1_902113032, Tn, KP1_175570504 * Tk); cannam@95: Tq = FMA(KP1_902113032, Tk, KP1_175570504 * Tn); cannam@95: Tf = FNMS(KP500000000, Te, Tb); cannam@95: Th = Tf - Tg; cannam@95: Tp = Tg + Tf; cannam@95: R0[WS(rs, 1)] = Th - To; cannam@95: R0[WS(rs, 2)] = Tp + Tq; cannam@95: R0[WS(rs, 4)] = Th + To; cannam@95: R0[WS(rs, 3)] = Tp - Tq; cannam@95: } cannam@95: { cannam@95: E Tw, Ty, Tt, Tx, Tr; cannam@95: Tw = FNMS(KP1_902113032, Tv, KP1_175570504 * Tu); cannam@95: Ty = FMA(KP1_902113032, Tu, KP1_175570504 * Tv); cannam@95: Tr = FNMS(KP500000000, Ta, T3); cannam@95: Tt = Tr - Ts; cannam@95: Tx = Ts + Tr; cannam@95: R1[WS(rs, 3)] = Tt - Tw; cannam@95: R1[WS(rs, 4)] = Tx + Ty; cannam@95: R1[WS(rs, 1)] = Tt + Tw; cannam@95: R1[0] = Tx - Ty; cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 10, "r2cb_10", {26, 6, 8, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cb_10) (planner *p) { cannam@95: X(kr2c_register) (p, r2cb_10, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */