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 07:41:34 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 8 -name r2cbIII_8 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 22 FP additions, 12 FP multiplications, cannam@95: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), cannam@95: * 23 stack variables, 4 constants, and 16 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_8(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(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { cannam@95: E T4, T7, T3, Tl, Tf, T5, T8, T9, T6, Tc; cannam@95: { cannam@95: E T1, T2, Td, Te; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 3)]; cannam@95: Td = Ci[0]; cannam@95: Te = Ci[WS(csi, 3)]; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: T7 = T1 - T2; cannam@95: T3 = T1 + T2; cannam@95: Tl = Te - Td; cannam@95: Tf = Td + Te; cannam@95: T5 = Cr[WS(csr, 1)]; cannam@95: T8 = Ci[WS(csi, 2)]; cannam@95: T9 = Ci[WS(csi, 1)]; cannam@95: } cannam@95: T6 = T4 + T5; cannam@95: Tc = T4 - T5; cannam@95: { cannam@95: E Ta, Tk, Tg, Th; cannam@95: Ta = T8 + T9; cannam@95: Tk = T8 - T9; cannam@95: Tg = Tc + Tf; cannam@95: Th = Tc - Tf; cannam@95: { cannam@95: E Tj, Tm, Tb, Ti; cannam@95: Tj = T3 - T6; cannam@95: R0[0] = KP2_000000000 * (T3 + T6); cannam@95: Tm = Tk + Tl; cannam@95: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); cannam@95: Tb = T7 - Ta; cannam@95: Ti = T7 + Ta; cannam@95: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); cannam@95: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); cannam@95: R1[WS(rs, 3)] = -(KP1_847759065 * (FNMS(KP414213562, Th, Ti))); cannam@95: R1[WS(rs, 1)] = KP1_847759065 * (FMA(KP414213562, Ti, Th)); cannam@95: R1[WS(rs, 2)] = -(KP1_847759065 * (FMA(KP414213562, Tb, Tg))); cannam@95: R1[0] = KP1_847759065 * (FNMS(KP414213562, Tg, Tb)); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_8) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_8, &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 8 -name r2cbIII_8 -dft-III -include r2cbIII.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 22 FP additions, 12 FP multiplications, cannam@95: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), cannam@95: * 19 stack variables, 4 constants, and 16 memory accesses cannam@95: */ cannam@95: #include "r2cbIII.h" cannam@95: cannam@95: static void r2cbIII_8(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_414213562, +1.414213562373095048801688724209698078569671875); cannam@95: DK(KP765366864, +0.765366864730179543456919968060797733522689125); cannam@95: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@95: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { cannam@95: E T3, T7, Tf, Tl, T6, Tc, Ta, Tk, Tb, Tg; cannam@95: { cannam@95: E T1, T2, Td, Te; cannam@95: T1 = Cr[0]; cannam@95: T2 = Cr[WS(csr, 3)]; cannam@95: T3 = T1 + T2; cannam@95: T7 = T1 - T2; cannam@95: Td = Ci[0]; cannam@95: Te = Ci[WS(csi, 3)]; cannam@95: Tf = Td + Te; cannam@95: Tl = Te - Td; cannam@95: } cannam@95: { cannam@95: E T4, T5, T8, T9; cannam@95: T4 = Cr[WS(csr, 2)]; cannam@95: T5 = Cr[WS(csr, 1)]; cannam@95: T6 = T4 + T5; cannam@95: Tc = T4 - T5; cannam@95: T8 = Ci[WS(csi, 2)]; cannam@95: T9 = Ci[WS(csi, 1)]; cannam@95: Ta = T8 + T9; cannam@95: Tk = T8 - T9; cannam@95: } cannam@95: R0[0] = KP2_000000000 * (T3 + T6); cannam@95: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); cannam@95: Tb = T7 - Ta; cannam@95: Tg = Tc + Tf; cannam@95: R1[0] = FNMS(KP765366864, Tg, KP1_847759065 * Tb); cannam@95: R1[WS(rs, 2)] = -(FMA(KP765366864, Tb, KP1_847759065 * Tg)); cannam@95: { cannam@95: E Th, Ti, Tj, Tm; cannam@95: Th = T7 + Ta; cannam@95: Ti = Tc - Tf; cannam@95: R1[WS(rs, 1)] = FMA(KP765366864, Th, KP1_847759065 * Ti); cannam@95: R1[WS(rs, 3)] = FNMS(KP1_847759065, Th, KP765366864 * Ti); cannam@95: Tj = T3 - T6; cannam@95: Tm = Tk + Tl; cannam@95: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); cannam@95: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: cannam@95: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; cannam@95: cannam@95: void X(codelet_r2cbIII_8) (planner *p) { cannam@95: X(kr2c_register) (p, r2cbIII_8, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */