cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:50:38 EDT 2016 */ cannam@127: cannam@127: #include "codelet-rdft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* 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@127: cannam@127: /* cannam@127: * This function contains 22 FP additions, 12 FP multiplications, cannam@127: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), cannam@127: * 23 stack variables, 4 constants, and 16 memory accesses cannam@127: */ cannam@127: #include "r2cbIII.h" cannam@127: cannam@127: 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@127: { cannam@127: DK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@127: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@127: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@127: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@127: { cannam@127: INT i; cannam@127: 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@127: E T4, T7, T3, Tl, Tf, T5, T8, T9, T6, Tc; cannam@127: { cannam@127: E T1, T2, Td, Te; cannam@127: T1 = Cr[0]; cannam@127: T2 = Cr[WS(csr, 3)]; cannam@127: Td = Ci[0]; cannam@127: Te = Ci[WS(csi, 3)]; cannam@127: T4 = Cr[WS(csr, 2)]; cannam@127: T7 = T1 - T2; cannam@127: T3 = T1 + T2; cannam@127: Tl = Te - Td; cannam@127: Tf = Td + Te; cannam@127: T5 = Cr[WS(csr, 1)]; cannam@127: T8 = Ci[WS(csi, 2)]; cannam@127: T9 = Ci[WS(csi, 1)]; cannam@127: } cannam@127: T6 = T4 + T5; cannam@127: Tc = T4 - T5; cannam@127: { cannam@127: E Ta, Tk, Tg, Th; cannam@127: Ta = T8 + T9; cannam@127: Tk = T8 - T9; cannam@127: Tg = Tc + Tf; cannam@127: Th = Tc - Tf; cannam@127: { cannam@127: E Tj, Tm, Tb, Ti; cannam@127: Tj = T3 - T6; cannam@127: R0[0] = KP2_000000000 * (T3 + T6); cannam@127: Tm = Tk + Tl; cannam@127: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); cannam@127: Tb = T7 - Ta; cannam@127: Ti = T7 + Ta; cannam@127: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); cannam@127: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); cannam@127: R1[WS(rs, 3)] = -(KP1_847759065 * (FNMS(KP414213562, Th, Ti))); cannam@127: R1[WS(rs, 1)] = KP1_847759065 * (FMA(KP414213562, Ti, Th)); cannam@127: R1[WS(rs, 2)] = -(KP1_847759065 * (FMA(KP414213562, Tb, Tg))); cannam@127: R1[0] = KP1_847759065 * (FNMS(KP414213562, Tg, Tb)); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cbIII_8) (planner *p) { cannam@127: X(kr2c_register) (p, r2cbIII_8, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* 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@127: cannam@127: /* cannam@127: * This function contains 22 FP additions, 12 FP multiplications, cannam@127: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), cannam@127: * 19 stack variables, 4 constants, and 16 memory accesses cannam@127: */ cannam@127: #include "r2cbIII.h" cannam@127: cannam@127: 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@127: { cannam@127: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); cannam@127: DK(KP765366864, +0.765366864730179543456919968060797733522689125); cannam@127: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); cannam@127: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@127: { cannam@127: INT i; cannam@127: 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@127: E T3, T7, Tf, Tl, T6, Tc, Ta, Tk, Tb, Tg; cannam@127: { cannam@127: E T1, T2, Td, Te; cannam@127: T1 = Cr[0]; cannam@127: T2 = Cr[WS(csr, 3)]; cannam@127: T3 = T1 + T2; cannam@127: T7 = T1 - T2; cannam@127: Td = Ci[0]; cannam@127: Te = Ci[WS(csi, 3)]; cannam@127: Tf = Td + Te; cannam@127: Tl = Te - Td; cannam@127: } cannam@127: { cannam@127: E T4, T5, T8, T9; cannam@127: T4 = Cr[WS(csr, 2)]; cannam@127: T5 = Cr[WS(csr, 1)]; cannam@127: T6 = T4 + T5; cannam@127: Tc = T4 - T5; cannam@127: T8 = Ci[WS(csi, 2)]; cannam@127: T9 = Ci[WS(csi, 1)]; cannam@127: Ta = T8 + T9; cannam@127: Tk = T8 - T9; cannam@127: } cannam@127: R0[0] = KP2_000000000 * (T3 + T6); cannam@127: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); cannam@127: Tb = T7 - Ta; cannam@127: Tg = Tc + Tf; cannam@127: R1[0] = FNMS(KP765366864, Tg, KP1_847759065 * Tb); cannam@127: R1[WS(rs, 2)] = -(FMA(KP765366864, Tb, KP1_847759065 * Tg)); cannam@127: { cannam@127: E Th, Ti, Tj, Tm; cannam@127: Th = T7 + Ta; cannam@127: Ti = Tc - Tf; cannam@127: R1[WS(rs, 1)] = FMA(KP765366864, Th, KP1_847759065 * Ti); cannam@127: R1[WS(rs, 3)] = FNMS(KP1_847759065, Th, KP765366864 * Ti); cannam@127: Tj = T3 - T6; cannam@127: Tm = Tk + Tl; cannam@127: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); cannam@127: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cbIII_8) (planner *p) { cannam@127: X(kr2c_register) (p, r2cbIII_8, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */