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 07:41:34 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 8 -name r2cbIII_8 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 22 FP additions, 12 FP multiplications, Chris@10: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), Chris@10: * 23 stack variables, 4 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_8(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(KP414213562, +0.414213562373095048801688724209698078569671875); Chris@10: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); Chris@10: DK(KP1_414213562, +1.414213562373095048801688724209698078569671875); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E T4, T7, T3, Tl, Tf, T5, T8, T9, T6, Tc; Chris@10: { Chris@10: E T1, T2, Td, Te; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 3)]; Chris@10: Td = Ci[0]; Chris@10: Te = Ci[WS(csi, 3)]; Chris@10: T4 = Cr[WS(csr, 2)]; Chris@10: T7 = T1 - T2; Chris@10: T3 = T1 + T2; Chris@10: Tl = Te - Td; Chris@10: Tf = Td + Te; Chris@10: T5 = Cr[WS(csr, 1)]; Chris@10: T8 = Ci[WS(csi, 2)]; Chris@10: T9 = Ci[WS(csi, 1)]; Chris@10: } Chris@10: T6 = T4 + T5; Chris@10: Tc = T4 - T5; Chris@10: { Chris@10: E Ta, Tk, Tg, Th; Chris@10: Ta = T8 + T9; Chris@10: Tk = T8 - T9; Chris@10: Tg = Tc + Tf; Chris@10: Th = Tc - Tf; Chris@10: { Chris@10: E Tj, Tm, Tb, Ti; Chris@10: Tj = T3 - T6; Chris@10: R0[0] = KP2_000000000 * (T3 + T6); Chris@10: Tm = Tk + Tl; Chris@10: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); Chris@10: Tb = T7 - Ta; Chris@10: Ti = T7 + Ta; Chris@10: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); Chris@10: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); Chris@10: R1[WS(rs, 3)] = -(KP1_847759065 * (FNMS(KP414213562, Th, Ti))); Chris@10: R1[WS(rs, 1)] = KP1_847759065 * (FMA(KP414213562, Ti, Th)); Chris@10: R1[WS(rs, 2)] = -(KP1_847759065 * (FMA(KP414213562, Tb, Tg))); Chris@10: R1[0] = KP1_847759065 * (FNMS(KP414213562, Tg, Tb)); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_8, &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 8 -name r2cbIII_8 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 22 FP additions, 12 FP multiplications, Chris@10: * (or, 18 additions, 8 multiplications, 4 fused multiply/add), Chris@10: * 19 stack variables, 4 constants, and 16 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_8(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_414213562, +1.414213562373095048801688724209698078569671875); Chris@10: DK(KP765366864, +0.765366864730179543456919968060797733522689125); Chris@10: DK(KP1_847759065, +1.847759065022573512256366378793576573644833252); 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(32, rs), MAKE_VOLATILE_STRIDE(32, csr), MAKE_VOLATILE_STRIDE(32, csi)) { Chris@10: E T3, T7, Tf, Tl, T6, Tc, Ta, Tk, Tb, Tg; Chris@10: { Chris@10: E T1, T2, Td, Te; Chris@10: T1 = Cr[0]; Chris@10: T2 = Cr[WS(csr, 3)]; Chris@10: T3 = T1 + T2; Chris@10: T7 = T1 - T2; Chris@10: Td = Ci[0]; Chris@10: Te = Ci[WS(csi, 3)]; Chris@10: Tf = Td + Te; Chris@10: Tl = Te - Td; Chris@10: } Chris@10: { Chris@10: E T4, T5, T8, T9; Chris@10: T4 = Cr[WS(csr, 2)]; Chris@10: T5 = Cr[WS(csr, 1)]; Chris@10: T6 = T4 + T5; Chris@10: Tc = T4 - T5; Chris@10: T8 = Ci[WS(csi, 2)]; Chris@10: T9 = Ci[WS(csi, 1)]; Chris@10: Ta = T8 + T9; Chris@10: Tk = T8 - T9; Chris@10: } Chris@10: R0[0] = KP2_000000000 * (T3 + T6); Chris@10: R0[WS(rs, 2)] = KP2_000000000 * (Tl - Tk); Chris@10: Tb = T7 - Ta; Chris@10: Tg = Tc + Tf; Chris@10: R1[0] = FNMS(KP765366864, Tg, KP1_847759065 * Tb); Chris@10: R1[WS(rs, 2)] = -(FMA(KP765366864, Tb, KP1_847759065 * Tg)); Chris@10: { Chris@10: E Th, Ti, Tj, Tm; Chris@10: Th = T7 + Ta; Chris@10: Ti = Tc - Tf; Chris@10: R1[WS(rs, 1)] = FMA(KP765366864, Th, KP1_847759065 * Ti); Chris@10: R1[WS(rs, 3)] = FNMS(KP1_847759065, Th, KP765366864 * Ti); Chris@10: Tj = T3 - T6; Chris@10: Tm = Tk + Tl; Chris@10: R0[WS(rs, 1)] = KP1_414213562 * (Tj + Tm); Chris@10: R0[WS(rs, 3)] = KP1_414213562 * (Tm - Tj); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 8, "r2cbIII_8", {18, 8, 4, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_8) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_8, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */