Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:04:54 EDT 2018 */ Chris@82: Chris@82: #include "dft/codelet-dft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 5 -name n1bv_5 -include dft/simd/n1b.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 16 FP additions, 11 FP multiplications, Chris@82: * (or, 7 additions, 2 multiplications, 9 fused multiply/add), Chris@82: * 18 stack variables, 4 constants, and 10 memory accesses Chris@82: */ Chris@82: #include "dft/simd/n1b.h" Chris@82: Chris@82: static void n1bv_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: { Chris@82: INT i; Chris@82: const R *xi; Chris@82: R *xo; Chris@82: xi = ii; Chris@82: xo = io; Chris@82: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(10, is), MAKE_VOLATILE_STRIDE(10, os)) { Chris@82: V T1, T8, Td, Ta, Tc; Chris@82: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@82: { Chris@82: V T2, T3, T4, T5, T6, T7; Chris@82: T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@82: T3 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@82: T4 = VADD(T2, T3); Chris@82: T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@82: T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@82: T7 = VADD(T5, T6); Chris@82: T8 = VADD(T4, T7); Chris@82: Td = VSUB(T5, T6); Chris@82: Ta = VSUB(T4, T7); Chris@82: Tc = VSUB(T2, T3); Chris@82: } Chris@82: ST(&(xo[0]), VADD(T1, T8), ovs, &(xo[0])); Chris@82: { Chris@82: V Te, Tg, Tb, Tf, T9; Chris@82: Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc)); Chris@82: Tg = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td)); Chris@82: T9 = VFNMS(LDK(KP250000000), T8, T1); Chris@82: Tb = VFMA(LDK(KP559016994), Ta, T9); Chris@82: Tf = VFNMS(LDK(KP559016994), Ta, T9); Chris@82: ST(&(xo[WS(os, 1)]), VFMAI(Te, Tb), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 3)]), VFMAI(Tg, Tf), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 4)]), VFNMSI(Te, Tb), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 2)]), VFNMSI(Tg, Tf), ovs, &(xo[0])); Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 5, XSIMD_STRING("n1bv_5"), {7, 2, 9, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_n1bv_5) (planner *p) { Chris@82: X(kdft_register) (p, n1bv_5, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 5 -name n1bv_5 -include dft/simd/n1b.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 16 FP additions, 6 FP multiplications, Chris@82: * (or, 13 additions, 3 multiplications, 3 fused multiply/add), Chris@82: * 18 stack variables, 4 constants, and 10 memory accesses Chris@82: */ Chris@82: #include "dft/simd/n1b.h" Chris@82: Chris@82: static void n1bv_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: { Chris@82: INT i; Chris@82: const R *xi; Chris@82: R *xo; Chris@82: xi = ii; Chris@82: xo = io; Chris@82: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(10, is), MAKE_VOLATILE_STRIDE(10, os)) { Chris@82: V Tb, T3, Tc, T6, Ta; Chris@82: Tb = LD(&(xi[0]), ivs, &(xi[0])); Chris@82: { Chris@82: V T1, T2, T8, T4, T5, T9; Chris@82: T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@82: T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@82: T8 = VADD(T1, T2); Chris@82: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@82: T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@82: T9 = VADD(T4, T5); Chris@82: T3 = VSUB(T1, T2); Chris@82: Tc = VADD(T8, T9); Chris@82: T6 = VSUB(T4, T5); Chris@82: Ta = VMUL(LDK(KP559016994), VSUB(T8, T9)); Chris@82: } Chris@82: ST(&(xo[0]), VADD(Tb, Tc), ovs, &(xo[0])); Chris@82: { Chris@82: V T7, Tf, Te, Tg, Td; Chris@82: T7 = VBYI(VFMA(LDK(KP951056516), T3, VMUL(LDK(KP587785252), T6))); Chris@82: Tf = VBYI(VFNMS(LDK(KP951056516), T6, VMUL(LDK(KP587785252), T3))); Chris@82: Td = VFNMS(LDK(KP250000000), Tc, Tb); Chris@82: Te = VADD(Ta, Td); Chris@82: Tg = VSUB(Td, Ta); Chris@82: ST(&(xo[WS(os, 1)]), VADD(T7, Te), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 3)]), VSUB(Tg, Tf), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 4)]), VSUB(Te, T7), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 2)]), VADD(Tf, Tg), ovs, &(xo[0])); Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 5, XSIMD_STRING("n1bv_5"), {13, 3, 3, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_n1bv_5) (planner *p) { Chris@82: X(kdft_register) (p, n1bv_5, &desc); Chris@82: } Chris@82: Chris@82: #endif