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 08:05:25 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fuv_7 -include dft/simd/t1fu.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 36 FP additions, 36 FP multiplications, Chris@82: * (or, 15 additions, 15 multiplications, 21 fused multiply/add), Chris@82: * 30 stack variables, 6 constants, and 14 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t1fu.h" Chris@82: Chris@82: static void t1fuv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP801937735, +0.801937735804838252472204639014890102331838324); Chris@82: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@82: DVK(KP554958132, +0.554958132087371191422194871006410481067288862); Chris@82: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@82: DVK(KP692021471, +0.692021471630095869627814897002069140197260599); Chris@82: DVK(KP356895867, +0.356895867892209443894399510021300583399127187); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ri; Chris@82: for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) { Chris@82: V T1, Tk, Tm, Tl, T6, Tg, Tb, Th, Tu, Tp; Chris@82: T1 = LD(&(x[0]), ms, &(x[0])); Chris@82: { Chris@82: V T3, T5, Tf, Td, Ta, T8; Chris@82: { Chris@82: V T2, T4, Te, Tc, T9, T7; Chris@82: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: T3 = BYTWJ(&(W[0]), T2); Chris@82: T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@82: T5 = BYTWJ(&(W[TWVL * 10]), T4); Chris@82: Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tf = BYTWJ(&(W[TWVL * 6]), Te); Chris@82: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: Td = BYTWJ(&(W[TWVL * 4]), Tc); Chris@82: T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: Ta = BYTWJ(&(W[TWVL * 8]), T9); Chris@82: T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: T8 = BYTWJ(&(W[TWVL * 2]), T7); Chris@82: } Chris@82: Tk = VSUB(T5, T3); Chris@82: Tm = VSUB(Ta, T8); Chris@82: Tl = VSUB(Tf, Td); Chris@82: T6 = VADD(T3, T5); Chris@82: Tg = VADD(Td, Tf); Chris@82: Tb = VADD(T8, Ta); Chris@82: Th = VFNMS(LDK(KP356895867), T6, Tg); Chris@82: Tu = VFNMS(LDK(KP356895867), Tg, Tb); Chris@82: Tp = VFNMS(LDK(KP356895867), Tb, T6); Chris@82: } Chris@82: ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); Chris@82: { Chris@82: V Tw, Ty, Tv, Tx; Chris@82: Tv = VFNMS(LDK(KP692021471), Tu, T6); Chris@82: Tw = VFNMS(LDK(KP900968867), Tv, T1); Chris@82: Tx = VFNMS(LDK(KP554958132), Tk, Tm); Chris@82: Ty = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tx, Tl)); Chris@82: ST(&(x[WS(rs, 4)]), VFNMSI(Ty, Tw), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 3)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)])); Chris@82: } Chris@82: { Chris@82: V Tj, To, Ti, Tn; Chris@82: Ti = VFNMS(LDK(KP692021471), Th, Tb); Chris@82: Tj = VFNMS(LDK(KP900968867), Ti, T1); Chris@82: Tn = VFMA(LDK(KP554958132), Tm, Tl); Chris@82: To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk)); Chris@82: ST(&(x[WS(rs, 5)]), VFNMSI(To, Tj), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 2)]), VFMAI(To, Tj), ms, &(x[0])); Chris@82: } Chris@82: { Chris@82: V Tr, Tt, Tq, Ts; Chris@82: Tq = VFNMS(LDK(KP692021471), Tp, Tg); Chris@82: Tr = VFNMS(LDK(KP900968867), Tq, T1); Chris@82: Ts = VFMA(LDK(KP554958132), Tl, Tk); Chris@82: Tt = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Ts, Tm)); Chris@82: ST(&(x[WS(rs, 6)]), VFNMSI(Tt, Tr), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 1)]), VFMAI(Tt, Tr), ms, &(x[WS(rs, 1)])); Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const tw_instr twinstr[] = { Chris@82: VTW(0, 1), Chris@82: VTW(0, 2), Chris@82: VTW(0, 3), Chris@82: VTW(0, 4), Chris@82: VTW(0, 5), Chris@82: VTW(0, 6), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 7, XSIMD_STRING("t1fuv_7"), twinstr, &GENUS, {15, 15, 21, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t1fuv_7) (planner *p) { Chris@82: X(kdft_dit_register) (p, t1fuv_7, &desc); Chris@82: } Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fuv_7 -include dft/simd/t1fu.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 36 FP additions, 30 FP multiplications, Chris@82: * (or, 24 additions, 18 multiplications, 12 fused multiply/add), Chris@82: * 21 stack variables, 6 constants, and 14 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t1fu.h" Chris@82: Chris@82: static void t1fuv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@82: DVK(KP222520933, +0.222520933956314404288902564496794759466355569); Chris@82: DVK(KP623489801, +0.623489801858733530525004884004239810632274731); Chris@82: DVK(KP781831482, +0.781831482468029808708444526674057750232334519); Chris@82: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@82: DVK(KP433883739, +0.433883739117558120475768332848358754609990728); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ri; Chris@82: for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) { Chris@82: V T1, Tg, Tj, T6, Ti, Tb, Tk, Tp, To; Chris@82: T1 = LD(&(x[0]), ms, &(x[0])); Chris@82: { Chris@82: V Td, Tf, Tc, Te; Chris@82: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: Td = BYTWJ(&(W[TWVL * 4]), Tc); Chris@82: Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tf = BYTWJ(&(W[TWVL * 6]), Te); Chris@82: Tg = VADD(Td, Tf); Chris@82: Tj = VSUB(Tf, Td); Chris@82: } Chris@82: { Chris@82: V T3, T5, T2, T4; Chris@82: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: T3 = BYTWJ(&(W[0]), T2); Chris@82: T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@82: T5 = BYTWJ(&(W[TWVL * 10]), T4); Chris@82: T6 = VADD(T3, T5); Chris@82: Ti = VSUB(T5, T3); Chris@82: } Chris@82: { Chris@82: V T8, Ta, T7, T9; Chris@82: T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: T8 = BYTWJ(&(W[TWVL * 2]), T7); Chris@82: T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: Ta = BYTWJ(&(W[TWVL * 8]), T9); Chris@82: Tb = VADD(T8, Ta); Chris@82: Tk = VSUB(Ta, T8); Chris@82: } Chris@82: ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); Chris@82: Tp = VBYI(VFMA(LDK(KP433883739), Ti, VFNMS(LDK(KP781831482), Tk, VMUL(LDK(KP974927912), Tj)))); Chris@82: To = VFMA(LDK(KP623489801), Tb, VFNMS(LDK(KP222520933), Tg, VFNMS(LDK(KP900968867), T6, T1))); Chris@82: ST(&(x[WS(rs, 4)]), VSUB(To, Tp), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)])); Chris@82: { Chris@82: V Tl, Th, Tn, Tm; Chris@82: Tl = VBYI(VFNMS(LDK(KP781831482), Tj, VFNMS(LDK(KP433883739), Tk, VMUL(LDK(KP974927912), Ti)))); Chris@82: Th = VFMA(LDK(KP623489801), Tg, VFNMS(LDK(KP900968867), Tb, VFNMS(LDK(KP222520933), T6, T1))); Chris@82: ST(&(x[WS(rs, 5)]), VSUB(Th, Tl), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 2)]), VADD(Th, Tl), ms, &(x[0])); Chris@82: Tn = VBYI(VFMA(LDK(KP781831482), Ti, VFMA(LDK(KP974927912), Tk, VMUL(LDK(KP433883739), Tj)))); Chris@82: Tm = VFMA(LDK(KP623489801), T6, VFNMS(LDK(KP900968867), Tg, VFNMS(LDK(KP222520933), Tb, T1))); Chris@82: ST(&(x[WS(rs, 6)]), VSUB(Tm, Tn), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)])); Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const tw_instr twinstr[] = { Chris@82: VTW(0, 1), Chris@82: VTW(0, 2), Chris@82: VTW(0, 3), Chris@82: VTW(0, 4), Chris@82: VTW(0, 5), Chris@82: VTW(0, 6), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 7, XSIMD_STRING("t1fuv_7"), twinstr, &GENUS, {24, 18, 12, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t1fuv_7) (planner *p) { Chris@82: X(kdft_dit_register) (p, t1fuv_7, &desc); Chris@82: } Chris@82: #endif