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 07:38:01 EST 2012 */ cannam@95: cannam@95: #include "codelet-dft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fv_7 -include t1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 36 FP additions, 36 FP multiplications, cannam@95: * (or, 15 additions, 15 multiplications, 21 fused multiply/add), cannam@95: * 42 stack variables, 6 constants, and 14 memory accesses cannam@95: */ cannam@95: #include "t1f.h" cannam@95: cannam@95: static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@95: DVK(KP801937735, +0.801937735804838252472204639014890102331838324); cannam@95: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@95: DVK(KP692021471, +0.692021471630095869627814897002069140197260599); cannam@95: DVK(KP554958132, +0.554958132087371191422194871006410481067288862); cannam@95: DVK(KP356895867, +0.356895867892209443894399510021300583399127187); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: 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)) { cannam@95: V T1, T2, T4, Te, Tc, T9, T7; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@95: Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: { cannam@95: V T3, T5, Tf, Td, Ta, T8; cannam@95: T3 = BYTWJ(&(W[0]), T2); cannam@95: T5 = BYTWJ(&(W[TWVL * 10]), T4); cannam@95: Tf = BYTWJ(&(W[TWVL * 6]), Te); cannam@95: Td = BYTWJ(&(W[TWVL * 4]), Tc); cannam@95: Ta = BYTWJ(&(W[TWVL * 8]), T9); cannam@95: T8 = BYTWJ(&(W[TWVL * 2]), T7); cannam@95: { cannam@95: V T6, Tk, Tg, Tl, Tb, Tm; cannam@95: T6 = VADD(T3, T5); cannam@95: Tk = VSUB(T5, T3); cannam@95: Tg = VADD(Td, Tf); cannam@95: Tl = VSUB(Tf, Td); cannam@95: Tb = VADD(T8, Ta); cannam@95: Tm = VSUB(Ta, T8); cannam@95: { cannam@95: V Th, Ts, Tp, Tu, Tn, Tx, Ti, Tt; cannam@95: Th = VFNMS(LDK(KP356895867), T6, Tg); cannam@95: Ts = VFMA(LDK(KP554958132), Tl, Tk); cannam@95: ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); cannam@95: Tp = VFNMS(LDK(KP356895867), Tb, T6); cannam@95: Tu = VFNMS(LDK(KP356895867), Tg, Tb); cannam@95: Tn = VFMA(LDK(KP554958132), Tm, Tl); cannam@95: Tx = VFNMS(LDK(KP554958132), Tk, Tm); cannam@95: Ti = VFNMS(LDK(KP692021471), Th, Tb); cannam@95: Tt = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Ts, Tm)); cannam@95: { cannam@95: V Tq, Tv, To, Ty, Tj, Tr, Tw; cannam@95: Tq = VFNMS(LDK(KP692021471), Tp, Tg); cannam@95: Tv = VFNMS(LDK(KP692021471), Tu, T6); cannam@95: To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk)); cannam@95: Ty = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tx, Tl)); cannam@95: Tj = VFNMS(LDK(KP900968867), Ti, T1); cannam@95: Tr = VFNMS(LDK(KP900968867), Tq, T1); cannam@95: Tw = VFNMS(LDK(KP900968867), Tv, T1); cannam@95: ST(&(x[WS(rs, 2)]), VFMAI(To, Tj), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 5)]), VFNMSI(To, Tj), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 1)]), VFMAI(Tt, Tr), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 6)]), VFNMSI(Tt, Tr), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 3)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VFNMSI(Ty, Tw), ms, &(x[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: VTW(0, 6), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 7, XSIMD_STRING("t1fv_7"), twinstr, &GENUS, {15, 15, 21, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1fv_7) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1fv_7, &desc); cannam@95: } cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fv_7 -include t1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 36 FP additions, 30 FP multiplications, cannam@95: * (or, 24 additions, 18 multiplications, 12 fused multiply/add), cannam@95: * 21 stack variables, 6 constants, and 14 memory accesses cannam@95: */ cannam@95: #include "t1f.h" cannam@95: cannam@95: static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); cannam@95: DVK(KP222520933, +0.222520933956314404288902564496794759466355569); cannam@95: DVK(KP623489801, +0.623489801858733530525004884004239810632274731); cannam@95: DVK(KP781831482, +0.781831482468029808708444526674057750232334519); cannam@95: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); cannam@95: DVK(KP433883739, +0.433883739117558120475768332848358754609990728); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: 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)) { cannam@95: V T1, Tg, Tj, T6, Ti, Tb, Tk, Tp, To; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: { cannam@95: V Td, Tf, Tc, Te; cannam@95: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: Td = BYTWJ(&(W[TWVL * 4]), Tc); cannam@95: Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: Tf = BYTWJ(&(W[TWVL * 6]), Te); cannam@95: Tg = VADD(Td, Tf); cannam@95: Tj = VSUB(Tf, Td); cannam@95: } cannam@95: { cannam@95: V T3, T5, T2, T4; cannam@95: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T3 = BYTWJ(&(W[0]), T2); cannam@95: T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@95: T5 = BYTWJ(&(W[TWVL * 10]), T4); cannam@95: T6 = VADD(T3, T5); cannam@95: Ti = VSUB(T5, T3); cannam@95: } cannam@95: { cannam@95: V T8, Ta, T7, T9; cannam@95: T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: T8 = BYTWJ(&(W[TWVL * 2]), T7); cannam@95: T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: Ta = BYTWJ(&(W[TWVL * 8]), T9); cannam@95: Tb = VADD(T8, Ta); cannam@95: Tk = VSUB(Ta, T8); cannam@95: } cannam@95: ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); cannam@95: Tp = VBYI(VFMA(LDK(KP433883739), Ti, VFNMS(LDK(KP781831482), Tk, VMUL(LDK(KP974927912), Tj)))); cannam@95: To = VFMA(LDK(KP623489801), Tb, VFNMS(LDK(KP222520933), Tg, VFNMS(LDK(KP900968867), T6, T1))); cannam@95: ST(&(x[WS(rs, 4)]), VSUB(To, Tp), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)])); cannam@95: { cannam@95: V Tl, Th, Tn, Tm; cannam@95: Tl = VBYI(VFNMS(LDK(KP781831482), Tj, VFNMS(LDK(KP433883739), Tk, VMUL(LDK(KP974927912), Ti)))); cannam@95: Th = VFMA(LDK(KP623489801), Tg, VFNMS(LDK(KP900968867), Tb, VFNMS(LDK(KP222520933), T6, T1))); cannam@95: ST(&(x[WS(rs, 5)]), VSUB(Th, Tl), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 2)]), VADD(Th, Tl), ms, &(x[0])); cannam@95: Tn = VBYI(VFMA(LDK(KP781831482), Ti, VFMA(LDK(KP974927912), Tk, VMUL(LDK(KP433883739), Tj)))); cannam@95: Tm = VFMA(LDK(KP623489801), T6, VFNMS(LDK(KP900968867), Tg, VFNMS(LDK(KP222520933), Tb, T1))); cannam@95: ST(&(x[WS(rs, 6)]), VSUB(Tm, Tn), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: VTW(0, 6), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 7, XSIMD_STRING("t1fv_7"), twinstr, &GENUS, {24, 18, 12, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1fv_7) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1fv_7, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */