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 07:36:52 EST 2012 */ Chris@10: Chris@10: #include "codelet-dft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 14 -name n1fv_14 -include n1f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 74 FP additions, 48 FP multiplications, Chris@10: * (or, 32 additions, 6 multiplications, 42 fused multiply/add), Chris@10: * 63 stack variables, 6 constants, and 28 memory accesses Chris@10: */ Chris@10: #include "n1f.h" Chris@10: Chris@10: static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@10: DVK(KP801937735, +0.801937735804838252472204639014890102331838324); Chris@10: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@10: DVK(KP692021471, +0.692021471630095869627814897002069140197260599); Chris@10: DVK(KP554958132, +0.554958132087371191422194871006410481067288862); Chris@10: DVK(KP356895867, +0.356895867892209443894399510021300583399127187); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ri; Chris@10: xo = ro; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) { Chris@10: V TH, T3, TP, Tn, Ta, Ts, TW, TK, TO, Tk, TM, Tg, TL, Td, T1; Chris@10: V T2; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: { Chris@10: V Ti, TI, T6, TJ, T9, Tj, Te, Tf, Tb, Tc; Chris@10: { Chris@10: V T4, T5, T7, T8, Tl, Tm; Chris@10: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: TH = VADD(T1, T2); Chris@10: T3 = VSUB(T1, T2); Chris@10: TI = VADD(T4, T5); Chris@10: T6 = VSUB(T4, T5); Chris@10: TJ = VADD(T7, T8); Chris@10: T9 = VSUB(T7, T8); Chris@10: TP = VADD(Tl, Tm); Chris@10: Tn = VSUB(Tl, Tm); Chris@10: Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: } Chris@10: Ta = VADD(T6, T9); Chris@10: Ts = VSUB(T9, T6); Chris@10: TW = VSUB(TJ, TI); Chris@10: TK = VADD(TI, TJ); Chris@10: TO = VADD(Ti, Tj); Chris@10: Tk = VSUB(Ti, Tj); Chris@10: TM = VADD(Te, Tf); Chris@10: Tg = VSUB(Te, Tf); Chris@10: TL = VADD(Tb, Tc); Chris@10: Td = VSUB(Tb, Tc); Chris@10: } Chris@10: { Chris@10: V T18, TB, T13, TY, TG, Tw, T11, Tr, T16, TT, Tz, TE, TU, TQ; Chris@10: TU = VSUB(TO, TP); Chris@10: TQ = VADD(TO, TP); Chris@10: { Chris@10: V Tt, To, TV, TN; Chris@10: Tt = VSUB(Tn, Tk); Chris@10: To = VADD(Tk, Tn); Chris@10: TV = VSUB(TL, TM); Chris@10: TN = VADD(TL, TM); Chris@10: { Chris@10: V Tu, Th, TZ, T17; Chris@10: Tu = VSUB(Tg, Td); Chris@10: Th = VADD(Td, Tg); Chris@10: TZ = VFNMS(LDK(KP356895867), TK, TQ); Chris@10: T17 = VFNMS(LDK(KP554958132), TU, TW); Chris@10: { Chris@10: V Tp, TA, T14, TR; Chris@10: Tp = VFNMS(LDK(KP356895867), Ta, To); Chris@10: TA = VFMA(LDK(KP554958132), Tt, Ts); Chris@10: ST(&(xo[0]), VADD(TH, VADD(TK, VADD(TN, TQ))), ovs, &(xo[0])); Chris@10: T14 = VFNMS(LDK(KP356895867), TN, TK); Chris@10: TR = VFNMS(LDK(KP356895867), TQ, TN); Chris@10: { Chris@10: V T12, TX, Tx, TC; Chris@10: T12 = VFMA(LDK(KP554958132), TV, TU); Chris@10: TX = VFMA(LDK(KP554958132), TW, TV); Chris@10: ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)])); Chris@10: Tx = VFNMS(LDK(KP356895867), Th, Ta); Chris@10: TC = VFNMS(LDK(KP356895867), To, Th); Chris@10: { Chris@10: V TF, Tv, T10, Tq; Chris@10: TF = VFNMS(LDK(KP554958132), Ts, Tu); Chris@10: Tv = VFMA(LDK(KP554958132), Tu, Tt); Chris@10: T10 = VFNMS(LDK(KP692021471), TZ, TN); Chris@10: T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV)); Chris@10: Tq = VFNMS(LDK(KP692021471), Tp, Th); Chris@10: TB = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TA, Tu)); Chris@10: { Chris@10: V T15, TS, Ty, TD; Chris@10: T15 = VFNMS(LDK(KP692021471), T14, TQ); Chris@10: TS = VFNMS(LDK(KP692021471), TR, TK); Chris@10: T13 = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), T12, TW)); Chris@10: TY = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TX, TU)); Chris@10: Ty = VFNMS(LDK(KP692021471), Tx, To); Chris@10: TD = VFNMS(LDK(KP692021471), TC, Ta); Chris@10: TG = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TF, Tt)); Chris@10: Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts)); Chris@10: T11 = VFNMS(LDK(KP900968867), T10, TH); Chris@10: Tr = VFNMS(LDK(KP900968867), Tq, T3); Chris@10: T16 = VFNMS(LDK(KP900968867), T15, TH); Chris@10: TT = VFNMS(LDK(KP900968867), TS, TH); Chris@10: Tz = VFNMS(LDK(KP900968867), Ty, T3); Chris@10: TE = VFNMS(LDK(KP900968867), TD, T3); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: ST(&(xo[WS(os, 12)]), VFNMSI(T13, T11), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 2)]), VFMAI(T13, T11), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 9)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 5)]), VFNMSI(Tw, Tr), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 8)]), VFNMSI(T18, T16), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 6)]), VFMAI(T18, T16), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 10)]), VFNMSI(TY, TT), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 4)]), VFMAI(TY, TT), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 1)]), VFMAI(TB, Tz), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 13)]), VFNMSI(TB, Tz), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 3)]), VFMAI(TG, TE), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 11)]), VFNMSI(TG, TE), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), {32, 6, 42, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1fv_14) (planner *p) { Chris@10: X(kdft_register) (p, n1fv_14, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 14 -name n1fv_14 -include n1f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 74 FP additions, 36 FP multiplications, Chris@10: * (or, 50 additions, 12 multiplications, 24 fused multiply/add), Chris@10: * 33 stack variables, 6 constants, and 28 memory accesses Chris@10: */ Chris@10: #include "n1f.h" Chris@10: Chris@10: static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP222520933, +0.222520933956314404288902564496794759466355569); Chris@10: DVK(KP900968867, +0.900968867902419126236102319507445051165919162); Chris@10: DVK(KP623489801, +0.623489801858733530525004884004239810632274731); Chris@10: DVK(KP433883739, +0.433883739117558120475768332848358754609990728); Chris@10: DVK(KP781831482, +0.781831482468029808708444526674057750232334519); Chris@10: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ri; Chris@10: xo = ro; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) { Chris@10: V T3, Ty, To, TK, Tr, TE, Ta, TJ, Tq, TB, Th, TL, Ts, TH, T1; Chris@10: V T2; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = VSUB(T1, T2); Chris@10: Ty = VADD(T1, T2); Chris@10: { Chris@10: V Tk, TC, Tn, TD; Chris@10: { Chris@10: V Ti, Tj, Tl, Tm; Chris@10: Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tk = VSUB(Ti, Tj); Chris@10: TC = VADD(Ti, Tj); Chris@10: Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tn = VSUB(Tl, Tm); Chris@10: TD = VADD(Tl, Tm); Chris@10: } Chris@10: To = VADD(Tk, Tn); Chris@10: TK = VSUB(TC, TD); Chris@10: Tr = VSUB(Tn, Tk); Chris@10: TE = VADD(TC, TD); Chris@10: } Chris@10: { Chris@10: V T6, Tz, T9, TA; Chris@10: { Chris@10: V T4, T5, T7, T8; Chris@10: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T6 = VSUB(T4, T5); Chris@10: Tz = VADD(T4, T5); Chris@10: T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: T9 = VSUB(T7, T8); Chris@10: TA = VADD(T7, T8); Chris@10: } Chris@10: Ta = VADD(T6, T9); Chris@10: TJ = VSUB(TA, Tz); Chris@10: Tq = VSUB(T9, T6); Chris@10: TB = VADD(Tz, TA); Chris@10: } Chris@10: { Chris@10: V Td, TF, Tg, TG; Chris@10: { Chris@10: V Tb, Tc, Te, Tf; Chris@10: Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: Td = VSUB(Tb, Tc); Chris@10: TF = VADD(Tb, Tc); Chris@10: Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tg = VSUB(Te, Tf); Chris@10: TG = VADD(Te, Tf); Chris@10: } Chris@10: Th = VADD(Td, Tg); Chris@10: TL = VSUB(TF, TG); Chris@10: Ts = VSUB(Tg, Td); Chris@10: TH = VADD(TF, TG); Chris@10: } Chris@10: ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[0]), VADD(Ty, VADD(TB, VADD(TH, TE))), ovs, &(xo[0])); Chris@10: { Chris@10: V Tt, Tp, TP, TQ; Chris@10: Tt = VBYI(VFNMS(LDK(KP781831482), Tr, VFNMS(LDK(KP433883739), Ts, VMUL(LDK(KP974927912), Tq)))); Chris@10: Tp = VFMA(LDK(KP623489801), To, VFNMS(LDK(KP900968867), Th, VFNMS(LDK(KP222520933), Ta, T3))); Chris@10: ST(&(xo[WS(os, 5)]), VSUB(Tp, Tt), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 9)]), VADD(Tp, Tt), ovs, &(xo[WS(os, 1)])); Chris@10: TP = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK)))); Chris@10: TQ = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TB, Ty))); Chris@10: ST(&(xo[WS(os, 2)]), VADD(TP, TQ), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 12)]), VSUB(TQ, TP), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V Tv, Tu, TM, TI; Chris@10: Tv = VBYI(VFMA(LDK(KP781831482), Tq, VFMA(LDK(KP974927912), Ts, VMUL(LDK(KP433883739), Tr)))); Chris@10: Tu = VFMA(LDK(KP623489801), Ta, VFNMS(LDK(KP900968867), To, VFNMS(LDK(KP222520933), Th, T3))); Chris@10: ST(&(xo[WS(os, 13)]), VSUB(Tu, Tv), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 1)]), VADD(Tu, Tv), ovs, &(xo[WS(os, 1)])); Chris@10: TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ)))); Chris@10: TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TH, Ty))); Chris@10: ST(&(xo[WS(os, 6)]), VSUB(TI, TM), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 8)]), VADD(TM, TI), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V TO, TN, Tx, Tw; Chris@10: TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL)))); Chris@10: TN = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP222520933), TE, VFNMS(LDK(KP900968867), TB, Ty))); Chris@10: ST(&(xo[WS(os, 4)]), VSUB(TN, TO), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 10)]), VADD(TO, TN), ovs, &(xo[0])); Chris@10: Tx = VBYI(VFMA(LDK(KP433883739), Tq, VFNMS(LDK(KP781831482), Ts, VMUL(LDK(KP974927912), Tr)))); Chris@10: Tw = VFMA(LDK(KP623489801), Th, VFNMS(LDK(KP222520933), To, VFNMS(LDK(KP900968867), Ta, T3))); Chris@10: ST(&(xo[WS(os, 11)]), VSUB(Tw, Tx), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 3)]), VADD(Tw, Tx), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), {50, 12, 24, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1fv_14) (planner *p) { Chris@10: X(kdft_register) (p, n1fv_14, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */