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 07:37:30 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 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include n2b.h -store-multiple 2 */ 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: * 65 stack variables, 6 constants, and 35 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_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 = ii; Chris@10: xo = io; 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, Tu, TU, 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: Tu = VSUB(T6, T9); Chris@10: TU = VSUB(TI, TJ); 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 T19, T1a, T13, TG, TY, T18, TB, Tw, TT, Tz, T11, T16, TE, Tr, TV; Chris@10: V TQ; Chris@10: TV = VSUB(TP, TO); Chris@10: TQ = VADD(TO, TP); Chris@10: { Chris@10: V Ts, To, TW, TN; Chris@10: Ts = VSUB(Tk, Tn); Chris@10: To = VADD(Tk, Tn); Chris@10: TW = VSUB(TM, TL); Chris@10: TN = VADD(TL, TM); Chris@10: { Chris@10: V Tt, Th, TR, T12; Chris@10: Tt = VSUB(Td, Tg); Chris@10: Th = VADD(Td, Tg); Chris@10: TR = VFNMS(LDK(KP356895867), TK, TQ); Chris@10: T12 = VFNMS(LDK(KP554958132), TV, TU); Chris@10: { Chris@10: V Tx, TF, TZ, T14; Chris@10: Tx = VFNMS(LDK(KP356895867), Ta, To); Chris@10: TF = VFMA(LDK(KP554958132), Ts, Tu); Chris@10: T19 = VADD(TH, VADD(TK, VADD(TN, TQ))); Chris@10: STM2(&(xo[0]), T19, ovs, &(xo[0])); Chris@10: TZ = VFNMS(LDK(KP356895867), TN, TK); Chris@10: T14 = VFNMS(LDK(KP356895867), TQ, TN); Chris@10: { Chris@10: V TX, T17, TC, Tp; Chris@10: TX = VFMA(LDK(KP554958132), TW, TV); Chris@10: T17 = VFMA(LDK(KP554958132), TU, TW); Chris@10: T1a = VADD(T3, VADD(Ta, VADD(Th, To))); Chris@10: STM2(&(xo[14]), T1a, ovs, &(xo[2])); Chris@10: TC = VFNMS(LDK(KP356895867), Th, Ta); Chris@10: Tp = VFNMS(LDK(KP356895867), To, Th); Chris@10: { Chris@10: V TA, Tv, TS, Ty; Chris@10: TA = VFMA(LDK(KP554958132), Tt, Ts); Chris@10: Tv = VFNMS(LDK(KP554958132), Tu, Tt); Chris@10: TS = VFNMS(LDK(KP692021471), TR, TN); Chris@10: T13 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T12, TW)); Chris@10: Ty = VFNMS(LDK(KP692021471), Tx, Th); Chris@10: TG = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TF, Tt)); Chris@10: { Chris@10: V T10, T15, TD, Tq; Chris@10: T10 = VFNMS(LDK(KP692021471), TZ, TQ); Chris@10: T15 = VFNMS(LDK(KP692021471), T14, TK); Chris@10: TY = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TX, TU)); Chris@10: T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV)); Chris@10: TD = VFNMS(LDK(KP692021471), TC, To); Chris@10: Tq = VFNMS(LDK(KP692021471), Tp, Ta); Chris@10: TB = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TA, Tu)); Chris@10: Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts)); Chris@10: TT = VFNMS(LDK(KP900968867), TS, TH); Chris@10: Tz = VFNMS(LDK(KP900968867), Ty, T3); Chris@10: T11 = VFNMS(LDK(KP900968867), T10, TH); Chris@10: T16 = VFNMS(LDK(KP900968867), T15, TH); Chris@10: TE = VFNMS(LDK(KP900968867), TD, T3); Chris@10: Tr = VFNMS(LDK(KP900968867), Tq, T3); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1b, T1c, T1d, T1e; Chris@10: T1b = VFMAI(TY, TT); Chris@10: STM2(&(xo[4]), T1b, ovs, &(xo[0])); Chris@10: T1c = VFNMSI(TY, TT); Chris@10: STM2(&(xo[24]), T1c, ovs, &(xo[0])); Chris@10: T1d = VFMAI(TB, Tz); Chris@10: STM2(&(xo[18]), T1d, ovs, &(xo[2])); Chris@10: T1e = VFNMSI(TB, Tz); Chris@10: STM2(&(xo[10]), T1e, ovs, &(xo[2])); Chris@10: { Chris@10: V T1f, T1g, T1h, T1i; Chris@10: T1f = VFMAI(T13, T11); Chris@10: STM2(&(xo[12]), T1f, ovs, &(xo[0])); Chris@10: STN2(&(xo[12]), T1f, T1a, ovs); Chris@10: T1g = VFNMSI(T13, T11); Chris@10: STM2(&(xo[16]), T1g, ovs, &(xo[0])); Chris@10: STN2(&(xo[16]), T1g, T1d, ovs); Chris@10: T1h = VFMAI(T18, T16); Chris@10: STM2(&(xo[8]), T1h, ovs, &(xo[0])); Chris@10: STN2(&(xo[8]), T1h, T1e, ovs); Chris@10: T1i = VFNMSI(T18, T16); Chris@10: STM2(&(xo[20]), T1i, ovs, &(xo[0])); Chris@10: { Chris@10: V T1j, T1k, T1l, T1m; Chris@10: T1j = VFNMSI(TG, TE); Chris@10: STM2(&(xo[26]), T1j, ovs, &(xo[2])); Chris@10: STN2(&(xo[24]), T1c, T1j, ovs); Chris@10: T1k = VFMAI(TG, TE); Chris@10: STM2(&(xo[2]), T1k, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), T19, T1k, ovs); Chris@10: T1l = VFNMSI(Tw, Tr); Chris@10: STM2(&(xo[22]), T1l, ovs, &(xo[2])); Chris@10: STN2(&(xo[20]), T1i, T1l, ovs); Chris@10: T1m = VFMAI(Tw, Tr); Chris@10: STM2(&(xo[6]), T1m, ovs, &(xo[2])); Chris@10: STN2(&(xo[4]), T1b, T1m, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {32, 6, 42, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_14) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_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 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include n2b.h -store-multiple 2 */ 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: * 41 stack variables, 6 constants, and 35 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_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(KP222520933, +0.222520933956314404288902564496794759466355569); Chris@10: DVK(KP623489801, +0.623489801858733530525004884004239810632274731); Chris@10: DVK(KP781831482, +0.781831482468029808708444526674057750232334519); Chris@10: DVK(KP974927912, +0.974927912181823607018131682993931217232785801); Chris@10: DVK(KP433883739, +0.433883739117558120475768332848358754609990728); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; 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 Tp, Ty, Tl, TL, Tq, TE, T7, TJ, Ts, TB, Te, TK, Tr, TH, Tn; Chris@10: V To; Chris@10: Tn = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: To = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tp = VSUB(Tn, To); Chris@10: Ty = VADD(Tn, To); Chris@10: { Chris@10: V Th, TC, Tk, TD; Chris@10: { Chris@10: V Tf, Tg, Ti, Tj; Chris@10: Tf = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: Tg = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: Th = VSUB(Tf, Tg); Chris@10: TC = VADD(Tf, Tg); Chris@10: Ti = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: Tj = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tk = VSUB(Ti, Tj); Chris@10: TD = VADD(Ti, Tj); Chris@10: } Chris@10: Tl = VSUB(Th, Tk); Chris@10: TL = VSUB(TD, TC); Chris@10: Tq = VADD(Th, Tk); Chris@10: TE = VADD(TC, TD); Chris@10: } Chris@10: { Chris@10: V T3, Tz, T6, TA; Chris@10: { Chris@10: V T1, T2, T4, T5; Chris@10: T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = VSUB(T1, T2); Chris@10: Tz = VADD(T1, T2); Chris@10: T4 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: T6 = VSUB(T4, T5); Chris@10: TA = VADD(T4, T5); Chris@10: } Chris@10: T7 = VSUB(T3, T6); Chris@10: TJ = VSUB(Tz, TA); Chris@10: Ts = VADD(T3, T6); Chris@10: TB = VADD(Tz, TA); Chris@10: } Chris@10: { Chris@10: V Ta, TF, Td, TG; Chris@10: { Chris@10: V T8, T9, Tb, Tc; Chris@10: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: T9 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Ta = VSUB(T8, T9); Chris@10: TF = VADD(T8, T9); Chris@10: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: Td = VSUB(Tb, Tc); Chris@10: TG = VADD(Tb, Tc); Chris@10: } Chris@10: Te = VSUB(Ta, Td); Chris@10: TK = VSUB(TG, TF); Chris@10: Tr = VADD(Ta, Td); Chris@10: TH = VADD(TF, TG); Chris@10: } Chris@10: { Chris@10: V TR, TS, TU, TV; Chris@10: TR = VADD(Tp, VADD(Ts, VADD(Tq, Tr))); Chris@10: STM2(&(xo[14]), TR, ovs, &(xo[2])); Chris@10: TS = VADD(Ty, VADD(TB, VADD(TE, TH))); Chris@10: STM2(&(xo[0]), TS, ovs, &(xo[0])); Chris@10: { Chris@10: V TT, Tm, Tt, TQ, TP, TW; Chris@10: Tm = VBYI(VFMA(LDK(KP433883739), T7, VFNMS(LDK(KP781831482), Tl, VMUL(LDK(KP974927912), Te)))); Chris@10: Tt = VFMA(LDK(KP623489801), Tq, VFNMS(LDK(KP222520933), Tr, VFNMS(LDK(KP900968867), Ts, Tp))); Chris@10: TT = VADD(Tm, Tt); Chris@10: STM2(&(xo[6]), TT, ovs, &(xo[2])); Chris@10: TU = VSUB(Tt, Tm); Chris@10: STM2(&(xo[22]), TU, ovs, &(xo[2])); Chris@10: TQ = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK)))); Chris@10: TP = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TB, Ty))); Chris@10: TV = VSUB(TP, TQ); Chris@10: STM2(&(xo[24]), TV, ovs, &(xo[0])); Chris@10: TW = VADD(TP, TQ); Chris@10: STM2(&(xo[4]), TW, ovs, &(xo[0])); Chris@10: STN2(&(xo[4]), TW, TT, ovs); Chris@10: } Chris@10: { Chris@10: V T10, TM, TI, TZ; Chris@10: { Chris@10: V Tu, Tv, TX, TY; Chris@10: Tu = VBYI(VFMA(LDK(KP781831482), T7, VFMA(LDK(KP974927912), Tl, VMUL(LDK(KP433883739), Te)))); Chris@10: Tv = VFMA(LDK(KP623489801), Ts, VFNMS(LDK(KP900968867), Tr, VFNMS(LDK(KP222520933), Tq, Tp))); Chris@10: TX = VADD(Tu, Tv); Chris@10: STM2(&(xo[2]), TX, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), TS, TX, ovs); Chris@10: TY = VSUB(Tv, Tu); Chris@10: STM2(&(xo[26]), TY, ovs, &(xo[2])); Chris@10: STN2(&(xo[24]), TV, TY, ovs); Chris@10: } 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), TH, VFNMS(LDK(KP222520933), TE, Ty))); Chris@10: TZ = VSUB(TI, TM); Chris@10: STM2(&(xo[12]), TZ, ovs, &(xo[0])); Chris@10: STN2(&(xo[12]), TZ, TR, ovs); Chris@10: T10 = VADD(TI, TM); Chris@10: STM2(&(xo[16]), T10, ovs, &(xo[0])); Chris@10: { Chris@10: V T11, TO, TN, T12; Chris@10: TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL)))); Chris@10: TN = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP222520933), TH, VFNMS(LDK(KP900968867), TB, Ty))); Chris@10: T11 = VSUB(TN, TO); Chris@10: STM2(&(xo[8]), T11, ovs, &(xo[0])); Chris@10: T12 = VADD(TN, TO); Chris@10: STM2(&(xo[20]), T12, ovs, &(xo[0])); Chris@10: STN2(&(xo[20]), T12, TU, ovs); Chris@10: { Chris@10: V Tx, Tw, T13, T14; Chris@10: Tx = VBYI(VFNMS(LDK(KP781831482), Te, VFNMS(LDK(KP433883739), Tl, VMUL(LDK(KP974927912), T7)))); Chris@10: Tw = VFMA(LDK(KP623489801), Tr, VFNMS(LDK(KP900968867), Tq, VFNMS(LDK(KP222520933), Ts, Tp))); Chris@10: T13 = VSUB(Tw, Tx); Chris@10: STM2(&(xo[10]), T13, ovs, &(xo[2])); Chris@10: STN2(&(xo[8]), T11, T13, ovs); Chris@10: T14 = VADD(Tx, Tw); Chris@10: STM2(&(xo[18]), T14, ovs, &(xo[2])); Chris@10: STN2(&(xo[16]), T10, T14, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {50, 12, 24, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_14) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_14, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */