d@0: /* d@0: * Copyright (c) 2003, 2007-8 Matteo Frigo d@0: * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology d@0: * d@0: * This program is free software; you can redistribute it and/or modify d@0: * it under the terms of the GNU General Public License as published by d@0: * the Free Software Foundation; either version 2 of the License, or d@0: * (at your option) any later version. d@0: * d@0: * This program is distributed in the hope that it will be useful, d@0: * but WITHOUT ANY WARRANTY; without even the implied warranty of d@0: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the d@0: * GNU General Public License for more details. d@0: * d@0: * You should have received a copy of the GNU General Public License d@0: * along with this program; if not, write to the Free Software d@0: * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA d@0: * d@0: */ d@0: /* Generated by: ../../genfft/gen_twiddle_c -standalone -fma -reorder-insns -simd -compact -variables 100000 -include fftw-spu.h -trivial-stores -n 9 -name X(spu_t1fv_9) */ d@0: d@0: /* d@0: * This function contains 54 FP additions, 54 FP multiplications, d@0: * (or, 20 additions, 20 multiplications, 34 fused multiply/add), d@0: * 84 stack variables, 19 constants, and 18 memory accesses d@0: */ d@0: #include "fftw-spu.h" d@0: d@0: void X(spu_t1fv_9) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { d@0: DVK(KP666666666, +0.666666666666666666666666666666666666666666667); d@0: DVK(KP852868531, +0.852868531952443209628250963940074071936020296); d@0: DVK(KP673648177, +0.673648177666930348851716626769314796000375677); d@0: DVK(KP898197570, +0.898197570222573798468955502359086394667167570); d@0: DVK(KP879385241, +0.879385241571816768108218554649462939872416269); d@0: DVK(KP984807753, +0.984807753012208059366743024589523013670643252); d@0: DVK(KP826351822, +0.826351822333069651148283373230685203999624323); d@0: DVK(KP420276625, +0.420276625461206169731530603237061658838781920); d@0: DVK(KP939692620, +0.939692620785908384054109277324731469936208134); d@0: DVK(KP907603734, +0.907603734547952313649323976213898122064543220); d@0: DVK(KP347296355, +0.347296355333860697703433253538629592000751354); d@0: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); d@0: DVK(KP152703644, +0.152703644666139302296566746461370407999248646); d@0: DVK(KP203604859, +0.203604859554852403062088995281827210665664861); d@0: DVK(KP726681596, +0.726681596905677465811651808188092531873167623); d@0: DVK(KP968908795, +0.968908795874236621082202410917456709164223497); d@0: DVK(KP439692620, +0.439692620785908384054109277324731469936208134); d@0: DVK(KP586256827, +0.586256827714544512072145703099641959914944179); d@0: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); d@0: INT m; d@0: R *x; d@0: x = ri; d@0: for (m = mb, W = W + (mb * ((TWVL / VL) * 16)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 16), MAKE_VOLATILE_STRIDE(rs)) { d@0: V TG, T7, Tu, Tw, Tx, Tf, Ty, Tv, Tn, Tz, TD, TP, TL, TK, TO; d@0: V T1, T5, T3, T4, T2, T6, T9, Th, Te, Tm, T8, Tg, Tb, Td, Ta; d@0: V Tc, Tj, Tl, Ti, Tk, Tr, Tp, Tq, To, Ts, Tt, TC, TH, TB, TA; d@0: V TF, TE, TJ, TI, TN, TW, TR, TZ, TX, TQ, TM, TV, TU, TY, TT; d@0: V T10, TS, T11; d@0: T1 = LD(&(x[0]), ms, &(x[0])); d@0: T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); d@0: T5 = BYTWJ(&(W[TWVL * 10]), T4); d@0: T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); d@0: T3 = BYTWJ(&(W[TWVL * 4]), T2); d@0: TG = VSUB(T5, T3); d@0: T6 = VADD(T3, T5); d@0: T7 = VADD(T1, T6); d@0: Tu = VFNMS(LDK(KP500000000), T6, T1); d@0: T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); d@0: T9 = BYTWJ(&(W[0]), T8); d@0: Tg = LD(&(x[WS(rs, 2)]), ms, &(x[0])); d@0: Th = BYTWJ(&(W[TWVL * 2]), Tg); d@0: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); d@0: Tb = BYTWJ(&(W[TWVL * 6]), Ta); d@0: Tc = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); d@0: Td = BYTWJ(&(W[TWVL * 12]), Tc); d@0: Te = VADD(Tb, Td); d@0: Tw = VSUB(Tb, Td); d@0: Ti = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); d@0: Tj = BYTWJ(&(W[TWVL * 8]), Ti); d@0: Tk = LD(&(x[WS(rs, 8)]), ms, &(x[0])); d@0: Tl = BYTWJ(&(W[TWVL * 14]), Tk); d@0: Tx = VSUB(Tl, Tj); d@0: Tm = VADD(Tj, Tl); d@0: Tf = VADD(T9, Te); d@0: Ty = VFNMS(LDK(KP500000000), Te, T9); d@0: Tv = VFNMS(LDK(KP500000000), Tm, Th); d@0: Tn = VADD(Th, Tm); d@0: Tz = VFNMS(LDK(KP586256827), Ty, Tx); d@0: TD = VFNMS(LDK(KP439692620), Tw, Tv); d@0: TP = VFMA(LDK(KP968908795), Ty, Tw); d@0: TL = VFNMS(LDK(KP726681596), Tw, Ty); d@0: TK = VFMA(LDK(KP203604859), Tv, Tx); d@0: TO = VFNMS(LDK(KP152703644), Tx, Tv); d@0: Tr = VMUL(LDK(KP866025403), VSUB(Tn, Tf)); d@0: To = VADD(Tf, Tn); d@0: Tp = VADD(T7, To); d@0: Tq = VFNMS(LDK(KP500000000), To, T7); d@0: ST(&(x[0]), Tp, ms, &(x[0])); d@0: Ts = VFNMSI(Tr, Tq); d@0: Tt = VFMAI(Tr, Tq); d@0: ST(&(x[WS(rs, 3)]), Tt, ms, &(x[WS(rs, 1)])); d@0: ST(&(x[WS(rs, 6)]), Ts, ms, &(x[0])); d@0: TA = VFNMS(LDK(KP347296355), Tz, Tw); d@0: TB = VFNMS(LDK(KP907603734), TA, Tv); d@0: TC = VFNMS(LDK(KP939692620), TB, Tu); d@0: TE = VFNMS(LDK(KP420276625), TD, Tx); d@0: TF = VFNMS(LDK(KP826351822), TE, Ty); d@0: TH = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), TG, TF)); d@0: TJ = VFMAI(TH, TC); d@0: TI = VFNMSI(TH, TC); d@0: ST(&(x[WS(rs, 2)]), TI, ms, &(x[0])); d@0: ST(&(x[WS(rs, 7)]), TJ, ms, &(x[WS(rs, 1)])); d@0: TX = VFNMS(LDK(KP898197570), TL, TK); d@0: TM = VFMA(LDK(KP898197570), TL, TK); d@0: TU = VFNMS(LDK(KP673648177), TP, TO); d@0: TQ = VFMA(LDK(KP673648177), TP, TO); d@0: TV = VFNMS(LDK(KP500000000), TM, TU); d@0: TN = VFMA(LDK(KP852868531), TM, Tu); d@0: TW = VFMA(LDK(KP852868531), TV, Tu); d@0: TY = VFMA(LDK(KP666666666), TQ, TX); d@0: TR = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), TG, TQ)); d@0: TZ = VMUL(LDK(KP866025403), VFMA(LDK(KP852868531), TY, TG)); d@0: TT = VFMAI(TR, TN); d@0: TS = VFNMSI(TR, TN); d@0: ST(&(x[WS(rs, 1)]), TS, ms, &(x[WS(rs, 1)])); d@0: T10 = VFNMSI(TZ, TW); d@0: T11 = VFMAI(TZ, TW); d@0: ST(&(x[WS(rs, 4)]), T11, ms, &(x[0])); d@0: ST(&(x[WS(rs, 8)]), TT, ms, &(x[0])); d@0: ST(&(x[WS(rs, 5)]), T10, ms, &(x[WS(rs, 1)])); d@0: } d@0: }