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 16 -name X(spu_t1fv_16) */ d@0: d@0: /* d@0: * This function contains 87 FP additions, 64 FP multiplications, d@0: * (or, 53 additions, 30 multiplications, 34 fused multiply/add), d@0: * 108 stack variables, 3 constants, and 32 memory accesses d@0: */ d@0: #include "fftw-spu.h" d@0: d@0: void X(spu_t1fv_16) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { d@0: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); d@0: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); d@0: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); d@0: INT m; d@0: R *x; d@0: x = ri; d@0: for (m = mb, W = W + (mb * ((TWVL / VL) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(rs)) { d@0: V T14, T1h, Ta, TS, TJ, TT, T17, T1i, Tl, TV, T1b, T1k, Tw, TW, T1e; d@0: V T1l, T1, T8, T3, T6, T7, T2, T5, T4, T9, TA, TH, TC, TF, Tz; d@0: V TG, TB, TE, T16, T15, TD, TI, Tc, Tj, Te, Th, Tb, Ti, Td, Tg; d@0: V T19, T1a, Tf, Tk, Tn, Tu, Tp, Ts, Tm, Tt, To, Tr, T1c, T1d, Tq; d@0: V Tv, T10, T11, TY, TZ, TU, TX, T12, T13, TO, Ty, TL, TP, TK, Tx; d@0: V TN, TQ, TM, TR, T1u, T1y, T1x, T1v, T1g, T1q, T1n, T1r, T18, T1f, T1j; d@0: V T1m, T1p, T1s, T1o, T1t, T1w, T1C, T1z, T1D, T1B, T1E, T1A, T1F; d@0: T1 = LD(&(x[0]), ms, &(x[0])); d@0: T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0])); d@0: T8 = BYTWJ(&(W[TWVL * 22]), T7); d@0: T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); d@0: T3 = BYTWJ(&(W[TWVL * 14]), T2); d@0: T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); d@0: T6 = BYTWJ(&(W[TWVL * 6]), T5); d@0: T14 = VSUB(T1, T3); d@0: T4 = VADD(T1, T3); d@0: T9 = VADD(T6, T8); d@0: T1h = VSUB(T6, T8); d@0: Ta = VSUB(T4, T9); d@0: TS = VADD(T4, T9); d@0: Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); d@0: TA = BYTWJ(&(W[TWVL * 26]), Tz); d@0: TG = LD(&(x[WS(rs, 10)]), ms, &(x[0])); d@0: TH = BYTWJ(&(W[TWVL * 18]), TG); d@0: TB = LD(&(x[WS(rs, 6)]), ms, &(x[0])); d@0: TC = BYTWJ(&(W[TWVL * 10]), TB); d@0: TE = LD(&(x[WS(rs, 2)]), ms, &(x[0])); d@0: TF = BYTWJ(&(W[TWVL * 2]), TE); d@0: TD = VADD(TA, TC); d@0: T16 = VSUB(TA, TC); d@0: T15 = VSUB(TF, TH); d@0: TI = VADD(TF, TH); d@0: TJ = VSUB(TD, TI); d@0: TT = VADD(TI, TD); d@0: T17 = VADD(T15, T16); d@0: T1i = VSUB(T16, T15); d@0: Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); d@0: Tc = BYTWJ(&(W[0]), Tb); d@0: Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); d@0: Tj = BYTWJ(&(W[TWVL * 24]), Ti); d@0: Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); d@0: Te = BYTWJ(&(W[TWVL * 16]), Td); d@0: Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); d@0: Th = BYTWJ(&(W[TWVL * 8]), Tg); d@0: Tf = VADD(Tc, Te); d@0: T19 = VSUB(Tc, Te); d@0: T1a = VSUB(Th, Tj); d@0: Tk = VADD(Th, Tj); d@0: Tl = VSUB(Tf, Tk); d@0: TV = VADD(Tf, Tk); d@0: T1b = VFNMS(LDK(KP414213562), T1a, T19); d@0: T1k = VFMA(LDK(KP414213562), T19, T1a); d@0: Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); d@0: Tn = BYTWJ(&(W[TWVL * 28]), Tm); d@0: Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); d@0: Tu = BYTWJ(&(W[TWVL * 20]), Tt); d@0: To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); d@0: Tp = BYTWJ(&(W[TWVL * 12]), To); d@0: Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); d@0: Ts = BYTWJ(&(W[TWVL * 4]), Tr); d@0: Tq = VADD(Tn, Tp); d@0: T1c = VSUB(Tn, Tp); d@0: T1d = VSUB(Tu, Ts); d@0: Tv = VADD(Ts, Tu); d@0: Tw = VSUB(Tq, Tv); d@0: TW = VADD(Tq, Tv); d@0: T1e = VFNMS(LDK(KP414213562), T1d, T1c); d@0: T1l = VFMA(LDK(KP414213562), T1c, T1d); d@0: TU = VADD(TS, TT); d@0: T10 = VSUB(TS, TT); d@0: T11 = VSUB(TW, TV); d@0: TX = VADD(TV, TW); d@0: TY = VSUB(TU, TX); d@0: TZ = VADD(TU, TX); d@0: ST(&(x[WS(rs, 8)]), TY, ms, &(x[0])); d@0: ST(&(x[0]), TZ, ms, &(x[0])); d@0: T12 = VFNMSI(T11, T10); d@0: T13 = VFMAI(T11, T10); d@0: ST(&(x[WS(rs, 12)]), T12, ms, &(x[0])); d@0: ST(&(x[WS(rs, 4)]), T13, ms, &(x[0])); d@0: Tx = VADD(Tl, Tw); d@0: TK = VSUB(Tw, Tl); d@0: TO = VFMA(LDK(KP707106781), Tx, Ta); d@0: Ty = VFNMS(LDK(KP707106781), Tx, Ta); d@0: TL = VFNMS(LDK(KP707106781), TK, TJ); d@0: TP = VFMA(LDK(KP707106781), TK, TJ); d@0: TM = VFNMSI(TL, Ty); d@0: TN = VFMAI(TL, Ty); d@0: ST(&(x[WS(rs, 6)]), TM, ms, &(x[0])); d@0: TR = VFMAI(TP, TO); d@0: TQ = VFNMSI(TP, TO); d@0: ST(&(x[WS(rs, 2)]), TR, ms, &(x[0])); d@0: ST(&(x[WS(rs, 10)]), TN, ms, &(x[0])); d@0: ST(&(x[WS(rs, 14)]), TQ, ms, &(x[0])); d@0: T18 = VFMA(LDK(KP707106781), T17, T14); d@0: T1u = VFNMS(LDK(KP707106781), T17, T14); d@0: T1y = VSUB(T1e, T1b); d@0: T1f = VADD(T1b, T1e); d@0: T1g = VFNMS(LDK(KP923879532), T1f, T18); d@0: T1q = VFMA(LDK(KP923879532), T1f, T18); d@0: T1j = VFNMS(LDK(KP707106781), T1i, T1h); d@0: T1x = VFMA(LDK(KP707106781), T1i, T1h); d@0: T1v = VADD(T1k, T1l); d@0: T1m = VSUB(T1k, T1l); d@0: T1n = VFNMS(LDK(KP923879532), T1m, T1j); d@0: T1r = VFMA(LDK(KP923879532), T1m, T1j); d@0: T1o = VFNMSI(T1n, T1g); d@0: T1p = VFMAI(T1n, T1g); d@0: ST(&(x[WS(rs, 9)]), T1o, ms, &(x[WS(rs, 1)])); d@0: T1t = VFMAI(T1r, T1q); d@0: T1s = VFNMSI(T1r, T1q); d@0: ST(&(x[WS(rs, 15)]), T1t, ms, &(x[WS(rs, 1)])); d@0: ST(&(x[WS(rs, 7)]), T1p, ms, &(x[WS(rs, 1)])); d@0: ST(&(x[WS(rs, 1)]), T1s, ms, &(x[WS(rs, 1)])); d@0: T1w = VFNMS(LDK(KP923879532), T1v, T1u); d@0: T1C = VFMA(LDK(KP923879532), T1v, T1u); d@0: T1z = VFNMS(LDK(KP923879532), T1y, T1x); d@0: T1D = VFMA(LDK(KP923879532), T1y, T1x); d@0: T1A = VFNMSI(T1z, T1w); d@0: T1B = VFMAI(T1z, T1w); d@0: ST(&(x[WS(rs, 5)]), T1A, ms, &(x[WS(rs, 1)])); d@0: T1F = VFNMSI(T1D, T1C); d@0: T1E = VFMAI(T1D, T1C); d@0: ST(&(x[WS(rs, 13)]), T1F, ms, &(x[WS(rs, 1)])); d@0: ST(&(x[WS(rs, 11)]), T1B, ms, &(x[WS(rs, 1)])); d@0: ST(&(x[WS(rs, 3)]), T1E, ms, &(x[WS(rs, 1)])); d@0: } d@0: }