/* * Copyright (c) 2003, 2007-8 Matteo Frigo * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* 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) */ /* * This function contains 87 FP additions, 64 FP multiplications, * (or, 53 additions, 30 multiplications, 34 fused multiply/add), * 108 stack variables, 3 constants, and 32 memory accesses */ #include "fftw-spu.h" void X(spu_t1fv_16) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP923879532, +0.923879532511286756128183189396788286822416626); DVK(KP707106781, +0.707106781186547524400844362104849039284835938); DVK(KP414213562, +0.414213562373095048801688724209698078569671875); INT m; R *x; x = ri; 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)) { V T14, T1h, Ta, TS, TJ, TT, T17, T1i, Tl, TV, T1b, T1k, Tw, TW, T1e; V T1l, T1, T8, T3, T6, T7, T2, T5, T4, T9, TA, TH, TC, TF, Tz; V TG, TB, TE, T16, T15, TD, TI, Tc, Tj, Te, Th, Tb, Ti, Td, Tg; V T19, T1a, Tf, Tk, Tn, Tu, Tp, Ts, Tm, Tt, To, Tr, T1c, T1d, Tq; V Tv, T10, T11, TY, TZ, TU, TX, T12, T13, TO, Ty, TL, TP, TK, Tx; V TN, TQ, TM, TR, T1u, T1y, T1x, T1v, T1g, T1q, T1n, T1r, T18, T1f, T1j; V T1m, T1p, T1s, T1o, T1t, T1w, T1C, T1z, T1D, T1B, T1E, T1A, T1F; T1 = LD(&(x[0]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0])); T8 = BYTWJ(&(W[TWVL * 22]), T7); T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 14]), T2); T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T6 = BYTWJ(&(W[TWVL * 6]), T5); T14 = VSUB(T1, T3); T4 = VADD(T1, T3); T9 = VADD(T6, T8); T1h = VSUB(T6, T8); Ta = VSUB(T4, T9); TS = VADD(T4, T9); Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); TA = BYTWJ(&(W[TWVL * 26]), Tz); TG = LD(&(x[WS(rs, 10)]), ms, &(x[0])); TH = BYTWJ(&(W[TWVL * 18]), TG); TB = LD(&(x[WS(rs, 6)]), ms, &(x[0])); TC = BYTWJ(&(W[TWVL * 10]), TB); TE = LD(&(x[WS(rs, 2)]), ms, &(x[0])); TF = BYTWJ(&(W[TWVL * 2]), TE); TD = VADD(TA, TC); T16 = VSUB(TA, TC); T15 = VSUB(TF, TH); TI = VADD(TF, TH); TJ = VSUB(TD, TI); TT = VADD(TI, TD); T17 = VADD(T15, T16); T1i = VSUB(T16, T15); Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tc = BYTWJ(&(W[0]), Tb); Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Tj = BYTWJ(&(W[TWVL * 24]), Ti); Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Te = BYTWJ(&(W[TWVL * 16]), Td); Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Th = BYTWJ(&(W[TWVL * 8]), Tg); Tf = VADD(Tc, Te); T19 = VSUB(Tc, Te); T1a = VSUB(Th, Tj); Tk = VADD(Th, Tj); Tl = VSUB(Tf, Tk); TV = VADD(Tf, Tk); T1b = VFNMS(LDK(KP414213562), T1a, T19); T1k = VFMA(LDK(KP414213562), T19, T1a); Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); Tn = BYTWJ(&(W[TWVL * 28]), Tm); Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Tu = BYTWJ(&(W[TWVL * 20]), Tt); To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tp = BYTWJ(&(W[TWVL * 12]), To); Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Ts = BYTWJ(&(W[TWVL * 4]), Tr); Tq = VADD(Tn, Tp); T1c = VSUB(Tn, Tp); T1d = VSUB(Tu, Ts); Tv = VADD(Ts, Tu); Tw = VSUB(Tq, Tv); TW = VADD(Tq, Tv); T1e = VFNMS(LDK(KP414213562), T1d, T1c); T1l = VFMA(LDK(KP414213562), T1c, T1d); TU = VADD(TS, TT); T10 = VSUB(TS, TT); T11 = VSUB(TW, TV); TX = VADD(TV, TW); TY = VSUB(TU, TX); TZ = VADD(TU, TX); ST(&(x[WS(rs, 8)]), TY, ms, &(x[0])); ST(&(x[0]), TZ, ms, &(x[0])); T12 = VFNMSI(T11, T10); T13 = VFMAI(T11, T10); ST(&(x[WS(rs, 12)]), T12, ms, &(x[0])); ST(&(x[WS(rs, 4)]), T13, ms, &(x[0])); Tx = VADD(Tl, Tw); TK = VSUB(Tw, Tl); TO = VFMA(LDK(KP707106781), Tx, Ta); Ty = VFNMS(LDK(KP707106781), Tx, Ta); TL = VFNMS(LDK(KP707106781), TK, TJ); TP = VFMA(LDK(KP707106781), TK, TJ); TM = VFNMSI(TL, Ty); TN = VFMAI(TL, Ty); ST(&(x[WS(rs, 6)]), TM, ms, &(x[0])); TR = VFMAI(TP, TO); TQ = VFNMSI(TP, TO); ST(&(x[WS(rs, 2)]), TR, ms, &(x[0])); ST(&(x[WS(rs, 10)]), TN, ms, &(x[0])); ST(&(x[WS(rs, 14)]), TQ, ms, &(x[0])); T18 = VFMA(LDK(KP707106781), T17, T14); T1u = VFNMS(LDK(KP707106781), T17, T14); T1y = VSUB(T1e, T1b); T1f = VADD(T1b, T1e); T1g = VFNMS(LDK(KP923879532), T1f, T18); T1q = VFMA(LDK(KP923879532), T1f, T18); T1j = VFNMS(LDK(KP707106781), T1i, T1h); T1x = VFMA(LDK(KP707106781), T1i, T1h); T1v = VADD(T1k, T1l); T1m = VSUB(T1k, T1l); T1n = VFNMS(LDK(KP923879532), T1m, T1j); T1r = VFMA(LDK(KP923879532), T1m, T1j); T1o = VFNMSI(T1n, T1g); T1p = VFMAI(T1n, T1g); ST(&(x[WS(rs, 9)]), T1o, ms, &(x[WS(rs, 1)])); T1t = VFMAI(T1r, T1q); T1s = VFNMSI(T1r, T1q); ST(&(x[WS(rs, 15)]), T1t, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), T1p, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), T1s, ms, &(x[WS(rs, 1)])); T1w = VFNMS(LDK(KP923879532), T1v, T1u); T1C = VFMA(LDK(KP923879532), T1v, T1u); T1z = VFNMS(LDK(KP923879532), T1y, T1x); T1D = VFMA(LDK(KP923879532), T1y, T1x); T1A = VFNMSI(T1z, T1w); T1B = VFMAI(T1z, T1w); ST(&(x[WS(rs, 5)]), T1A, ms, &(x[WS(rs, 1)])); T1F = VFNMSI(T1D, T1C); T1E = VFMAI(T1D, T1C); ST(&(x[WS(rs, 13)]), T1F, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 11)]), T1B, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), T1E, ms, &(x[WS(rs, 1)])); } }