/* * 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 15 -name X(spu_t1fv_15) */ /* * This function contains 92 FP additions, 77 FP multiplications, * (or, 50 additions, 35 multiplications, 42 fused multiply/add), * 117 stack variables, 8 constants, and 30 memory accesses */ #include "fftw-spu.h" void X(spu_t1fv_15) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP823639103, +0.823639103546331925877420039278190003029660514); DVK(KP910592997, +0.910592997310029334643087372129977886038870291); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP618033988, +0.618033988749894848204586834365638117720309180); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 28)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 28), MAKE_VOLATILE_STRIDE(rs)) { V T1g, T7, TU, T17, T1a, To, TL, TK, TF, T1j, T1l, T1d, T1e, T11, T13; V T1, T5, T3, T4, T2, T6, T9, Tq, Ty, Th, Te, T15, Tv, T18, TD; V T19, Tm, T16, T8, Tp, Tx, Tg, Tb, Td, Ta, Tc, Ts, Tu, Tr, Tt; V TA, TC, Tz, TB, Tj, Tl, Ti, Tk, T1h, T1i, TV, TW, Tf, Tn, TY; V TZ, Tw, TE, TX, T10, T12, T1k, T1J, T1I, T1G, T1H, TQ, TM, TT, TJ; V TP, TI, TH, TG, TR, TS, TO, TN, T1r, T1n, T1D, T1z, T1q, T1c, T1C; V T1w, T1f, T1x, T1y, T1m, T1v, T1b, T1u, T14, T1p, T1F, T1o, T1E, T1t, T1B; V T1s, T1A; T1 = LD(&(x[0]), ms, &(x[0])); T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); T5 = BYTWJ(&(W[TWVL * 18]), T4); T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[TWVL * 8]), T2); T1g = VSUB(T5, T3); T6 = VADD(T3, T5); T7 = VADD(T1, T6); TU = VFNMS(LDK(KP500000000), T6, T1); T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[TWVL * 4]), T8); Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Tq = BYTWJ(&(W[TWVL * 10]), Tp); Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Ty = BYTWJ(&(W[TWVL * 16]), Tx); Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Th = BYTWJ(&(W[TWVL * 22]), Tg); Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Tb = BYTWJ(&(W[TWVL * 14]), Ta); Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[TWVL * 24]), Tc); Te = VADD(Tb, Td); T15 = VSUB(Td, Tb); Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Ts = BYTWJ(&(W[TWVL * 20]), Tr); Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tu = BYTWJ(&(W[0]), Tt); Tv = VADD(Ts, Tu); T18 = VSUB(Tu, Ts); Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); TA = BYTWJ(&(W[TWVL * 26]), Tz); TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); TC = BYTWJ(&(W[TWVL * 6]), TB); TD = VADD(TA, TC); T19 = VSUB(TC, TA); Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tj = BYTWJ(&(W[TWVL * 2]), Ti); Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tl = BYTWJ(&(W[TWVL * 12]), Tk); Tm = VADD(Tj, Tl); T16 = VSUB(Tl, Tj); T17 = VSUB(T15, T16); T1h = VADD(T15, T16); T1i = VADD(T18, T19); T1a = VSUB(T18, T19); Tf = VADD(T9, Te); TV = VFNMS(LDK(KP500000000), Te, T9); TW = VFNMS(LDK(KP500000000), Tm, Th); Tn = VADD(Th, Tm); To = VADD(Tf, Tn); TL = VSUB(Tf, Tn); TY = VFNMS(LDK(KP500000000), Tv, Tq); Tw = VADD(Tq, Tv); TE = VADD(Ty, TD); TZ = VFNMS(LDK(KP500000000), TD, Ty); TK = VSUB(Tw, TE); TF = VADD(Tw, TE); T1j = VADD(T1h, T1i); T1l = VSUB(T1h, T1i); TX = VADD(TV, TW); T1d = VSUB(TV, TW); T1e = VSUB(TY, TZ); T10 = VADD(TY, TZ); T11 = VADD(TX, T10); T13 = VSUB(TX, T10); T12 = VFNMS(LDK(KP250000000), T11, TU); T1G = VADD(TU, T11); T1H = VMUL(LDK(KP866025403), VADD(T1g, T1j)); T1k = VFNMS(LDK(KP250000000), T1j, T1g); T1J = VFMAI(T1H, T1G); T1I = VFNMSI(T1H, T1G); ST(&(x[WS(rs, 5)]), T1I, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 10)]), T1J, ms, &(x[0])); TQ = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TK, TL)); TM = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TL, TK)); TG = VADD(To, TF); TI = VSUB(To, TF); TT = VADD(T7, TG); TH = VFNMS(LDK(KP250000000), TG, T7); TJ = VFNMS(LDK(KP559016994), TI, TH); TP = VFMA(LDK(KP559016994), TI, TH); ST(&(x[0]), TT, ms, &(x[0])); TS = VFMAI(TQ, TP); TR = VFNMSI(TQ, TP); ST(&(x[WS(rs, 9)]), TS, ms, &(x[WS(rs, 1)])); TN = VFNMSI(TM, TJ); TO = VFMAI(TM, TJ); ST(&(x[WS(rs, 3)]), TN, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 12)]), TO, ms, &(x[0])); ST(&(x[WS(rs, 6)]), TR, ms, &(x[0])); T1f = VFMA(LDK(KP618033988), T1e, T1d); T1x = VFNMS(LDK(KP618033988), T1d, T1e); T1y = VFNMS(LDK(KP559016994), T1l, T1k); T1m = VFMA(LDK(KP559016994), T1l, T1k); T1r = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1m, T1f)); T1n = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1m, T1f)); T1D = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1y, T1x)); T1z = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1y, T1x)); T1v = VFNMS(LDK(KP618033988), T17, T1a); T1b = VFMA(LDK(KP618033988), T1a, T17); T1u = VFNMS(LDK(KP559016994), T13, T12); T14 = VFMA(LDK(KP559016994), T13, T12); T1q = VFNMS(LDK(KP823639103), T1b, T14); T1c = VFMA(LDK(KP823639103), T1b, T14); T1C = VFNMS(LDK(KP823639103), T1v, T1u); T1w = VFMA(LDK(KP823639103), T1v, T1u); T1p = VFMAI(T1n, T1c); T1o = VFNMSI(T1n, T1c); ST(&(x[WS(rs, 1)]), T1o, ms, &(x[WS(rs, 1)])); T1F = VFMAI(T1D, T1C); T1E = VFNMSI(T1D, T1C); ST(&(x[WS(rs, 8)]), T1E, ms, &(x[0])); ST(&(x[WS(rs, 7)]), T1F, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 14)]), T1p, ms, &(x[0])); T1t = VFMAI(T1r, T1q); T1s = VFNMSI(T1r, T1q); ST(&(x[WS(rs, 11)]), T1s, ms, &(x[WS(rs, 1)])); T1B = VFMAI(T1z, T1w); T1A = VFNMSI(T1z, T1w); ST(&(x[WS(rs, 13)]), T1A, ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), T1B, ms, &(x[0])); ST(&(x[WS(rs, 4)]), T1t, ms, &(x[0])); } }