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