cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:38:04 EST 2012 */ cannam@95: cannam@95: #include "codelet-dft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name t1fv_15 -include t1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 92 FP additions, 77 FP multiplications, cannam@95: * (or, 50 additions, 35 multiplications, 42 fused multiply/add), cannam@95: * 81 stack variables, 8 constants, and 30 memory accesses cannam@95: */ cannam@95: #include "t1f.h" cannam@95: cannam@95: static void t1fv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP823639103, +0.823639103546331925877420039278190003029660514); cannam@95: DVK(KP910592997, +0.910592997310029334643087372129977886038870291); cannam@95: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: 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(15, rs)) { cannam@95: V Tq, Ty, Th, T1b, T10, Ts, TP, T7, Tu, TA, TC, Tj, Tk, TQ, Tf; cannam@95: { cannam@95: V T1, T4, T2, T9, Te; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@95: T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: { cannam@95: V T8, Tp, Tx, Tg; cannam@95: T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@95: Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@95: Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@95: { cannam@95: V Tb, Td, Tr, T6, Tt, Tz, TB, Ti; cannam@95: { cannam@95: V T5, T3, Ta, Tc; cannam@95: Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@95: Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@95: T5 = BYTWJ(&(W[TWVL * 18]), T4); cannam@95: T3 = BYTWJ(&(W[TWVL * 8]), T2); cannam@95: T9 = BYTWJ(&(W[TWVL * 4]), T8); cannam@95: Tq = BYTWJ(&(W[TWVL * 10]), Tp); cannam@95: Ty = BYTWJ(&(W[TWVL * 16]), Tx); cannam@95: Th = BYTWJ(&(W[TWVL * 22]), Tg); cannam@95: Tb = BYTWJ(&(W[TWVL * 14]), Ta); cannam@95: Td = BYTWJ(&(W[TWVL * 24]), Tc); cannam@95: Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@95: T1b = VSUB(T5, T3); cannam@95: T6 = VADD(T3, T5); cannam@95: Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@95: TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: Te = VADD(Tb, Td); cannam@95: T10 = VSUB(Td, Tb); cannam@95: Ts = BYTWJ(&(W[TWVL * 20]), Tr); cannam@95: TP = VFNMS(LDK(KP500000000), T6, T1); cannam@95: T7 = VADD(T1, T6); cannam@95: Tu = BYTWJ(&(W[0]), Tt); cannam@95: TA = BYTWJ(&(W[TWVL * 26]), Tz); cannam@95: TC = BYTWJ(&(W[TWVL * 6]), TB); cannam@95: Tj = BYTWJ(&(W[TWVL * 2]), Ti); cannam@95: Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: } cannam@95: TQ = VFNMS(LDK(KP500000000), Te, T9); cannam@95: Tf = VADD(T9, Te); cannam@95: } cannam@95: { cannam@95: V Tv, T13, TD, T14, Tl; cannam@95: Tv = VADD(Ts, Tu); cannam@95: T13 = VSUB(Tu, Ts); cannam@95: TD = VADD(TA, TC); cannam@95: T14 = VSUB(TC, TA); cannam@95: Tl = BYTWJ(&(W[TWVL * 12]), Tk); cannam@95: { cannam@95: V TT, Tw, T1d, T15, TU, TE, T11, Tm; cannam@95: TT = VFNMS(LDK(KP500000000), Tv, Tq); cannam@95: Tw = VADD(Tq, Tv); cannam@95: T1d = VADD(T13, T14); cannam@95: T15 = VSUB(T13, T14); cannam@95: TU = VFNMS(LDK(KP500000000), TD, Ty); cannam@95: TE = VADD(Ty, TD); cannam@95: T11 = VSUB(Tl, Tj); cannam@95: Tm = VADD(Tj, Tl); cannam@95: { cannam@95: V T19, TV, TK, TF, T1c, T12, TR, Tn; cannam@95: T19 = VSUB(TT, TU); cannam@95: TV = VADD(TT, TU); cannam@95: TK = VSUB(Tw, TE); cannam@95: TF = VADD(Tw, TE); cannam@95: T1c = VADD(T10, T11); cannam@95: T12 = VSUB(T10, T11); cannam@95: TR = VFNMS(LDK(KP500000000), Tm, Th); cannam@95: Tn = VADD(Th, Tm); cannam@95: { cannam@95: V T1g, T1e, T1m, T16, T18, TS, TL, To, T1f, T1u; cannam@95: T1g = VSUB(T1c, T1d); cannam@95: T1e = VADD(T1c, T1d); cannam@95: T1m = VFNMS(LDK(KP618033988), T12, T15); cannam@95: T16 = VFMA(LDK(KP618033988), T15, T12); cannam@95: T18 = VSUB(TQ, TR); cannam@95: TS = VADD(TQ, TR); cannam@95: TL = VSUB(Tf, Tn); cannam@95: To = VADD(Tf, Tn); cannam@95: T1f = VFNMS(LDK(KP250000000), T1e, T1b); cannam@95: T1u = VMUL(LDK(KP866025403), VADD(T1b, T1e)); cannam@95: { cannam@95: V T1o, T1a, TY, TO, TM, TG, TI, T1p, T1h, T1t, TX, TW; cannam@95: T1o = VFNMS(LDK(KP618033988), T18, T19); cannam@95: T1a = VFMA(LDK(KP618033988), T19, T18); cannam@95: TW = VADD(TS, TV); cannam@95: TY = VSUB(TS, TV); cannam@95: TO = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TK, TL)); cannam@95: TM = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TL, TK)); cannam@95: TG = VADD(To, TF); cannam@95: TI = VSUB(To, TF); cannam@95: T1p = VFNMS(LDK(KP559016994), T1g, T1f); cannam@95: T1h = VFMA(LDK(KP559016994), T1g, T1f); cannam@95: T1t = VADD(TP, TW); cannam@95: TX = VFNMS(LDK(KP250000000), TW, TP); cannam@95: { cannam@95: V T1q, T1s, T1k, T1i, T1l, TZ, TJ, TN, TH; cannam@95: ST(&(x[0]), VADD(T7, TG), ms, &(x[0])); cannam@95: TH = VFNMS(LDK(KP250000000), TG, T7); cannam@95: T1q = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1p, T1o)); cannam@95: T1s = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1p, T1o)); cannam@95: T1k = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1h, T1a)); cannam@95: T1i = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1h, T1a)); cannam@95: ST(&(x[WS(rs, 10)]), VFMAI(T1u, T1t), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 5)]), VFNMSI(T1u, T1t), ms, &(x[WS(rs, 1)])); cannam@95: T1l = VFNMS(LDK(KP559016994), TY, TX); cannam@95: TZ = VFMA(LDK(KP559016994), TY, TX); cannam@95: TJ = VFNMS(LDK(KP559016994), TI, TH); cannam@95: TN = VFMA(LDK(KP559016994), TI, TH); cannam@95: { cannam@95: V T1n, T1r, T1j, T17; cannam@95: T1n = VFMA(LDK(KP823639103), T1m, T1l); cannam@95: T1r = VFNMS(LDK(KP823639103), T1m, T1l); cannam@95: T1j = VFNMS(LDK(KP823639103), T16, TZ); cannam@95: T17 = VFMA(LDK(KP823639103), T16, TZ); cannam@95: ST(&(x[WS(rs, 12)]), VFMAI(TM, TJ), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 3)]), VFNMSI(TM, TJ), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 9)]), VFMAI(TO, TN), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 6)]), VFNMSI(TO, TN), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 2)]), VFMAI(T1q, T1n), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 13)]), VFNMSI(T1q, T1n), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 7)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 8)]), VFNMSI(T1s, T1r), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 4)]), VFMAI(T1k, T1j), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 11)]), VFNMSI(T1k, T1j), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 14)]), VFMAI(T1i, T17), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 1)]), VFNMSI(T1i, T17), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: VTW(0, 6), cannam@95: VTW(0, 7), cannam@95: VTW(0, 8), cannam@95: VTW(0, 9), cannam@95: VTW(0, 10), cannam@95: VTW(0, 11), cannam@95: VTW(0, 12), cannam@95: VTW(0, 13), cannam@95: VTW(0, 14), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 15, XSIMD_STRING("t1fv_15"), twinstr, &GENUS, {50, 35, 42, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1fv_15) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1fv_15, &desc); cannam@95: } cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name t1fv_15 -include t1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 92 FP additions, 53 FP multiplications, cannam@95: * (or, 78 additions, 39 multiplications, 14 fused multiply/add), cannam@95: * 52 stack variables, 10 constants, and 30 memory accesses cannam@95: */ cannam@95: #include "t1f.h" cannam@95: cannam@95: static void t1fv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP216506350, +0.216506350946109661690930792688234045867850657); cannam@95: DVK(KP484122918, +0.484122918275927110647408174972799951354115213); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DVK(KP509036960, +0.509036960455127183450980863393907648510733164); cannam@95: DVK(KP823639103, +0.823639103546331925877420039278190003029660514); cannam@95: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: 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(15, rs)) { cannam@95: V T1e, T7, TP, T12, T15, Tf, Tn, To, T1b, T1c, T1f, TQ, TR, TS, Tw; cannam@95: V TE, TF, TT, TU, TV; cannam@95: { cannam@95: V T1, T5, T3, T4, T2, T6; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@95: T5 = BYTWJ(&(W[TWVL * 18]), T4); cannam@95: T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: T3 = BYTWJ(&(W[TWVL * 8]), T2); cannam@95: T1e = VSUB(T5, T3); cannam@95: T6 = VADD(T3, T5); cannam@95: T7 = VADD(T1, T6); cannam@95: TP = VFNMS(LDK(KP500000000), T6, T1); cannam@95: } cannam@95: { cannam@95: V T9, Tq, Ty, Th, Te, T13, Tv, T10, TD, T11, Tm, T14; cannam@95: { cannam@95: V T8, Tp, Tx, Tg; cannam@95: T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: T9 = BYTWJ(&(W[TWVL * 4]), T8); cannam@95: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@95: Tq = BYTWJ(&(W[TWVL * 10]), Tp); cannam@95: Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@95: Ty = BYTWJ(&(W[TWVL * 16]), Tx); cannam@95: Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@95: Th = BYTWJ(&(W[TWVL * 22]), Tg); cannam@95: } cannam@95: { cannam@95: V Tb, Td, Ta, Tc; cannam@95: Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@95: Tb = BYTWJ(&(W[TWVL * 14]), Ta); cannam@95: Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@95: Td = BYTWJ(&(W[TWVL * 24]), Tc); cannam@95: Te = VADD(Tb, Td); cannam@95: T13 = VSUB(Td, Tb); cannam@95: } cannam@95: { cannam@95: V Ts, Tu, Tr, Tt; cannam@95: Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@95: Ts = BYTWJ(&(W[TWVL * 20]), Tr); cannam@95: Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: Tu = BYTWJ(&(W[0]), Tt); cannam@95: Tv = VADD(Ts, Tu); cannam@95: T10 = VSUB(Tu, Ts); cannam@95: } cannam@95: { cannam@95: V TA, TC, Tz, TB; cannam@95: Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@95: TA = BYTWJ(&(W[TWVL * 26]), Tz); cannam@95: TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: TC = BYTWJ(&(W[TWVL * 6]), TB); cannam@95: TD = VADD(TA, TC); cannam@95: T11 = VSUB(TC, TA); cannam@95: } cannam@95: { cannam@95: V Tj, Tl, Ti, Tk; cannam@95: Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: Tj = BYTWJ(&(W[TWVL * 2]), Ti); cannam@95: Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@95: Tl = BYTWJ(&(W[TWVL * 12]), Tk); cannam@95: Tm = VADD(Tj, Tl); cannam@95: T14 = VSUB(Tl, Tj); cannam@95: } cannam@95: T12 = VSUB(T10, T11); cannam@95: T15 = VSUB(T13, T14); cannam@95: Tf = VADD(T9, Te); cannam@95: Tn = VADD(Th, Tm); cannam@95: To = VADD(Tf, Tn); cannam@95: T1b = VADD(T13, T14); cannam@95: T1c = VADD(T10, T11); cannam@95: T1f = VADD(T1b, T1c); cannam@95: TQ = VFNMS(LDK(KP500000000), Te, T9); cannam@95: TR = VFNMS(LDK(KP500000000), Tm, Th); cannam@95: TS = VADD(TQ, TR); cannam@95: Tw = VADD(Tq, Tv); cannam@95: TE = VADD(Ty, TD); cannam@95: TF = VADD(Tw, TE); cannam@95: TT = VFNMS(LDK(KP500000000), Tv, Tq); cannam@95: TU = VFNMS(LDK(KP500000000), TD, Ty); cannam@95: TV = VADD(TT, TU); cannam@95: } cannam@95: { cannam@95: V TI, TG, TH, TM, TO, TK, TL, TN, TJ; cannam@95: TI = VMUL(LDK(KP559016994), VSUB(To, TF)); cannam@95: TG = VADD(To, TF); cannam@95: TH = VFNMS(LDK(KP250000000), TG, T7); cannam@95: TK = VSUB(Tw, TE); cannam@95: TL = VSUB(Tf, Tn); cannam@95: TM = VBYI(VFNMS(LDK(KP587785252), TL, VMUL(LDK(KP951056516), TK))); cannam@95: TO = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TK))); cannam@95: ST(&(x[0]), VADD(T7, TG), ms, &(x[0])); cannam@95: TN = VADD(TI, TH); cannam@95: ST(&(x[WS(rs, 6)]), VSUB(TN, TO), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 9)]), VADD(TO, TN), ms, &(x[WS(rs, 1)])); cannam@95: TJ = VSUB(TH, TI); cannam@95: ST(&(x[WS(rs, 3)]), VSUB(TJ, TM), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 12)]), VADD(TM, TJ), ms, &(x[0])); cannam@95: } cannam@95: { cannam@95: V T16, T1m, T1u, T1h, T1o, T1a, T1p, TZ, T1t, T1l, T1d, T1g; cannam@95: T16 = VFNMS(LDK(KP509036960), T15, VMUL(LDK(KP823639103), T12)); cannam@95: T1m = VFMA(LDK(KP823639103), T15, VMUL(LDK(KP509036960), T12)); cannam@95: T1u = VBYI(VMUL(LDK(KP866025403), VADD(T1e, T1f))); cannam@95: T1d = VMUL(LDK(KP484122918), VSUB(T1b, T1c)); cannam@95: T1g = VFNMS(LDK(KP216506350), T1f, VMUL(LDK(KP866025403), T1e)); cannam@95: T1h = VSUB(T1d, T1g); cannam@95: T1o = VADD(T1d, T1g); cannam@95: { cannam@95: V T18, T19, TY, TW, TX; cannam@95: T18 = VSUB(TT, TU); cannam@95: T19 = VSUB(TQ, TR); cannam@95: T1a = VFNMS(LDK(KP587785252), T19, VMUL(LDK(KP951056516), T18)); cannam@95: T1p = VFMA(LDK(KP951056516), T19, VMUL(LDK(KP587785252), T18)); cannam@95: TY = VMUL(LDK(KP559016994), VSUB(TS, TV)); cannam@95: TW = VADD(TS, TV); cannam@95: TX = VFNMS(LDK(KP250000000), TW, TP); cannam@95: TZ = VSUB(TX, TY); cannam@95: T1t = VADD(TP, TW); cannam@95: T1l = VADD(TY, TX); cannam@95: } cannam@95: { cannam@95: V T17, T1i, T1r, T1s; cannam@95: ST(&(x[WS(rs, 5)]), VSUB(T1t, T1u), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 10)]), VADD(T1t, T1u), ms, &(x[0])); cannam@95: T17 = VSUB(TZ, T16); cannam@95: T1i = VBYI(VSUB(T1a, T1h)); cannam@95: ST(&(x[WS(rs, 8)]), VSUB(T17, T1i), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 7)]), VADD(T17, T1i), ms, &(x[WS(rs, 1)])); cannam@95: T1r = VSUB(T1l, T1m); cannam@95: T1s = VBYI(VADD(T1p, T1o)); cannam@95: ST(&(x[WS(rs, 11)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VADD(T1r, T1s), ms, &(x[0])); cannam@95: { cannam@95: V T1n, T1q, T1j, T1k; cannam@95: T1n = VADD(T1l, T1m); cannam@95: T1q = VBYI(VSUB(T1o, T1p)); cannam@95: ST(&(x[WS(rs, 14)]), VSUB(T1n, T1q), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 1)]), VADD(T1n, T1q), ms, &(x[WS(rs, 1)])); cannam@95: T1j = VADD(TZ, T16); cannam@95: T1k = VBYI(VADD(T1a, T1h)); cannam@95: ST(&(x[WS(rs, 13)]), VSUB(T1j, T1k), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 2)]), VADD(T1j, T1k), ms, &(x[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: VTW(0, 6), cannam@95: VTW(0, 7), cannam@95: VTW(0, 8), cannam@95: VTW(0, 9), cannam@95: VTW(0, 10), cannam@95: VTW(0, 11), cannam@95: VTW(0, 12), cannam@95: VTW(0, 13), cannam@95: VTW(0, 14), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 15, XSIMD_STRING("t1fv_15"), twinstr, &GENUS, {78, 39, 14, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1fv_15) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1fv_15, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */