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