cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:44:21 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name t1bv_15 -include t1b.h -sign 1 */ cannam@127: cannam@127: /* cannam@127: * This function contains 92 FP additions, 77 FP multiplications, cannam@127: * (or, 50 additions, 35 multiplications, 42 fused multiply/add), cannam@127: * 81 stack variables, 8 constants, and 30 memory accesses cannam@127: */ cannam@127: #include "t1b.h" cannam@127: cannam@127: static void t1bv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP823639103, +0.823639103546331925877420039278190003029660514); cannam@127: DVK(KP910592997, +0.910592997310029334643087372129977886038870291); cannam@127: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@127: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@127: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ii; cannam@127: 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@127: V Tq, Ty, Th, TV, TK, Ts, T1f, T7, Tu, TA, TC, Tj, Tk, T1g, Tf; cannam@127: { cannam@127: V T1, T4, T2, T9, Te; cannam@127: T1 = LD(&(x[0]), ms, &(x[0])); cannam@127: T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@127: T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@127: { cannam@127: V T8, Tp, Tx, Tg; cannam@127: T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@127: Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@127: Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@127: { cannam@127: V Tb, Td, Tr, T6, Tt, Tz, TB, Ti; cannam@127: { cannam@127: V T5, T3, Ta, Tc; cannam@127: Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@127: Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@127: T5 = BYTW(&(W[TWVL * 18]), T4); cannam@127: T3 = BYTW(&(W[TWVL * 8]), T2); cannam@127: T9 = BYTW(&(W[TWVL * 4]), T8); cannam@127: Tq = BYTW(&(W[TWVL * 10]), Tp); cannam@127: Ty = BYTW(&(W[TWVL * 16]), Tx); cannam@127: Th = BYTW(&(W[TWVL * 22]), Tg); cannam@127: Tb = BYTW(&(W[TWVL * 14]), Ta); cannam@127: Td = BYTW(&(W[TWVL * 24]), Tc); cannam@127: Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@127: TV = VSUB(T3, T5); cannam@127: T6 = VADD(T3, T5); cannam@127: Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@127: TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@127: Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: Te = VADD(Tb, Td); cannam@127: TK = VSUB(Tb, Td); cannam@127: Ts = BYTW(&(W[TWVL * 20]), Tr); cannam@127: T1f = VADD(T1, T6); cannam@127: T7 = VFNMS(LDK(KP500000000), T6, T1); cannam@127: Tu = BYTW(&(W[0]), Tt); cannam@127: TA = BYTW(&(W[TWVL * 26]), Tz); cannam@127: TC = BYTW(&(W[TWVL * 6]), TB); cannam@127: Tj = BYTW(&(W[TWVL * 2]), Ti); cannam@127: Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: T1g = VADD(T9, Te); cannam@127: Tf = VFNMS(LDK(KP500000000), Te, T9); cannam@127: } cannam@127: { cannam@127: V Tv, TN, TD, TO, Tl; cannam@127: Tv = VADD(Ts, Tu); cannam@127: TN = VSUB(Ts, Tu); cannam@127: TD = VADD(TA, TC); cannam@127: TO = VSUB(TA, TC); cannam@127: Tl = BYTW(&(W[TWVL * 12]), Tk); cannam@127: { cannam@127: V Tw, T1j, TX, TP, TE, T1k, TL, Tm; cannam@127: Tw = VFNMS(LDK(KP500000000), Tv, Tq); cannam@127: T1j = VADD(Tq, Tv); cannam@127: TX = VADD(TN, TO); cannam@127: TP = VSUB(TN, TO); cannam@127: TE = VFNMS(LDK(KP500000000), TD, Ty); cannam@127: T1k = VADD(Ty, TD); cannam@127: TL = VSUB(Tj, Tl); cannam@127: Tm = VADD(Tj, Tl); cannam@127: { cannam@127: V TT, TF, T1q, T1l, TW, TM, T1h, Tn; cannam@127: TT = VSUB(Tw, TE); cannam@127: TF = VADD(Tw, TE); cannam@127: T1q = VSUB(T1j, T1k); cannam@127: T1l = VADD(T1j, T1k); cannam@127: TW = VADD(TK, TL); cannam@127: TM = VSUB(TK, TL); cannam@127: T1h = VADD(Th, Tm); cannam@127: Tn = VFNMS(LDK(KP500000000), Tm, Th); cannam@127: { cannam@127: V T10, TY, T16, TQ, T1r, T1i, TS, To, TZ, T1e; cannam@127: T10 = VSUB(TW, TX); cannam@127: TY = VADD(TW, TX); cannam@127: T16 = VFNMS(LDK(KP618033988), TM, TP); cannam@127: TQ = VFMA(LDK(KP618033988), TP, TM); cannam@127: T1r = VSUB(T1g, T1h); cannam@127: T1i = VADD(T1g, T1h); cannam@127: TS = VSUB(Tf, Tn); cannam@127: To = VADD(Tf, Tn); cannam@127: TZ = VFNMS(LDK(KP250000000), TY, TV); cannam@127: T1e = VMUL(LDK(KP866025403), VADD(TV, TY)); cannam@127: { cannam@127: V T1u, T1s, T1o, T18, TU, TG, TI, T19, T11, T1n, T1m; cannam@127: T1u = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1q, T1r)); cannam@127: T1s = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1r, T1q)); cannam@127: T1m = VADD(T1i, T1l); cannam@127: T1o = VSUB(T1i, T1l); cannam@127: T18 = VFNMS(LDK(KP618033988), TS, TT); cannam@127: TU = VFMA(LDK(KP618033988), TT, TS); cannam@127: TG = VADD(To, TF); cannam@127: TI = VSUB(To, TF); cannam@127: T19 = VFNMS(LDK(KP559016994), T10, TZ); cannam@127: T11 = VFMA(LDK(KP559016994), T10, TZ); cannam@127: ST(&(x[0]), VADD(T1f, T1m), ms, &(x[0])); cannam@127: T1n = VFNMS(LDK(KP250000000), T1m, T1f); cannam@127: { cannam@127: V T1a, T1c, T14, T12, T1p, T1t, T15, TJ, T1d, TH; cannam@127: T1d = VADD(T7, TG); cannam@127: TH = VFNMS(LDK(KP250000000), TG, T7); cannam@127: T1a = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T19, T18)); cannam@127: T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T19, T18)); cannam@127: T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T11, TU)); cannam@127: T12 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T11, TU)); cannam@127: T1p = VFNMS(LDK(KP559016994), T1o, T1n); cannam@127: T1t = VFMA(LDK(KP559016994), T1o, T1n); cannam@127: ST(&(x[WS(rs, 10)]), VFMAI(T1e, T1d), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 5)]), VFNMSI(T1e, T1d), ms, &(x[WS(rs, 1)])); cannam@127: T15 = VFNMS(LDK(KP559016994), TI, TH); cannam@127: TJ = VFMA(LDK(KP559016994), TI, TH); cannam@127: { cannam@127: V T17, T1b, T13, TR; cannam@127: ST(&(x[WS(rs, 12)]), VFNMSI(T1s, T1p), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 3)]), VFMAI(T1s, T1p), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 9)]), VFNMSI(T1u, T1t), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 6)]), VFMAI(T1u, T1t), ms, &(x[0])); cannam@127: T17 = VFNMS(LDK(KP823639103), T16, T15); cannam@127: T1b = VFMA(LDK(KP823639103), T16, T15); cannam@127: T13 = VFMA(LDK(KP823639103), TQ, TJ); cannam@127: TR = VFNMS(LDK(KP823639103), TQ, TJ); cannam@127: ST(&(x[WS(rs, 13)]), VFMAI(T1a, T17), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 2)]), VFNMSI(T1a, T17), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 8)]), VFMAI(T1c, T1b), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 7)]), VFNMSI(T1c, T1b), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 11)]), VFMAI(T14, T13), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 4)]), VFNMSI(T14, T13), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 14)]), VFNMSI(T12, TR), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 1)]), VFMAI(T12, TR), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 2), cannam@127: VTW(0, 3), cannam@127: VTW(0, 4), cannam@127: VTW(0, 5), cannam@127: VTW(0, 6), cannam@127: VTW(0, 7), cannam@127: VTW(0, 8), cannam@127: VTW(0, 9), cannam@127: VTW(0, 10), cannam@127: VTW(0, 11), cannam@127: VTW(0, 12), cannam@127: VTW(0, 13), cannam@127: VTW(0, 14), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 15, XSIMD_STRING("t1bv_15"), twinstr, &GENUS, {50, 35, 42, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t1bv_15) (planner *p) { cannam@127: X(kdft_dit_register) (p, t1bv_15, &desc); cannam@127: } cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name t1bv_15 -include t1b.h -sign 1 */ cannam@127: cannam@127: /* cannam@127: * This function contains 92 FP additions, 53 FP multiplications, cannam@127: * (or, 78 additions, 39 multiplications, 14 fused multiply/add), cannam@127: * 52 stack variables, 10 constants, and 30 memory accesses cannam@127: */ cannam@127: #include "t1b.h" cannam@127: cannam@127: static void t1bv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP216506350, +0.216506350946109661690930792688234045867850657); cannam@127: DVK(KP484122918, +0.484122918275927110647408174972799951354115213); cannam@127: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@127: DVK(KP509036960, +0.509036960455127183450980863393907648510733164); cannam@127: DVK(KP823639103, +0.823639103546331925877420039278190003029660514); cannam@127: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@127: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ii; cannam@127: 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@127: V Ts, TV, T1f, TZ, T10, Tb, Tm, Tt, T1j, T1k, T1l, TI, TM, TR, Tz; cannam@127: V TD, TQ, T1g, T1h, T1i; cannam@127: { cannam@127: V TT, Tr, Tp, Tq, To, TU; cannam@127: TT = LD(&(x[0]), ms, &(x[0])); cannam@127: Tq = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@127: Tr = BYTW(&(W[TWVL * 18]), Tq); cannam@127: To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@127: Tp = BYTW(&(W[TWVL * 8]), To); cannam@127: Ts = VSUB(Tp, Tr); cannam@127: TU = VADD(Tp, Tr); cannam@127: TV = VFNMS(LDK(KP500000000), TU, TT); cannam@127: T1f = VADD(TT, TU); cannam@127: } cannam@127: { cannam@127: V Tx, TG, TK, TB, T5, Ty, Tg, TH, Tl, TL, Ta, TC; cannam@127: { cannam@127: V Tw, TF, TJ, TA; cannam@127: Tw = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: Tx = BYTW(&(W[TWVL * 4]), Tw); cannam@127: TF = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@127: TG = BYTW(&(W[TWVL * 10]), TF); cannam@127: TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@127: TK = BYTW(&(W[TWVL * 16]), TJ); cannam@127: TA = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@127: TB = BYTW(&(W[TWVL * 22]), TA); cannam@127: } cannam@127: { cannam@127: V T2, T4, T1, T3; cannam@127: T1 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@127: T2 = BYTW(&(W[TWVL * 14]), T1); cannam@127: T3 = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@127: T4 = BYTW(&(W[TWVL * 24]), T3); cannam@127: T5 = VSUB(T2, T4); cannam@127: Ty = VADD(T2, T4); cannam@127: } cannam@127: { cannam@127: V Td, Tf, Tc, Te; cannam@127: Tc = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@127: Td = BYTW(&(W[TWVL * 20]), Tc); cannam@127: Te = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: Tf = BYTW(&(W[0]), Te); cannam@127: Tg = VSUB(Td, Tf); cannam@127: TH = VADD(Td, Tf); cannam@127: } cannam@127: { cannam@127: V Ti, Tk, Th, Tj; cannam@127: Th = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@127: Ti = BYTW(&(W[TWVL * 26]), Th); cannam@127: Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@127: Tk = BYTW(&(W[TWVL * 6]), Tj); cannam@127: Tl = VSUB(Ti, Tk); cannam@127: TL = VADD(Ti, Tk); cannam@127: } cannam@127: { cannam@127: V T7, T9, T6, T8; cannam@127: T6 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: T7 = BYTW(&(W[TWVL * 2]), T6); cannam@127: T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@127: T9 = BYTW(&(W[TWVL * 12]), T8); cannam@127: Ta = VSUB(T7, T9); cannam@127: TC = VADD(T7, T9); cannam@127: } cannam@127: TZ = VSUB(T5, Ta); cannam@127: T10 = VSUB(Tg, Tl); cannam@127: Tb = VADD(T5, Ta); cannam@127: Tm = VADD(Tg, Tl); cannam@127: Tt = VADD(Tb, Tm); cannam@127: T1j = VADD(TG, TH); cannam@127: T1k = VADD(TK, TL); cannam@127: T1l = VADD(T1j, T1k); cannam@127: TI = VFNMS(LDK(KP500000000), TH, TG); cannam@127: TM = VFNMS(LDK(KP500000000), TL, TK); cannam@127: TR = VADD(TI, TM); cannam@127: Tz = VFNMS(LDK(KP500000000), Ty, Tx); cannam@127: TD = VFNMS(LDK(KP500000000), TC, TB); cannam@127: TQ = VADD(Tz, TD); cannam@127: T1g = VADD(Tx, Ty); cannam@127: T1h = VADD(TB, TC); cannam@127: T1i = VADD(T1g, T1h); cannam@127: } cannam@127: { cannam@127: V T1o, T1m, T1n, T1s, T1t, T1q, T1r, T1u, T1p; cannam@127: T1o = VMUL(LDK(KP559016994), VSUB(T1i, T1l)); cannam@127: T1m = VADD(T1i, T1l); cannam@127: T1n = VFNMS(LDK(KP250000000), T1m, T1f); cannam@127: T1q = VSUB(T1g, T1h); cannam@127: T1r = VSUB(T1j, T1k); cannam@127: T1s = VBYI(VFNMS(LDK(KP951056516), T1r, VMUL(LDK(KP587785252), T1q))); cannam@127: T1t = VBYI(VFMA(LDK(KP951056516), T1q, VMUL(LDK(KP587785252), T1r))); cannam@127: ST(&(x[0]), VADD(T1f, T1m), ms, &(x[0])); cannam@127: T1u = VADD(T1o, T1n); cannam@127: ST(&(x[WS(rs, 6)]), VADD(T1t, T1u), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 9)]), VSUB(T1u, T1t), ms, &(x[WS(rs, 1)])); cannam@127: T1p = VSUB(T1n, T1o); cannam@127: ST(&(x[WS(rs, 3)]), VSUB(T1p, T1s), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 12)]), VADD(T1s, T1p), ms, &(x[0])); cannam@127: } cannam@127: { cannam@127: V T11, T18, T1e, TO, T16, Tv, T15, TY, T1d, T19, TE, TN; cannam@127: T11 = VFMA(LDK(KP823639103), TZ, VMUL(LDK(KP509036960), T10)); cannam@127: T18 = VFNMS(LDK(KP823639103), T10, VMUL(LDK(KP509036960), TZ)); cannam@127: T1e = VBYI(VMUL(LDK(KP866025403), VADD(Ts, Tt))); cannam@127: TE = VSUB(Tz, TD); cannam@127: TN = VSUB(TI, TM); cannam@127: TO = VFMA(LDK(KP951056516), TE, VMUL(LDK(KP587785252), TN)); cannam@127: T16 = VFNMS(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TE)); cannam@127: { cannam@127: V Tn, Tu, TS, TW, TX; cannam@127: Tn = VMUL(LDK(KP484122918), VSUB(Tb, Tm)); cannam@127: Tu = VFNMS(LDK(KP216506350), Tt, VMUL(LDK(KP866025403), Ts)); cannam@127: Tv = VADD(Tn, Tu); cannam@127: T15 = VSUB(Tn, Tu); cannam@127: TS = VMUL(LDK(KP559016994), VSUB(TQ, TR)); cannam@127: TW = VADD(TQ, TR); cannam@127: TX = VFNMS(LDK(KP250000000), TW, TV); cannam@127: TY = VADD(TS, TX); cannam@127: T1d = VADD(TV, TW); cannam@127: T19 = VSUB(TX, TS); cannam@127: } cannam@127: { cannam@127: V TP, T12, T1b, T1c; cannam@127: ST(&(x[WS(rs, 5)]), VSUB(T1d, T1e), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 10)]), VADD(T1e, T1d), ms, &(x[0])); cannam@127: TP = VBYI(VADD(Tv, TO)); cannam@127: T12 = VSUB(TY, T11); cannam@127: ST(&(x[WS(rs, 1)]), VADD(TP, T12), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 14)]), VSUB(T12, TP), ms, &(x[0])); cannam@127: T1b = VBYI(VSUB(T16, T15)); cannam@127: T1c = VSUB(T19, T18); cannam@127: ST(&(x[WS(rs, 7)]), VADD(T1b, T1c), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 8)]), VSUB(T1c, T1b), ms, &(x[0])); cannam@127: { cannam@127: V T17, T1a, T13, T14; cannam@127: T17 = VBYI(VADD(T15, T16)); cannam@127: T1a = VADD(T18, T19); cannam@127: ST(&(x[WS(rs, 2)]), VADD(T17, T1a), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 13)]), VSUB(T1a, T17), ms, &(x[WS(rs, 1)])); cannam@127: T13 = VBYI(VSUB(Tv, TO)); cannam@127: T14 = VADD(T11, TY); cannam@127: ST(&(x[WS(rs, 4)]), VADD(T13, T14), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 11)]), VSUB(T14, T13), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 2), cannam@127: VTW(0, 3), cannam@127: VTW(0, 4), cannam@127: VTW(0, 5), cannam@127: VTW(0, 6), cannam@127: VTW(0, 7), cannam@127: VTW(0, 8), cannam@127: VTW(0, 9), cannam@127: VTW(0, 10), cannam@127: VTW(0, 11), cannam@127: VTW(0, 12), cannam@127: VTW(0, 13), cannam@127: VTW(0, 14), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 15, XSIMD_STRING("t1bv_15"), twinstr, &GENUS, {78, 39, 14, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t1bv_15) (planner *p) { cannam@127: X(kdft_dit_register) (p, t1bv_15, &desc); cannam@127: } cannam@127: #endif /* HAVE_FMA */