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 16:43:52 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3fv_16 -include t3f.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 98 FP additions, 86 FP multiplications, cannam@127: * (or, 64 additions, 52 multiplications, 34 fused multiply/add), cannam@127: * 70 stack variables, 3 constants, and 32 memory accesses cannam@127: */ cannam@127: #include "t3f.h" cannam@127: cannam@127: static void t3fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@127: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ri; cannam@127: for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) { cannam@127: V T13, Tg, TY, T14, T1A, T1q, T1f, T1x, T1r, T1i, Tt, T16, TB, T1j, T1k; cannam@127: V TH; cannam@127: { cannam@127: V T2, T8, Tu, T3; cannam@127: T2 = LDW(&(W[0])); cannam@127: T8 = LDW(&(W[TWVL * 2])); cannam@127: Tu = LDW(&(W[TWVL * 6])); cannam@127: T3 = LDW(&(W[TWVL * 4])); cannam@127: { cannam@127: V Ty, T1o, Tf, T1b, T7, Tr, TR, TX, T1g, Tl, To, Tw, TG, Tz, T1p; cannam@127: V T1e, TC; cannam@127: { cannam@127: V T1, T5, Ta, Td; cannam@127: T1 = LD(&(x[0]), ms, &(x[0])); cannam@127: T5 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@127: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@127: Td = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@127: { cannam@127: V Tx, TO, TE, Tb, Tm, Tp, TN, Te, T6, TW, TP, TS; cannam@127: { cannam@127: V TM, T9, TL, Tc, TU, T4, TV; cannam@127: TM = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@127: Tx = VZMULJ(T2, T8); cannam@127: T9 = VZMUL(T2, T8); cannam@127: TL = VZMULJ(T2, Tu); cannam@127: TO = VZMULJ(T8, T3); cannam@127: Tc = VZMUL(T8, T3); cannam@127: TU = VZMUL(T2, T3); cannam@127: T4 = VZMULJ(T2, T3); cannam@127: TV = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@127: TE = VZMUL(Tx, T3); cannam@127: Ty = VZMULJ(Tx, T3); cannam@127: Tb = VZMULJ(T9, Ta); cannam@127: Tm = VZMULJ(T9, T3); cannam@127: Tp = VZMUL(T9, T3); cannam@127: TN = VZMULJ(TL, TM); cannam@127: Te = VZMULJ(Tc, Td); cannam@127: T6 = VZMULJ(T4, T5); cannam@127: TW = VZMULJ(TU, TV); cannam@127: } cannam@127: TP = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@127: TS = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: { cannam@127: V TQ, TT, Ti, Tk, Tn, Th, Tq, Tj; cannam@127: Th = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: Tq = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@127: Tj = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@127: T1o = VSUB(Tb, Te); cannam@127: Tf = VADD(Tb, Te); cannam@127: T1b = VSUB(T1, T6); cannam@127: T7 = VADD(T1, T6); cannam@127: TQ = VZMULJ(TO, TP); cannam@127: TT = VZMULJ(Tx, TS); cannam@127: Ti = VZMULJ(T2, Th); cannam@127: Tr = VZMULJ(Tp, Tq); cannam@127: Tk = VZMULJ(T3, Tj); cannam@127: Tn = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@127: { cannam@127: V T1d, T1c, Tv, TF; cannam@127: Tv = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); cannam@127: TF = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@127: T1d = VSUB(TN, TQ); cannam@127: TR = VADD(TN, TQ); cannam@127: T1c = VSUB(TT, TW); cannam@127: TX = VADD(TT, TW); cannam@127: T1g = VSUB(Ti, Tk); cannam@127: Tl = VADD(Ti, Tk); cannam@127: To = VZMULJ(Tm, Tn); cannam@127: Tw = VZMULJ(Tu, Tv); cannam@127: TG = VZMULJ(TE, TF); cannam@127: Tz = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@127: T1p = VSUB(T1d, T1c); cannam@127: T1e = VADD(T1c, T1d); cannam@127: TC = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: V T1h, Ts, TA, TD; cannam@127: T13 = VADD(T7, Tf); cannam@127: Tg = VSUB(T7, Tf); cannam@127: T1h = VSUB(To, Tr); cannam@127: Ts = VADD(To, Tr); cannam@127: TY = VSUB(TR, TX); cannam@127: T14 = VADD(TX, TR); cannam@127: TA = VZMULJ(Ty, Tz); cannam@127: T1A = VFMA(LDK(KP707106781), T1p, T1o); cannam@127: T1q = VFNMS(LDK(KP707106781), T1p, T1o); cannam@127: T1f = VFMA(LDK(KP707106781), T1e, T1b); cannam@127: T1x = VFNMS(LDK(KP707106781), T1e, T1b); cannam@127: TD = VZMULJ(T8, TC); cannam@127: T1r = VFMA(LDK(KP414213562), T1g, T1h); cannam@127: T1i = VFNMS(LDK(KP414213562), T1h, T1g); cannam@127: Tt = VSUB(Tl, Ts); cannam@127: T16 = VADD(Tl, Ts); cannam@127: TB = VADD(Tw, TA); cannam@127: T1j = VSUB(Tw, TA); cannam@127: T1k = VSUB(TG, TD); cannam@127: TH = VADD(TD, TG); cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: V T15, T19, T1l, T1s, TI, T17; cannam@127: T15 = VADD(T13, T14); cannam@127: T19 = VSUB(T13, T14); cannam@127: T1l = VFNMS(LDK(KP414213562), T1k, T1j); cannam@127: T1s = VFMA(LDK(KP414213562), T1j, T1k); cannam@127: TI = VSUB(TB, TH); cannam@127: T17 = VADD(TB, TH); cannam@127: { cannam@127: V T1y, T1t, T1B, T1m; cannam@127: T1y = VADD(T1r, T1s); cannam@127: T1t = VSUB(T1r, T1s); cannam@127: T1B = VSUB(T1l, T1i); cannam@127: T1m = VADD(T1i, T1l); cannam@127: { cannam@127: V T18, T1a, TJ, TZ; cannam@127: T18 = VADD(T16, T17); cannam@127: T1a = VSUB(T17, T16); cannam@127: TJ = VADD(Tt, TI); cannam@127: TZ = VSUB(TI, Tt); cannam@127: { cannam@127: V T1u, T1w, T1z, T1D; cannam@127: T1u = VFNMS(LDK(KP923879532), T1t, T1q); cannam@127: T1w = VFMA(LDK(KP923879532), T1t, T1q); cannam@127: T1z = VFNMS(LDK(KP923879532), T1y, T1x); cannam@127: T1D = VFMA(LDK(KP923879532), T1y, T1x); cannam@127: { cannam@127: V T1n, T1v, T1C, T1E; cannam@127: T1n = VFNMS(LDK(KP923879532), T1m, T1f); cannam@127: T1v = VFMA(LDK(KP923879532), T1m, T1f); cannam@127: T1C = VFNMS(LDK(KP923879532), T1B, T1A); cannam@127: T1E = VFMA(LDK(KP923879532), T1B, T1A); cannam@127: ST(&(x[WS(rs, 12)]), VFNMSI(T1a, T19), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 4)]), VFMAI(T1a, T19), ms, &(x[0])); cannam@127: ST(&(x[0]), VADD(T15, T18), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 8)]), VSUB(T15, T18), ms, &(x[0])); cannam@127: { cannam@127: V T10, T12, TK, T11; cannam@127: T10 = VFNMS(LDK(KP707106781), TZ, TY); cannam@127: T12 = VFMA(LDK(KP707106781), TZ, TY); cannam@127: TK = VFNMS(LDK(KP707106781), TJ, Tg); cannam@127: T11 = VFMA(LDK(KP707106781), TJ, Tg); cannam@127: ST(&(x[WS(rs, 1)]), VFNMSI(T1w, T1v), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 15)]), VFMAI(T1w, T1v), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 7)]), VFMAI(T1u, T1n), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 9)]), VFNMSI(T1u, T1n), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 3)]), VFMAI(T1E, T1D), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 13)]), VFNMSI(T1E, T1D), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 11)]), VFMAI(T1C, T1z), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 5)]), VFNMSI(T1C, T1z), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 14)]), VFNMSI(T12, T11), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 2)]), VFMAI(T12, T11), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 10)]), VFMAI(T10, TK), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 6)]), VFNMSI(T10, TK), ms, &(x[0])); 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, 3), cannam@127: VTW(0, 9), cannam@127: VTW(0, 15), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 16, XSIMD_STRING("t3fv_16"), twinstr, &GENUS, {64, 52, 34, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t3fv_16) (planner *p) { cannam@127: X(kdft_dit_register) (p, t3fv_16, &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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3fv_16 -include t3f.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 98 FP additions, 64 FP multiplications, cannam@127: * (or, 94 additions, 60 multiplications, 4 fused multiply/add), cannam@127: * 51 stack variables, 3 constants, and 32 memory accesses cannam@127: */ cannam@127: #include "t3f.h" cannam@127: cannam@127: static void t3fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@127: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ri; cannam@127: for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) { cannam@127: V T4, T5, T6, To, T1, Ty, T7, T8, TO, TV, Te, Tp, TB, TH, Ts; cannam@127: T4 = LDW(&(W[0])); cannam@127: T5 = LDW(&(W[TWVL * 2])); cannam@127: T6 = VZMULJ(T4, T5); cannam@127: To = VZMUL(T4, T5); cannam@127: T1 = LDW(&(W[TWVL * 6])); cannam@127: Ty = VZMULJ(T4, T1); cannam@127: T7 = LDW(&(W[TWVL * 4])); cannam@127: T8 = VZMULJ(T6, T7); cannam@127: TO = VZMUL(T5, T7); cannam@127: TV = VZMULJ(T4, T7); cannam@127: Te = VZMUL(T6, T7); cannam@127: Tp = VZMULJ(To, T7); cannam@127: TB = VZMULJ(T5, T7); cannam@127: TH = VZMUL(T4, T7); cannam@127: Ts = VZMUL(To, T7); cannam@127: { cannam@127: V TY, T1f, TR, T1g, T1q, T1r, TL, TZ, T1l, T1m, T1n, Ti, T12, T1i, T1j; cannam@127: V T1k, Tw, T11, TU, TX, TW; cannam@127: TU = LD(&(x[0]), ms, &(x[0])); cannam@127: TW = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@127: TX = VZMULJ(TV, TW); cannam@127: TY = VSUB(TU, TX); cannam@127: T1f = VADD(TU, TX); cannam@127: { cannam@127: V TN, TQ, TM, TP; cannam@127: TM = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@127: TN = VZMULJ(To, TM); cannam@127: TP = LD(&(x[WS(rs, 12)]), ms, &(x[0])); cannam@127: TQ = VZMULJ(TO, TP); cannam@127: TR = VSUB(TN, TQ); cannam@127: T1g = VADD(TN, TQ); cannam@127: } cannam@127: { cannam@127: V TA, TJ, TD, TG, TE, TK; cannam@127: { cannam@127: V Tz, TI, TC, TF; cannam@127: Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); cannam@127: TA = VZMULJ(Ty, Tz); cannam@127: TI = LD(&(x[WS(rs, 10)]), ms, &(x[0])); cannam@127: TJ = VZMULJ(TH, TI); cannam@127: TC = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@127: TD = VZMULJ(TB, TC); cannam@127: TF = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: TG = VZMULJ(T6, TF); cannam@127: } cannam@127: T1q = VADD(TA, TD); cannam@127: T1r = VADD(TG, TJ); cannam@127: TE = VSUB(TA, TD); cannam@127: TK = VSUB(TG, TJ); cannam@127: TL = VMUL(LDK(KP707106781), VSUB(TE, TK)); cannam@127: TZ = VMUL(LDK(KP707106781), VADD(TK, TE)); cannam@127: } cannam@127: { cannam@127: V T3, Tg, Ta, Td, Tb, Th; cannam@127: { cannam@127: V T2, Tf, T9, Tc; cannam@127: T2 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); cannam@127: T3 = VZMULJ(T1, T2); cannam@127: Tf = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); cannam@127: Tg = VZMULJ(Te, Tf); cannam@127: T9 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@127: Ta = VZMULJ(T8, T9); cannam@127: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: Td = VZMULJ(T5, Tc); cannam@127: } cannam@127: T1l = VADD(T3, Ta); cannam@127: T1m = VADD(Td, Tg); cannam@127: T1n = VSUB(T1l, T1m); cannam@127: Tb = VSUB(T3, Ta); cannam@127: Th = VSUB(Td, Tg); cannam@127: Ti = VFNMS(LDK(KP923879532), Th, VMUL(LDK(KP382683432), Tb)); cannam@127: T12 = VFMA(LDK(KP923879532), Tb, VMUL(LDK(KP382683432), Th)); cannam@127: } cannam@127: { cannam@127: V Tk, Tu, Tm, Tr, Tn, Tv; cannam@127: { cannam@127: V Tj, Tt, Tl, Tq; cannam@127: Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: Tk = VZMULJ(T4, Tj); cannam@127: Tt = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); cannam@127: Tu = VZMULJ(Ts, Tt); cannam@127: Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@127: Tm = VZMULJ(T7, Tl); cannam@127: Tq = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@127: Tr = VZMULJ(Tp, Tq); cannam@127: } cannam@127: T1i = VADD(Tk, Tm); cannam@127: T1j = VADD(Tr, Tu); cannam@127: T1k = VSUB(T1i, T1j); cannam@127: Tn = VSUB(Tk, Tm); cannam@127: Tv = VSUB(Tr, Tu); cannam@127: Tw = VFMA(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tv)); cannam@127: T11 = VFNMS(LDK(KP382683432), Tv, VMUL(LDK(KP923879532), Tn)); cannam@127: } cannam@127: { cannam@127: V T1p, T1v, T1u, T1w; cannam@127: { cannam@127: V T1h, T1o, T1s, T1t; cannam@127: T1h = VSUB(T1f, T1g); cannam@127: T1o = VMUL(LDK(KP707106781), VADD(T1k, T1n)); cannam@127: T1p = VADD(T1h, T1o); cannam@127: T1v = VSUB(T1h, T1o); cannam@127: T1s = VSUB(T1q, T1r); cannam@127: T1t = VMUL(LDK(KP707106781), VSUB(T1n, T1k)); cannam@127: T1u = VBYI(VADD(T1s, T1t)); cannam@127: T1w = VBYI(VSUB(T1t, T1s)); cannam@127: } cannam@127: ST(&(x[WS(rs, 14)]), VSUB(T1p, T1u), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 6)]), VADD(T1v, T1w), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 2)]), VADD(T1p, T1u), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 10)]), VSUB(T1v, T1w), ms, &(x[0])); cannam@127: } cannam@127: { cannam@127: V T1z, T1D, T1C, T1E; cannam@127: { cannam@127: V T1x, T1y, T1A, T1B; cannam@127: T1x = VADD(T1f, T1g); cannam@127: T1y = VADD(T1r, T1q); cannam@127: T1z = VADD(T1x, T1y); cannam@127: T1D = VSUB(T1x, T1y); cannam@127: T1A = VADD(T1i, T1j); cannam@127: T1B = VADD(T1l, T1m); cannam@127: T1C = VADD(T1A, T1B); cannam@127: T1E = VBYI(VSUB(T1B, T1A)); cannam@127: } cannam@127: ST(&(x[WS(rs, 8)]), VSUB(T1z, T1C), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 4)]), VADD(T1D, T1E), ms, &(x[0])); cannam@127: ST(&(x[0]), VADD(T1z, T1C), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 12)]), VSUB(T1D, T1E), ms, &(x[0])); cannam@127: } cannam@127: { cannam@127: V TT, T15, T14, T16; cannam@127: { cannam@127: V Tx, TS, T10, T13; cannam@127: Tx = VSUB(Ti, Tw); cannam@127: TS = VSUB(TL, TR); cannam@127: TT = VBYI(VSUB(Tx, TS)); cannam@127: T15 = VBYI(VADD(TS, Tx)); cannam@127: T10 = VADD(TY, TZ); cannam@127: T13 = VADD(T11, T12); cannam@127: T14 = VSUB(T10, T13); cannam@127: T16 = VADD(T10, T13); cannam@127: } cannam@127: ST(&(x[WS(rs, 7)]), VADD(TT, T14), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 15)]), VSUB(T16, T15), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 9)]), VSUB(T14, TT), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 1)]), VADD(T15, T16), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: { cannam@127: V T19, T1d, T1c, T1e; cannam@127: { cannam@127: V T17, T18, T1a, T1b; cannam@127: T17 = VSUB(TY, TZ); cannam@127: T18 = VADD(Tw, Ti); cannam@127: T19 = VADD(T17, T18); cannam@127: T1d = VSUB(T17, T18); cannam@127: T1a = VADD(TR, TL); cannam@127: T1b = VSUB(T12, T11); cannam@127: T1c = VBYI(VADD(T1a, T1b)); cannam@127: T1e = VBYI(VSUB(T1b, T1a)); cannam@127: } cannam@127: ST(&(x[WS(rs, 13)]), VSUB(T19, T1c), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 5)]), VADD(T1d, T1e), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 3)]), VADD(T19, T1c), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 11)]), VSUB(T1d, T1e), ms, &(x[WS(rs, 1)])); 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, 3), cannam@127: VTW(0, 9), cannam@127: VTW(0, 15), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 16, XSIMD_STRING("t3fv_16"), twinstr, &GENUS, {94, 60, 4, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t3fv_16) (planner *p) { cannam@127: X(kdft_dit_register) (p, t3fv_16, &desc); cannam@127: } cannam@127: #endif /* HAVE_FMA */