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 07:39:04 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 t1bv_15 -include t1b.h -sign 1 */ 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 "t1b.h" Chris@10: Chris@10: static void t1bv_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 = ii; 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, TV, TK, Ts, T1f, T7, Tu, TA, TC, Tj, Tk, T1g, 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 = BYTW(&(W[TWVL * 18]), T4); Chris@10: T3 = BYTW(&(W[TWVL * 8]), T2); Chris@10: T9 = BYTW(&(W[TWVL * 4]), T8); Chris@10: Tq = BYTW(&(W[TWVL * 10]), Tp); Chris@10: Ty = BYTW(&(W[TWVL * 16]), Tx); Chris@10: Th = BYTW(&(W[TWVL * 22]), Tg); Chris@10: Tb = BYTW(&(W[TWVL * 14]), Ta); Chris@10: Td = BYTW(&(W[TWVL * 24]), Tc); Chris@10: Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Chris@10: TV = VSUB(T3, T5); 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: TK = VSUB(Tb, Td); Chris@10: Ts = BYTW(&(W[TWVL * 20]), Tr); Chris@10: T1f = VADD(T1, T6); Chris@10: T7 = VFNMS(LDK(KP500000000), T6, T1); Chris@10: Tu = BYTW(&(W[0]), Tt); Chris@10: TA = BYTW(&(W[TWVL * 26]), Tz); Chris@10: TC = BYTW(&(W[TWVL * 6]), TB); Chris@10: Tj = BYTW(&(W[TWVL * 2]), Ti); Chris@10: Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@10: } Chris@10: } Chris@10: T1g = VADD(T9, Te); Chris@10: Tf = VFNMS(LDK(KP500000000), Te, T9); Chris@10: } Chris@10: { Chris@10: V Tv, TN, TD, TO, Tl; Chris@10: Tv = VADD(Ts, Tu); Chris@10: TN = VSUB(Ts, Tu); Chris@10: TD = VADD(TA, TC); Chris@10: TO = VSUB(TA, TC); Chris@10: Tl = BYTW(&(W[TWVL * 12]), Tk); Chris@10: { Chris@10: V Tw, T1j, TX, TP, TE, T1k, TL, Tm; Chris@10: Tw = VFNMS(LDK(KP500000000), Tv, Tq); Chris@10: T1j = VADD(Tq, Tv); Chris@10: TX = VADD(TN, TO); Chris@10: TP = VSUB(TN, TO); Chris@10: TE = VFNMS(LDK(KP500000000), TD, Ty); Chris@10: T1k = VADD(Ty, TD); Chris@10: TL = VSUB(Tj, Tl); Chris@10: Tm = VADD(Tj, Tl); Chris@10: { Chris@10: V TT, TF, T1q, T1l, TW, TM, T1h, Tn; Chris@10: TT = VSUB(Tw, TE); Chris@10: TF = VADD(Tw, TE); Chris@10: T1q = VSUB(T1j, T1k); Chris@10: T1l = VADD(T1j, T1k); Chris@10: TW = VADD(TK, TL); Chris@10: TM = VSUB(TK, TL); Chris@10: T1h = VADD(Th, Tm); Chris@10: Tn = VFNMS(LDK(KP500000000), Tm, Th); Chris@10: { Chris@10: V T10, TY, T16, TQ, T1r, T1i, TS, To, TZ, T1e; Chris@10: T10 = VSUB(TW, TX); Chris@10: TY = VADD(TW, TX); Chris@10: T16 = VFNMS(LDK(KP618033988), TM, TP); Chris@10: TQ = VFMA(LDK(KP618033988), TP, TM); Chris@10: T1r = VSUB(T1g, T1h); Chris@10: T1i = VADD(T1g, T1h); Chris@10: TS = VSUB(Tf, Tn); Chris@10: To = VADD(Tf, Tn); Chris@10: TZ = VFNMS(LDK(KP250000000), TY, TV); Chris@10: T1e = VMUL(LDK(KP866025403), VADD(TV, TY)); Chris@10: { Chris@10: V T1u, T1s, T1o, T18, TU, TG, TI, T19, T11, T1n, T1m; Chris@10: T1u = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1q, T1r)); Chris@10: T1s = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1r, T1q)); Chris@10: T1m = VADD(T1i, T1l); Chris@10: T1o = VSUB(T1i, T1l); Chris@10: T18 = VFNMS(LDK(KP618033988), TS, TT); Chris@10: TU = VFMA(LDK(KP618033988), TT, TS); Chris@10: TG = VADD(To, TF); Chris@10: TI = VSUB(To, TF); Chris@10: T19 = VFNMS(LDK(KP559016994), T10, TZ); Chris@10: T11 = VFMA(LDK(KP559016994), T10, TZ); Chris@10: ST(&(x[0]), VADD(T1f, T1m), ms, &(x[0])); Chris@10: T1n = VFNMS(LDK(KP250000000), T1m, T1f); Chris@10: { Chris@10: V T1a, T1c, T14, T12, T1p, T1t, T15, TJ, T1d, TH; Chris@10: T1d = VADD(T7, TG); Chris@10: TH = VFNMS(LDK(KP250000000), TG, T7); Chris@10: T1a = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T19, T18)); Chris@10: T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T19, T18)); Chris@10: T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T11, TU)); Chris@10: T12 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T11, TU)); Chris@10: T1p = VFNMS(LDK(KP559016994), T1o, T1n); Chris@10: T1t = VFMA(LDK(KP559016994), T1o, T1n); Chris@10: ST(&(x[WS(rs, 10)]), VFMAI(T1e, T1d), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 5)]), VFNMSI(T1e, T1d), ms, &(x[WS(rs, 1)])); Chris@10: T15 = VFNMS(LDK(KP559016994), TI, TH); Chris@10: TJ = VFMA(LDK(KP559016994), TI, TH); Chris@10: { Chris@10: V T17, T1b, T13, TR; Chris@10: ST(&(x[WS(rs, 12)]), VFNMSI(T1s, T1p), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 3)]), VFMAI(T1s, T1p), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 9)]), VFNMSI(T1u, T1t), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 6)]), VFMAI(T1u, T1t), ms, &(x[0])); Chris@10: T17 = VFNMS(LDK(KP823639103), T16, T15); Chris@10: T1b = VFMA(LDK(KP823639103), T16, T15); Chris@10: T13 = VFMA(LDK(KP823639103), TQ, TJ); Chris@10: TR = VFNMS(LDK(KP823639103), TQ, TJ); Chris@10: ST(&(x[WS(rs, 13)]), VFMAI(T1a, T17), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 2)]), VFNMSI(T1a, T17), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 8)]), VFMAI(T1c, T1b), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 7)]), VFNMSI(T1c, T1b), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 11)]), VFMAI(T14, T13), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 4)]), VFNMSI(T14, T13), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 14)]), VFNMSI(T12, TR), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 1)]), VFMAI(T12, TR), 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("t1bv_15"), twinstr, &GENUS, {50, 35, 42, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t1bv_15) (planner *p) { Chris@10: X(kdft_dit_register) (p, t1bv_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 t1bv_15 -include t1b.h -sign 1 */ 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 "t1b.h" Chris@10: Chris@10: static void t1bv_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(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); 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 = ii; 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 Ts, TV, T1f, TZ, T10, Tb, Tm, Tt, T1j, T1k, T1l, TI, TM, TR, Tz; Chris@10: V TD, TQ, T1g, T1h, T1i; Chris@10: { Chris@10: V TT, Tr, Tp, Tq, To, TU; Chris@10: TT = LD(&(x[0]), ms, &(x[0])); Chris@10: Tq = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Chris@10: Tr = BYTW(&(W[TWVL * 18]), Tq); Chris@10: To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@10: Tp = BYTW(&(W[TWVL * 8]), To); Chris@10: Ts = VSUB(Tp, Tr); Chris@10: TU = VADD(Tp, Tr); Chris@10: TV = VFNMS(LDK(KP500000000), TU, TT); Chris@10: T1f = VADD(TT, TU); Chris@10: } Chris@10: { Chris@10: V Tx, TG, TK, TB, T5, Ty, Tg, TH, Tl, TL, Ta, TC; Chris@10: { Chris@10: V Tw, TF, TJ, TA; Chris@10: Tw = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@10: Tx = BYTW(&(W[TWVL * 4]), Tw); Chris@10: TF = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@10: TG = BYTW(&(W[TWVL * 10]), TF); Chris@10: TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@10: TK = BYTW(&(W[TWVL * 16]), TJ); Chris@10: TA = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Chris@10: TB = BYTW(&(W[TWVL * 22]), TA); Chris@10: } Chris@10: { Chris@10: V T2, T4, T1, T3; Chris@10: T1 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@10: T2 = BYTW(&(W[TWVL * 14]), T1); Chris@10: T3 = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Chris@10: T4 = BYTW(&(W[TWVL * 24]), T3); Chris@10: T5 = VSUB(T2, T4); Chris@10: Ty = VADD(T2, T4); Chris@10: } Chris@10: { Chris@10: V Td, Tf, Tc, Te; Chris@10: Tc = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Chris@10: Td = BYTW(&(W[TWVL * 20]), Tc); Chris@10: Te = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@10: Tf = BYTW(&(W[0]), Te); Chris@10: Tg = VSUB(Td, Tf); Chris@10: TH = VADD(Td, Tf); Chris@10: } Chris@10: { Chris@10: V Ti, Tk, Th, Tj; Chris@10: Th = LD(&(x[WS(rs, 14)]), ms, &(x[0])); Chris@10: Ti = BYTW(&(W[TWVL * 26]), Th); Chris@10: Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@10: Tk = BYTW(&(W[TWVL * 6]), Tj); Chris@10: Tl = VSUB(Ti, Tk); Chris@10: TL = VADD(Ti, Tk); Chris@10: } Chris@10: { Chris@10: V T7, T9, T6, T8; Chris@10: T6 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@10: T7 = BYTW(&(W[TWVL * 2]), T6); Chris@10: T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@10: T9 = BYTW(&(W[TWVL * 12]), T8); Chris@10: Ta = VSUB(T7, T9); Chris@10: TC = VADD(T7, T9); Chris@10: } Chris@10: TZ = VSUB(T5, Ta); Chris@10: T10 = VSUB(Tg, Tl); Chris@10: Tb = VADD(T5, Ta); Chris@10: Tm = VADD(Tg, Tl); Chris@10: Tt = VADD(Tb, Tm); Chris@10: T1j = VADD(TG, TH); Chris@10: T1k = VADD(TK, TL); Chris@10: T1l = VADD(T1j, T1k); Chris@10: TI = VFNMS(LDK(KP500000000), TH, TG); Chris@10: TM = VFNMS(LDK(KP500000000), TL, TK); Chris@10: TR = VADD(TI, TM); Chris@10: Tz = VFNMS(LDK(KP500000000), Ty, Tx); Chris@10: TD = VFNMS(LDK(KP500000000), TC, TB); Chris@10: TQ = VADD(Tz, TD); Chris@10: T1g = VADD(Tx, Ty); Chris@10: T1h = VADD(TB, TC); Chris@10: T1i = VADD(T1g, T1h); Chris@10: } Chris@10: { Chris@10: V T1o, T1m, T1n, T1s, T1t, T1q, T1r, T1u, T1p; Chris@10: T1o = VMUL(LDK(KP559016994), VSUB(T1i, T1l)); Chris@10: T1m = VADD(T1i, T1l); Chris@10: T1n = VFNMS(LDK(KP250000000), T1m, T1f); Chris@10: T1q = VSUB(T1g, T1h); Chris@10: T1r = VSUB(T1j, T1k); Chris@10: T1s = VBYI(VFNMS(LDK(KP951056516), T1r, VMUL(LDK(KP587785252), T1q))); Chris@10: T1t = VBYI(VFMA(LDK(KP951056516), T1q, VMUL(LDK(KP587785252), T1r))); Chris@10: ST(&(x[0]), VADD(T1f, T1m), ms, &(x[0])); Chris@10: T1u = VADD(T1o, T1n); Chris@10: ST(&(x[WS(rs, 6)]), VADD(T1t, T1u), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 9)]), VSUB(T1u, T1t), ms, &(x[WS(rs, 1)])); Chris@10: T1p = VSUB(T1n, T1o); Chris@10: ST(&(x[WS(rs, 3)]), VSUB(T1p, T1s), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 12)]), VADD(T1s, T1p), ms, &(x[0])); Chris@10: } Chris@10: { Chris@10: V T11, T18, T1e, TO, T16, Tv, T15, TY, T1d, T19, TE, TN; Chris@10: T11 = VFMA(LDK(KP823639103), TZ, VMUL(LDK(KP509036960), T10)); Chris@10: T18 = VFNMS(LDK(KP823639103), T10, VMUL(LDK(KP509036960), TZ)); Chris@10: T1e = VBYI(VMUL(LDK(KP866025403), VADD(Ts, Tt))); Chris@10: TE = VSUB(Tz, TD); Chris@10: TN = VSUB(TI, TM); Chris@10: TO = VFMA(LDK(KP951056516), TE, VMUL(LDK(KP587785252), TN)); Chris@10: T16 = VFNMS(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TE)); Chris@10: { Chris@10: V Tn, Tu, TS, TW, TX; Chris@10: Tn = VMUL(LDK(KP484122918), VSUB(Tb, Tm)); Chris@10: Tu = VFNMS(LDK(KP216506350), Tt, VMUL(LDK(KP866025403), Ts)); Chris@10: Tv = VADD(Tn, Tu); Chris@10: T15 = VSUB(Tn, Tu); Chris@10: TS = VMUL(LDK(KP559016994), VSUB(TQ, TR)); Chris@10: TW = VADD(TQ, TR); Chris@10: TX = VFNMS(LDK(KP250000000), TW, TV); Chris@10: TY = VADD(TS, TX); Chris@10: T1d = VADD(TV, TW); Chris@10: T19 = VSUB(TX, TS); Chris@10: } Chris@10: { Chris@10: V TP, T12, T1b, T1c; Chris@10: ST(&(x[WS(rs, 5)]), VSUB(T1d, T1e), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 10)]), VADD(T1e, T1d), ms, &(x[0])); Chris@10: TP = VBYI(VADD(Tv, TO)); Chris@10: T12 = VSUB(TY, T11); Chris@10: ST(&(x[WS(rs, 1)]), VADD(TP, T12), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 14)]), VSUB(T12, TP), ms, &(x[0])); Chris@10: T1b = VBYI(VSUB(T16, T15)); Chris@10: T1c = VSUB(T19, T18); Chris@10: ST(&(x[WS(rs, 7)]), VADD(T1b, T1c), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 8)]), VSUB(T1c, T1b), ms, &(x[0])); Chris@10: { Chris@10: V T17, T1a, T13, T14; Chris@10: T17 = VBYI(VADD(T15, T16)); Chris@10: T1a = VADD(T18, T19); Chris@10: ST(&(x[WS(rs, 2)]), VADD(T17, T1a), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 13)]), VSUB(T1a, T17), ms, &(x[WS(rs, 1)])); Chris@10: T13 = VBYI(VSUB(Tv, TO)); Chris@10: T14 = VADD(T11, TY); Chris@10: ST(&(x[WS(rs, 4)]), VADD(T13, T14), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 11)]), VSUB(T14, T13), ms, &(x[WS(rs, 1)])); 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("t1bv_15"), twinstr, &GENUS, {78, 39, 14, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t1bv_15) (planner *p) { Chris@10: X(kdft_dit_register) (p, t1bv_15, &desc); Chris@10: } Chris@10: #endif /* HAVE_FMA */