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 07:39:23 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include t3b.h -sign 1 */ Chris@10: Chris@10: /* Chris@10: * This function contains 138 FP additions, 118 FP multiplications, Chris@10: * (or, 92 additions, 72 multiplications, 46 fused multiply/add), Chris@10: * 90 stack variables, 4 constants, and 40 memory accesses Chris@10: */ Chris@10: #include "t3b.h" Chris@10: Chris@10: static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@10: { Chris@10: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT m; Chris@10: R *x; Chris@10: x = ii; Chris@10: 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(20, rs)) { Chris@10: V T19, T1u, T1p, T1x, T1m, T1w, T1t, TI; Chris@10: { Chris@10: V T2, T8, T3, Td; Chris@10: T2 = LDW(&(W[0])); Chris@10: T8 = LDW(&(W[TWVL * 2])); Chris@10: T3 = LDW(&(W[TWVL * 4])); Chris@10: Td = LDW(&(W[TWVL * 6])); Chris@10: { Chris@10: V T7, T1g, T1F, T23, T1n, Tp, T18, T27, T1P, T1I, TU, T1L, T28, T1S, T1o; Chris@10: V TE, T1l, T1j, T26, T2e; Chris@10: { Chris@10: V T1, T1e, T5, T1b; Chris@10: T1 = LD(&(x[0]), ms, &(x[0])); Chris@10: T1e = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); Chris@10: T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Chris@10: T1b = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@10: { Chris@10: V TA, Tx, TQ, T1O, T10, Th, T1G, T1R, T17, T1J, To, Ts, TR, Tv, TK; Chris@10: V TM, TP, Ty, TB; Chris@10: { Chris@10: V Tq, Tt, T13, T16, Tk, Tn; Chris@10: { Chris@10: V Tl, Ti, T11, T14, TV, Tc, T6, Tb, Tf, TW, TY, T1f; Chris@10: { Chris@10: V T1d, Ta, T9, T4; Chris@10: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@10: TA = VZMULJ(T2, T8); Chris@10: T9 = VZMUL(T2, T8); Chris@10: Tx = VZMUL(T8, T3); Chris@10: Tl = VZMULJ(T8, T3); Chris@10: T4 = VZMUL(T2, T3); Chris@10: Tq = VZMULJ(T2, T3); Chris@10: Tt = VZMULJ(T2, Td); Chris@10: Ti = VZMULJ(T8, Td); Chris@10: T11 = VZMULJ(TA, Td); Chris@10: T14 = VZMULJ(TA, T3); Chris@10: TQ = VZMUL(TA, T3); Chris@10: T1d = VZMULJ(T9, Td); Chris@10: TV = VZMUL(T9, T3); Chris@10: Tc = VZMULJ(T9, T3); Chris@10: T6 = VZMUL(T4, T5); Chris@10: Tb = VZMUL(T9, Ta); Chris@10: Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0])); Chris@10: TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Chris@10: TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@10: T1f = VZMUL(T1d, T1e); Chris@10: } Chris@10: { Chris@10: V T1D, TX, TZ, T15, T1E, Tg, T12, T1c, Te, Tj, Tm; Chris@10: T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)])); Chris@10: T1c = VZMUL(Tc, T1b); Chris@10: Te = VZMULJ(Tc, Td); Chris@10: T7 = VSUB(T1, T6); Chris@10: T1D = VADD(T1, T6); Chris@10: TX = VZMUL(TV, TW); Chris@10: TZ = VZMUL(T8, TY); Chris@10: T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@10: T13 = VZMUL(T11, T12); Chris@10: T1g = VSUB(T1c, T1f); Chris@10: T1E = VADD(T1c, T1f); Chris@10: Tg = VZMUL(Te, Tf); Chris@10: Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0])); Chris@10: Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@10: T1O = VADD(TX, TZ); Chris@10: T10 = VSUB(TX, TZ); Chris@10: T16 = VZMUL(T14, T15); Chris@10: T1F = VSUB(T1D, T1E); Chris@10: T23 = VADD(T1D, T1E); Chris@10: Th = VSUB(Tb, Tg); Chris@10: T1G = VADD(Tb, Tg); Chris@10: Tk = VZMUL(Ti, Tj); Chris@10: Tn = VZMUL(Tl, Tm); Chris@10: } Chris@10: } Chris@10: { Chris@10: V Tr, Tu, TJ, TL, TO; Chris@10: Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@10: T1R = VADD(T13, T16); Chris@10: T17 = VSUB(T13, T16); Chris@10: Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0])); Chris@10: TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@10: TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)])); Chris@10: TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@10: T1J = VADD(Tk, Tn); Chris@10: To = VSUB(Tk, Tn); Chris@10: Ts = VZMUL(Tq, Tr); Chris@10: TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Chris@10: Tv = VZMUL(Tt, Tu); Chris@10: TK = VZMUL(T3, TJ); Chris@10: TM = VZMUL(Td, TL); Chris@10: TP = VZMUL(T2, TO); Chris@10: Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Chris@10: TB = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1N, Tw, T1H, TN, Tz, TC, T1i, TT, T1K, TS; Chris@10: T1n = VSUB(Th, To); Chris@10: Tp = VADD(Th, To); Chris@10: TS = VZMUL(TQ, TR); Chris@10: T1N = VADD(Ts, Tv); Chris@10: Tw = VSUB(Ts, Tv); Chris@10: T1H = VADD(TK, TM); Chris@10: TN = VSUB(TK, TM); Chris@10: Tz = VZMUL(Tx, Ty); Chris@10: TC = VZMUL(TA, TB); Chris@10: T18 = VSUB(T10, T17); Chris@10: T1i = VADD(T10, T17); Chris@10: TT = VSUB(TP, TS); Chris@10: T1K = VADD(TP, TS); Chris@10: T27 = VADD(T1N, T1O); Chris@10: T1P = VSUB(T1N, T1O); Chris@10: { Chris@10: V TD, T1Q, T24, T1h, T25; Chris@10: TD = VSUB(Tz, TC); Chris@10: T1Q = VADD(Tz, TC); Chris@10: T1I = VSUB(T1G, T1H); Chris@10: T24 = VADD(T1G, T1H); Chris@10: T1h = VADD(TN, TT); Chris@10: TU = VSUB(TN, TT); Chris@10: T25 = VADD(T1J, T1K); Chris@10: T1L = VSUB(T1J, T1K); Chris@10: T28 = VADD(T1Q, T1R); Chris@10: T1S = VSUB(T1Q, T1R); Chris@10: T1o = VSUB(Tw, TD); Chris@10: TE = VADD(Tw, TD); Chris@10: T1l = VSUB(T1h, T1i); Chris@10: T1j = VADD(T1h, T1i); Chris@10: T26 = VADD(T24, T25); Chris@10: T2e = VSUB(T24, T25); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1M, T1Z, T1Y, T1T, T29, T2f, TH, TF, T1k, T1C; Chris@10: T1M = VADD(T1I, T1L); Chris@10: T1Z = VSUB(T1I, T1L); Chris@10: T1Y = VSUB(T1P, T1S); Chris@10: T1T = VADD(T1P, T1S); Chris@10: T29 = VADD(T27, T28); Chris@10: T2f = VSUB(T27, T28); Chris@10: TH = VSUB(Tp, TE); Chris@10: TF = VADD(Tp, TE); Chris@10: T1k = VFNMS(LDK(KP250000000), T1j, T1g); Chris@10: T1C = VADD(T1g, T1j); Chris@10: { Chris@10: V T1W, T2c, TG, T2i, T2g, T22, T20, T1V, T2b, T1U, T2a, T1B; Chris@10: T19 = VFMA(LDK(KP618033988), T18, TU); Chris@10: T1u = VFNMS(LDK(KP618033988), TU, T18); Chris@10: T1W = VSUB(T1M, T1T); Chris@10: T1U = VADD(T1M, T1T); Chris@10: T2c = VSUB(T26, T29); Chris@10: T2a = VADD(T26, T29); Chris@10: TG = VFNMS(LDK(KP250000000), TF, T7); Chris@10: T1B = VADD(T7, TF); Chris@10: T2i = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T2e, T2f)); Chris@10: T2g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T2f, T2e)); Chris@10: T22 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1Y, T1Z)); Chris@10: T20 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1Z, T1Y)); Chris@10: ST(&(x[WS(rs, 10)]), VADD(T1F, T1U), ms, &(x[0])); Chris@10: T1V = VFNMS(LDK(KP250000000), T1U, T1F); Chris@10: ST(&(x[0]), VADD(T23, T2a), ms, &(x[0])); Chris@10: T2b = VFNMS(LDK(KP250000000), T2a, T23); Chris@10: ST(&(x[WS(rs, 5)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 15)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)])); Chris@10: T1p = VFMA(LDK(KP618033988), T1o, T1n); Chris@10: T1x = VFNMS(LDK(KP618033988), T1n, T1o); Chris@10: { Chris@10: V T21, T1X, T2h, T2d; Chris@10: T21 = VFMA(LDK(KP559016994), T1W, T1V); Chris@10: T1X = VFNMS(LDK(KP559016994), T1W, T1V); Chris@10: T2h = VFNMS(LDK(KP559016994), T2c, T2b); Chris@10: T2d = VFMA(LDK(KP559016994), T2c, T2b); Chris@10: ST(&(x[WS(rs, 18)]), VFMAI(T20, T1X), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 2)]), VFNMSI(T20, T1X), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 14)]), VFNMSI(T22, T21), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 6)]), VFMAI(T22, T21), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 16)]), VFMAI(T2g, T2d), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 4)]), VFNMSI(T2g, T2d), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 12)]), VFNMSI(T2i, T2h), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 8)]), VFMAI(T2i, T2h), ms, &(x[0])); Chris@10: T1m = VFMA(LDK(KP559016994), T1l, T1k); Chris@10: T1w = VFNMS(LDK(KP559016994), T1l, T1k); Chris@10: T1t = VFNMS(LDK(KP559016994), TH, TG); Chris@10: TI = VFMA(LDK(KP559016994), TH, TG); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1A, T1y, T1q, T1s, T1a, T1r, T1z, T1v; Chris@10: T1A = VFMA(LDK(KP951056516), T1x, T1w); Chris@10: T1y = VFNMS(LDK(KP951056516), T1x, T1w); Chris@10: T1q = VFMA(LDK(KP951056516), T1p, T1m); Chris@10: T1s = VFNMS(LDK(KP951056516), T1p, T1m); Chris@10: T1a = VFNMS(LDK(KP951056516), T19, TI); Chris@10: T1r = VFMA(LDK(KP951056516), T19, TI); Chris@10: T1z = VFNMS(LDK(KP951056516), T1u, T1t); Chris@10: T1v = VFMA(LDK(KP951056516), T1u, T1t); Chris@10: ST(&(x[WS(rs, 9)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 11)]), VFNMSI(T1s, T1r), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 1)]), VFMAI(T1q, T1a), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 19)]), VFNMSI(T1q, T1a), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 17)]), VFMAI(T1y, T1v), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 3)]), VFNMSI(T1y, T1v), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 13)]), VFMAI(T1A, T1z), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 7)]), VFNMSI(T1A, T1z), ms, &(x[WS(rs, 1)])); 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, 3), Chris@10: VTW(0, 9), Chris@10: VTW(0, 19), Chris@10: {TW_NEXT, VL, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {92, 72, 46, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t3bv_20) (planner *p) { Chris@10: X(kdft_dit_register) (p, t3bv_20, &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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include t3b.h -sign 1 */ Chris@10: Chris@10: /* Chris@10: * This function contains 138 FP additions, 92 FP multiplications, Chris@10: * (or, 126 additions, 80 multiplications, 12 fused multiply/add), Chris@10: * 73 stack variables, 4 constants, and 40 memory accesses Chris@10: */ Chris@10: #include "t3b.h" Chris@10: Chris@10: static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@10: { 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: { Chris@10: INT m; Chris@10: R *x; Chris@10: x = ii; Chris@10: 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(20, rs)) { Chris@10: V T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te; Chris@10: V T1g, Ti, Tt, T11; Chris@10: T2 = LDW(&(W[0])); Chris@10: T8 = LDW(&(W[TWVL * 2])); Chris@10: T9 = VZMUL(T2, T8); Chris@10: TA = VZMULJ(T2, T8); Chris@10: T3 = LDW(&(W[TWVL * 4])); Chris@10: Tc = VZMULJ(T9, T3); Chris@10: T4 = VZMUL(T2, T3); Chris@10: TV = VZMUL(T9, T3); Chris@10: T14 = VZMULJ(TA, T3); Chris@10: Tl = VZMULJ(T8, T3); Chris@10: Tq = VZMULJ(T2, T3); Chris@10: Tx = VZMUL(T8, T3); Chris@10: TQ = VZMUL(TA, T3); Chris@10: Td = LDW(&(W[TWVL * 6])); Chris@10: Te = VZMULJ(Tc, Td); Chris@10: T1g = VZMULJ(T9, Td); Chris@10: Ti = VZMULJ(T8, Td); Chris@10: Tt = VZMULJ(T2, Td); Chris@10: T11 = VZMULJ(TA, Td); Chris@10: { Chris@10: V T7, T1j, T1U, T2a, TU, T1n, T1o, T18, Tp, TE, TF, T26, T27, T28, T1M; Chris@10: V T1P, T1W, T1b, T1c, T1k, T23, T24, T25, T1F, T1I, T1V, T1B, T1C; Chris@10: { Chris@10: V T1, T1i, T6, T1f, T1h, T5, T1e, T1S, T1T; Chris@10: T1 = LD(&(x[0]), ms, &(x[0])); Chris@10: T1h = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); Chris@10: T1i = VZMUL(T1g, T1h); Chris@10: T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Chris@10: T6 = VZMUL(T4, T5); Chris@10: T1e = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@10: T1f = VZMUL(Tc, T1e); Chris@10: T7 = VSUB(T1, T6); Chris@10: T1j = VSUB(T1f, T1i); Chris@10: T1S = VADD(T1, T6); Chris@10: T1T = VADD(T1f, T1i); Chris@10: T1U = VSUB(T1S, T1T); Chris@10: T2a = VADD(T1S, T1T); Chris@10: } Chris@10: { Chris@10: V Th, T1D, T10, T1L, T17, T1O, To, T1G, Tw, T1K, TN, T1E, TT, T1H, TD; Chris@10: V T1N; Chris@10: { Chris@10: V Tb, Tg, Ta, Tf; Chris@10: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@10: Tb = VZMUL(T9, Ta); Chris@10: Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0])); Chris@10: Tg = VZMUL(Te, Tf); Chris@10: Th = VSUB(Tb, Tg); Chris@10: T1D = VADD(Tb, Tg); Chris@10: } Chris@10: { Chris@10: V TX, TZ, TW, TY; Chris@10: TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Chris@10: TX = VZMUL(TV, TW); Chris@10: TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@10: TZ = VZMUL(T8, TY); Chris@10: T10 = VSUB(TX, TZ); Chris@10: T1L = VADD(TX, TZ); Chris@10: } Chris@10: { Chris@10: V T13, T16, T12, T15; Chris@10: T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)])); Chris@10: T13 = VZMUL(T11, T12); Chris@10: T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@10: T16 = VZMUL(T14, T15); Chris@10: T17 = VSUB(T13, T16); Chris@10: T1O = VADD(T13, T16); Chris@10: } Chris@10: { Chris@10: V Tk, Tn, Tj, Tm; Chris@10: Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0])); Chris@10: Tk = VZMUL(Ti, Tj); Chris@10: Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@10: Tn = VZMUL(Tl, Tm); Chris@10: To = VSUB(Tk, Tn); Chris@10: T1G = VADD(Tk, Tn); Chris@10: } Chris@10: { Chris@10: V Ts, Tv, Tr, Tu; Chris@10: Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@10: Ts = VZMUL(Tq, Tr); Chris@10: Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0])); Chris@10: Tv = VZMUL(Tt, Tu); Chris@10: Tw = VSUB(Ts, Tv); Chris@10: T1K = VADD(Ts, Tv); Chris@10: } Chris@10: { Chris@10: V TK, TM, TJ, TL; Chris@10: TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@10: TK = VZMUL(T3, TJ); Chris@10: TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)])); Chris@10: TM = VZMUL(Td, TL); Chris@10: TN = VSUB(TK, TM); Chris@10: T1E = VADD(TK, TM); Chris@10: } Chris@10: { Chris@10: V TP, TS, TO, TR; Chris@10: TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@10: TP = VZMUL(T2, TO); Chris@10: TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Chris@10: TS = VZMUL(TQ, TR); Chris@10: TT = VSUB(TP, TS); Chris@10: T1H = VADD(TP, TS); Chris@10: } Chris@10: { Chris@10: V Tz, TC, Ty, TB; Chris@10: Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Chris@10: Tz = VZMUL(Tx, Ty); Chris@10: TB = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@10: TC = VZMUL(TA, TB); Chris@10: TD = VSUB(Tz, TC); Chris@10: T1N = VADD(Tz, TC); Chris@10: } Chris@10: TU = VSUB(TN, TT); Chris@10: T1n = VSUB(Th, To); Chris@10: T1o = VSUB(Tw, TD); Chris@10: T18 = VSUB(T10, T17); Chris@10: Tp = VADD(Th, To); Chris@10: TE = VADD(Tw, TD); Chris@10: TF = VADD(Tp, TE); Chris@10: T26 = VADD(T1K, T1L); Chris@10: T27 = VADD(T1N, T1O); Chris@10: T28 = VADD(T26, T27); Chris@10: T1M = VSUB(T1K, T1L); Chris@10: T1P = VSUB(T1N, T1O); Chris@10: T1W = VADD(T1M, T1P); Chris@10: T1b = VADD(TN, TT); Chris@10: T1c = VADD(T10, T17); Chris@10: T1k = VADD(T1b, T1c); Chris@10: T23 = VADD(T1D, T1E); Chris@10: T24 = VADD(T1G, T1H); Chris@10: T25 = VADD(T23, T24); Chris@10: T1F = VSUB(T1D, T1E); Chris@10: T1I = VSUB(T1G, T1H); Chris@10: T1V = VADD(T1F, T1I); Chris@10: } Chris@10: T1B = VADD(T7, TF); Chris@10: T1C = VBYI(VADD(T1j, T1k)); Chris@10: ST(&(x[WS(rs, 15)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 5)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)])); Chris@10: { Chris@10: V T29, T2b, T2c, T2g, T2i, T2e, T2f, T2h, T2d; Chris@10: T29 = VMUL(LDK(KP559016994), VSUB(T25, T28)); Chris@10: T2b = VADD(T25, T28); Chris@10: T2c = VFNMS(LDK(KP250000000), T2b, T2a); Chris@10: T2e = VSUB(T23, T24); Chris@10: T2f = VSUB(T26, T27); Chris@10: T2g = VBYI(VFMA(LDK(KP951056516), T2e, VMUL(LDK(KP587785252), T2f))); Chris@10: T2i = VBYI(VFNMS(LDK(KP951056516), T2f, VMUL(LDK(KP587785252), T2e))); Chris@10: ST(&(x[0]), VADD(T2a, T2b), ms, &(x[0])); Chris@10: T2h = VSUB(T2c, T29); Chris@10: ST(&(x[WS(rs, 8)]), VSUB(T2h, T2i), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 12)]), VADD(T2i, T2h), ms, &(x[0])); Chris@10: T2d = VADD(T29, T2c); Chris@10: ST(&(x[WS(rs, 4)]), VSUB(T2d, T2g), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 16)]), VADD(T2g, T2d), ms, &(x[0])); Chris@10: } Chris@10: { Chris@10: V T1Z, T1X, T1Y, T1R, T21, T1J, T1Q, T22, T20; Chris@10: T1Z = VMUL(LDK(KP559016994), VSUB(T1V, T1W)); Chris@10: T1X = VADD(T1V, T1W); Chris@10: T1Y = VFNMS(LDK(KP250000000), T1X, T1U); Chris@10: T1J = VSUB(T1F, T1I); Chris@10: T1Q = VSUB(T1M, T1P); Chris@10: T1R = VBYI(VFNMS(LDK(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J))); Chris@10: T21 = VBYI(VFMA(LDK(KP951056516), T1J, VMUL(LDK(KP587785252), T1Q))); Chris@10: ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0])); Chris@10: T22 = VADD(T1Z, T1Y); Chris@10: ST(&(x[WS(rs, 6)]), VADD(T21, T22), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 14)]), VSUB(T22, T21), ms, &(x[0])); Chris@10: T20 = VSUB(T1Y, T1Z); Chris@10: ST(&(x[WS(rs, 2)]), VADD(T1R, T20), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 18)]), VSUB(T20, T1R), ms, &(x[0])); Chris@10: } Chris@10: { Chris@10: V T19, T1p, T1w, T1u, T1m, T1x, TI, T1t; Chris@10: T19 = VFNMS(LDK(KP951056516), T18, VMUL(LDK(KP587785252), TU)); Chris@10: T1p = VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1n)); Chris@10: T1w = VFMA(LDK(KP951056516), T1n, VMUL(LDK(KP587785252), T1o)); Chris@10: T1u = VFMA(LDK(KP951056516), TU, VMUL(LDK(KP587785252), T18)); Chris@10: { Chris@10: V T1d, T1l, TG, TH; Chris@10: T1d = VMUL(LDK(KP559016994), VSUB(T1b, T1c)); Chris@10: T1l = VFNMS(LDK(KP250000000), T1k, T1j); Chris@10: T1m = VSUB(T1d, T1l); Chris@10: T1x = VADD(T1d, T1l); Chris@10: TG = VFNMS(LDK(KP250000000), TF, T7); Chris@10: TH = VMUL(LDK(KP559016994), VSUB(Tp, TE)); Chris@10: TI = VSUB(TG, TH); Chris@10: T1t = VADD(TH, TG); Chris@10: } Chris@10: { Chris@10: V T1a, T1q, T1z, T1A; Chris@10: T1a = VSUB(TI, T19); Chris@10: T1q = VBYI(VSUB(T1m, T1p)); Chris@10: ST(&(x[WS(rs, 17)]), VSUB(T1a, T1q), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 3)]), VADD(T1a, T1q), ms, &(x[WS(rs, 1)])); Chris@10: T1z = VADD(T1t, T1u); Chris@10: T1A = VBYI(VSUB(T1x, T1w)); Chris@10: ST(&(x[WS(rs, 11)]), VSUB(T1z, T1A), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 9)]), VADD(T1z, T1A), ms, &(x[WS(rs, 1)])); Chris@10: } Chris@10: { Chris@10: V T1r, T1s, T1v, T1y; Chris@10: T1r = VADD(TI, T19); Chris@10: T1s = VBYI(VADD(T1p, T1m)); Chris@10: ST(&(x[WS(rs, 13)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 7)]), VADD(T1r, T1s), ms, &(x[WS(rs, 1)])); Chris@10: T1v = VSUB(T1t, T1u); Chris@10: T1y = VBYI(VADD(T1w, T1x)); Chris@10: ST(&(x[WS(rs, 19)]), VSUB(T1v, T1y), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 1)]), VADD(T1v, T1y), 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, 3), Chris@10: VTW(0, 9), Chris@10: VTW(0, 19), Chris@10: {TW_NEXT, VL, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {126, 80, 12, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t3bv_20) (planner *p) { Chris@10: X(kdft_dit_register) (p, t3bv_20, &desc); Chris@10: } Chris@10: #endif /* HAVE_FMA */