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 07:37:46 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_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include n2b.h -store-multiple 2 */ Chris@10: Chris@10: /* Chris@10: * This function contains 104 FP additions, 50 FP multiplications, Chris@10: * (or, 58 additions, 4 multiplications, 46 fused multiply/add), Chris@10: * 79 stack variables, 4 constants, and 50 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) { Chris@10: V T1H, T1I, TS, TA, TN, TV, T1M, T1N, T1O, T1P, T1R, T1S, TK, TU, TR; Chris@10: V Tl; Chris@10: { Chris@10: V T3, TE, T1r, T13, Ta, TL, Tz, TG, Ts, TF, Th, TM, T1u, T1C, T1n; Chris@10: V T1a, T1m, T1h, T1x, T1D, Tk, Ti; Chris@10: { Chris@10: V T1, T2, TC, TD; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: TC = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: TD = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@10: { Chris@10: V T14, T6, T1c, Tv, Tm, T1f, Ty, T17, T9, Tn, Tp, T1b, Td, Tq, Te; Chris@10: V Tf, T15, To; Chris@10: { Chris@10: V Tw, Tx, T7, T8, Tb, Tc; Chris@10: { Chris@10: V T4, T5, Tt, Tu, T11, T12; Chris@10: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@10: Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = VSUB(T1, T2); Chris@10: T11 = VADD(T1, T2); Chris@10: TE = VSUB(TC, TD); Chris@10: T12 = VADD(TC, TD); Chris@10: T14 = VADD(T4, T5); Chris@10: T6 = VSUB(T4, T5); Chris@10: T1c = VADD(Tt, Tu); Chris@10: Tv = VSUB(Tt, Tu); Chris@10: Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: T1r = VADD(T11, T12); Chris@10: T13 = VSUB(T11, T12); Chris@10: } Chris@10: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@10: Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1f = VADD(Tw, Tx); Chris@10: Ty = VSUB(Tw, Tx); Chris@10: T17 = VADD(T7, T8); Chris@10: T9 = VSUB(T7, T8); Chris@10: Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1b = VADD(Tb, Tc); Chris@10: Td = VSUB(Tb, Tc); Chris@10: Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: } Chris@10: Ta = VADD(T6, T9); Chris@10: TL = VSUB(T6, T9); Chris@10: T15 = VADD(Tm, Tn); Chris@10: To = VSUB(Tm, Tn); Chris@10: Tz = VSUB(Tv, Ty); Chris@10: TG = VADD(Tv, Ty); Chris@10: { Chris@10: V T1d, T1v, T18, Tr, T1e, Tg, T16, T1s; Chris@10: T1d = VSUB(T1b, T1c); Chris@10: T1v = VADD(T1b, T1c); Chris@10: T18 = VADD(Tp, Tq); Chris@10: Tr = VSUB(Tp, Tq); Chris@10: T1e = VADD(Te, Tf); Chris@10: Tg = VSUB(Te, Tf); Chris@10: T16 = VSUB(T14, T15); Chris@10: T1s = VADD(T14, T15); Chris@10: { Chris@10: V T1t, T19, T1w, T1g; Chris@10: T1t = VADD(T17, T18); Chris@10: T19 = VSUB(T17, T18); Chris@10: Ts = VSUB(To, Tr); Chris@10: TF = VADD(To, Tr); Chris@10: T1w = VADD(T1e, T1f); Chris@10: T1g = VSUB(T1e, T1f); Chris@10: Th = VADD(Td, Tg); Chris@10: TM = VSUB(Td, Tg); Chris@10: T1u = VADD(T1s, T1t); Chris@10: T1C = VSUB(T1s, T1t); Chris@10: T1n = VSUB(T16, T19); Chris@10: T1a = VADD(T16, T19); Chris@10: T1m = VSUB(T1d, T1g); Chris@10: T1h = VADD(T1d, T1g); Chris@10: T1x = VADD(T1v, T1w); Chris@10: T1D = VSUB(T1v, T1w); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Tk = VSUB(Ta, Th); Chris@10: Ti = VADD(Ta, Th); Chris@10: { Chris@10: V TJ, T1k, T1A, TZ, Tj, T1E, T1G, TI, T10, T1j, T1z, T1i, T1y, TH; Chris@10: TJ = VSUB(TF, TG); Chris@10: TH = VADD(TF, TG); Chris@10: T1i = VADD(T1a, T1h); Chris@10: T1k = VSUB(T1a, T1h); Chris@10: T1y = VADD(T1u, T1x); Chris@10: T1A = VSUB(T1u, T1x); Chris@10: TZ = VADD(T3, Ti); Chris@10: Tj = VFNMS(LDK(KP250000000), Ti, T3); Chris@10: T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C)); Chris@10: T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D)); Chris@10: TI = VFNMS(LDK(KP250000000), TH, TE); Chris@10: T10 = VADD(TE, TH); Chris@10: T1j = VFNMS(LDK(KP250000000), T1i, T13); Chris@10: T1H = VADD(T1r, T1y); Chris@10: STM2(&(xo[0]), T1H, ovs, &(xo[0])); Chris@10: T1z = VFNMS(LDK(KP250000000), T1y, T1r); Chris@10: T1I = VADD(T13, T1i); Chris@10: STM2(&(xo[20]), T1I, ovs, &(xo[0])); Chris@10: { Chris@10: V T1J, T1K, T1p, T1l, T1o, T1q, T1F, T1B, T1L, T1Q; Chris@10: TS = VFNMS(LDK(KP618033988), Ts, Tz); Chris@10: TA = VFMA(LDK(KP618033988), Tz, Ts); Chris@10: TN = VFMA(LDK(KP618033988), TM, TL); Chris@10: TV = VFNMS(LDK(KP618033988), TL, TM); Chris@10: T1J = VFMAI(T10, TZ); Chris@10: STM2(&(xo[10]), T1J, ovs, &(xo[2])); Chris@10: T1K = VFNMSI(T10, TZ); Chris@10: STM2(&(xo[30]), T1K, ovs, &(xo[2])); Chris@10: T1p = VFMA(LDK(KP559016994), T1k, T1j); Chris@10: T1l = VFNMS(LDK(KP559016994), T1k, T1j); Chris@10: T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m)); Chris@10: T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n)); Chris@10: T1F = VFNMS(LDK(KP559016994), T1A, T1z); Chris@10: T1B = VFMA(LDK(KP559016994), T1A, T1z); Chris@10: T1L = VFNMSI(T1q, T1p); Chris@10: STM2(&(xo[28]), T1L, ovs, &(xo[0])); Chris@10: STN2(&(xo[28]), T1L, T1K, ovs); Chris@10: T1M = VFMAI(T1q, T1p); Chris@10: STM2(&(xo[12]), T1M, ovs, &(xo[0])); Chris@10: T1N = VFMAI(T1o, T1l); Chris@10: STM2(&(xo[36]), T1N, ovs, &(xo[0])); Chris@10: T1O = VFNMSI(T1o, T1l); Chris@10: STM2(&(xo[4]), T1O, ovs, &(xo[0])); Chris@10: T1P = VFMAI(T1E, T1B); Chris@10: STM2(&(xo[32]), T1P, ovs, &(xo[0])); Chris@10: T1Q = VFNMSI(T1E, T1B); Chris@10: STM2(&(xo[8]), T1Q, ovs, &(xo[0])); Chris@10: STN2(&(xo[8]), T1Q, T1J, ovs); Chris@10: T1R = VFNMSI(T1G, T1F); Chris@10: STM2(&(xo[24]), T1R, ovs, &(xo[0])); Chris@10: T1S = VFMAI(T1G, T1F); Chris@10: STM2(&(xo[16]), T1S, ovs, &(xo[0])); Chris@10: TK = VFMA(LDK(KP559016994), TJ, TI); Chris@10: TU = VFNMS(LDK(KP559016994), TJ, TI); Chris@10: TR = VFNMS(LDK(KP559016994), Tk, Tj); Chris@10: Tl = VFMA(LDK(KP559016994), Tk, Tj); Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V TY, TW, TO, TQ, TB, TP, TX, TT; Chris@10: TY = VFMA(LDK(KP951056516), TV, TU); Chris@10: TW = VFNMS(LDK(KP951056516), TV, TU); Chris@10: TO = VFMA(LDK(KP951056516), TN, TK); Chris@10: TQ = VFNMS(LDK(KP951056516), TN, TK); Chris@10: TB = VFNMS(LDK(KP951056516), TA, Tl); Chris@10: TP = VFMA(LDK(KP951056516), TA, Tl); Chris@10: TX = VFNMS(LDK(KP951056516), TS, TR); Chris@10: TT = VFMA(LDK(KP951056516), TS, TR); Chris@10: { Chris@10: V T1T, T1U, T1V, T1W; Chris@10: T1T = VFMAI(TQ, TP); Chris@10: STM2(&(xo[18]), T1T, ovs, &(xo[2])); Chris@10: STN2(&(xo[16]), T1S, T1T, ovs); Chris@10: T1U = VFNMSI(TQ, TP); Chris@10: STM2(&(xo[22]), T1U, ovs, &(xo[2])); Chris@10: STN2(&(xo[20]), T1I, T1U, ovs); Chris@10: T1V = VFMAI(TO, TB); Chris@10: STM2(&(xo[2]), T1V, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), T1H, T1V, ovs); Chris@10: T1W = VFNMSI(TO, TB); Chris@10: STM2(&(xo[38]), T1W, ovs, &(xo[2])); Chris@10: STN2(&(xo[36]), T1N, T1W, ovs); Chris@10: { Chris@10: V T1X, T1Y, T1Z, T20; Chris@10: T1X = VFMAI(TW, TT); Chris@10: STM2(&(xo[34]), T1X, ovs, &(xo[2])); Chris@10: STN2(&(xo[32]), T1P, T1X, ovs); Chris@10: T1Y = VFNMSI(TW, TT); Chris@10: STM2(&(xo[6]), T1Y, ovs, &(xo[2])); Chris@10: STN2(&(xo[4]), T1O, T1Y, ovs); Chris@10: T1Z = VFMAI(TY, TX); Chris@10: STM2(&(xo[26]), T1Z, ovs, &(xo[2])); Chris@10: STN2(&(xo[24]), T1R, T1Z, ovs); Chris@10: T20 = VFNMSI(TY, TX); Chris@10: STM2(&(xo[14]), T20, ovs, &(xo[2])); Chris@10: STN2(&(xo[12]), T1M, T20, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), {58, 4, 46, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_20) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_20, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include n2b.h -store-multiple 2 */ Chris@10: Chris@10: /* Chris@10: * This function contains 104 FP additions, 24 FP multiplications, Chris@10: * (or, 92 additions, 12 multiplications, 12 fused multiply/add), Chris@10: * 57 stack variables, 4 constants, and 50 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) 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 i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) { Chris@10: V T3, T1y, TH, T1i, Ts, TL, TM, Tz, T13, T16, T1j, T1u, T1v, T1w, T1r; Chris@10: V T1s, T1t, T1a, T1d, T1k, Ti, Tk, TE, TI; Chris@10: { Chris@10: V T1, T2, T1g, TF, TG, T1h; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: T1g = VADD(T1, T2); Chris@10: TF = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: TG = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1h = VADD(TF, TG); Chris@10: T3 = VSUB(T1, T2); Chris@10: T1y = VADD(T1g, T1h); Chris@10: TH = VSUB(TF, TG); Chris@10: T1i = VSUB(T1g, T1h); Chris@10: } Chris@10: { Chris@10: V T6, T11, Tv, T19, Ty, T1c, T9, T14, Td, T18, To, T12, Tr, T15, Tg; Chris@10: V T1b; Chris@10: { Chris@10: V T4, T5, Tt, Tu; Chris@10: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@10: T6 = VSUB(T4, T5); Chris@10: T11 = VADD(T4, T5); Chris@10: Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tv = VSUB(Tt, Tu); Chris@10: T19 = VADD(Tt, Tu); Chris@10: } Chris@10: { Chris@10: V Tw, Tx, T7, T8; Chris@10: Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: Ty = VSUB(Tw, Tx); Chris@10: T1c = VADD(Tw, Tx); Chris@10: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: T9 = VSUB(T7, T8); Chris@10: T14 = VADD(T7, T8); Chris@10: } Chris@10: { Chris@10: V Tb, Tc, Tm, Tn; Chris@10: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@10: Td = VSUB(Tb, Tc); Chris@10: T18 = VADD(Tb, Tc); Chris@10: Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@10: To = VSUB(Tm, Tn); Chris@10: T12 = VADD(Tm, Tn); Chris@10: } Chris@10: { Chris@10: V Tp, Tq, Te, Tf; Chris@10: Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tr = VSUB(Tp, Tq); Chris@10: T15 = VADD(Tp, Tq); Chris@10: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: Tg = VSUB(Te, Tf); Chris@10: T1b = VADD(Te, Tf); Chris@10: } Chris@10: Ts = VSUB(To, Tr); Chris@10: TL = VSUB(T6, T9); Chris@10: TM = VSUB(Td, Tg); Chris@10: Tz = VSUB(Tv, Ty); Chris@10: T13 = VSUB(T11, T12); Chris@10: T16 = VSUB(T14, T15); Chris@10: T1j = VADD(T13, T16); Chris@10: T1u = VADD(T18, T19); Chris@10: T1v = VADD(T1b, T1c); Chris@10: T1w = VADD(T1u, T1v); Chris@10: T1r = VADD(T11, T12); Chris@10: T1s = VADD(T14, T15); Chris@10: T1t = VADD(T1r, T1s); Chris@10: T1a = VSUB(T18, T19); Chris@10: T1d = VSUB(T1b, T1c); Chris@10: T1k = VADD(T1a, T1d); Chris@10: { Chris@10: V Ta, Th, TC, TD; Chris@10: Ta = VADD(T6, T9); Chris@10: Th = VADD(Td, Tg); Chris@10: Ti = VADD(Ta, Th); Chris@10: Tk = VMUL(LDK(KP559016994), VSUB(Ta, Th)); Chris@10: TC = VADD(To, Tr); Chris@10: TD = VADD(Tv, Ty); Chris@10: TE = VMUL(LDK(KP559016994), VSUB(TC, TD)); Chris@10: TI = VADD(TC, TD); Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1H, T1J, T1K, T1L, T1N, T1I, TZ, T10; Chris@10: TZ = VADD(T3, Ti); Chris@10: T10 = VBYI(VADD(TH, TI)); Chris@10: T1H = VSUB(TZ, T10); Chris@10: STM2(&(xo[30]), T1H, ovs, &(xo[2])); Chris@10: T1I = VADD(TZ, T10); Chris@10: STM2(&(xo[10]), T1I, ovs, &(xo[2])); Chris@10: { Chris@10: V T1x, T1z, T1A, T1E, T1G, T1C, T1D, T1F, T1B, T1M; Chris@10: T1x = VMUL(LDK(KP559016994), VSUB(T1t, T1w)); Chris@10: T1z = VADD(T1t, T1w); Chris@10: T1A = VFNMS(LDK(KP250000000), T1z, T1y); Chris@10: T1C = VSUB(T1r, T1s); Chris@10: T1D = VSUB(T1u, T1v); Chris@10: T1E = VBYI(VFMA(LDK(KP951056516), T1C, VMUL(LDK(KP587785252), T1D))); Chris@10: T1G = VBYI(VFNMS(LDK(KP951056516), T1D, VMUL(LDK(KP587785252), T1C))); Chris@10: T1J = VADD(T1y, T1z); Chris@10: STM2(&(xo[0]), T1J, ovs, &(xo[0])); Chris@10: T1F = VSUB(T1A, T1x); Chris@10: T1K = VSUB(T1F, T1G); Chris@10: STM2(&(xo[16]), T1K, ovs, &(xo[0])); Chris@10: T1L = VADD(T1G, T1F); Chris@10: STM2(&(xo[24]), T1L, ovs, &(xo[0])); Chris@10: T1B = VADD(T1x, T1A); Chris@10: T1M = VSUB(T1B, T1E); Chris@10: STM2(&(xo[8]), T1M, ovs, &(xo[0])); Chris@10: STN2(&(xo[8]), T1M, T1I, ovs); Chris@10: T1N = VADD(T1E, T1B); Chris@10: STM2(&(xo[32]), T1N, ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V T1O, T1P, T1R, T1S; Chris@10: { Chris@10: V T1n, T1l, T1m, T1f, T1p, T17, T1e, T1q, T1Q, T1o; Chris@10: T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k)); Chris@10: T1l = VADD(T1j, T1k); Chris@10: T1m = VFNMS(LDK(KP250000000), T1l, T1i); Chris@10: T17 = VSUB(T13, T16); Chris@10: T1e = VSUB(T1a, T1d); Chris@10: T1f = VBYI(VFNMS(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17))); Chris@10: T1p = VBYI(VFMA(LDK(KP951056516), T17, VMUL(LDK(KP587785252), T1e))); Chris@10: T1O = VADD(T1i, T1l); Chris@10: STM2(&(xo[20]), T1O, ovs, &(xo[0])); Chris@10: T1q = VADD(T1n, T1m); Chris@10: T1P = VADD(T1p, T1q); Chris@10: STM2(&(xo[12]), T1P, ovs, &(xo[0])); Chris@10: T1Q = VSUB(T1q, T1p); Chris@10: STM2(&(xo[28]), T1Q, ovs, &(xo[0])); Chris@10: STN2(&(xo[28]), T1Q, T1H, ovs); Chris@10: T1o = VSUB(T1m, T1n); Chris@10: T1R = VADD(T1f, T1o); Chris@10: STM2(&(xo[4]), T1R, ovs, &(xo[0])); Chris@10: T1S = VSUB(T1o, T1f); Chris@10: STM2(&(xo[36]), T1S, ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V TA, TN, TU, TS, TK, TV, Tl, TR, TJ, Tj; Chris@10: TA = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts)); Chris@10: TN = VFNMS(LDK(KP951056516), TM, VMUL(LDK(KP587785252), TL)); Chris@10: TU = VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TM)); Chris@10: TS = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz)); Chris@10: TJ = VFNMS(LDK(KP250000000), TI, TH); Chris@10: TK = VSUB(TE, TJ); Chris@10: TV = VADD(TE, TJ); Chris@10: Tj = VFNMS(LDK(KP250000000), Ti, T3); Chris@10: Tl = VSUB(Tj, Tk); Chris@10: TR = VADD(Tk, Tj); Chris@10: { Chris@10: V TB, TO, T1T, T1U; Chris@10: TB = VSUB(Tl, TA); Chris@10: TO = VBYI(VSUB(TK, TN)); Chris@10: T1T = VSUB(TB, TO); Chris@10: STM2(&(xo[34]), T1T, ovs, &(xo[2])); Chris@10: STN2(&(xo[32]), T1N, T1T, ovs); Chris@10: T1U = VADD(TB, TO); Chris@10: STM2(&(xo[6]), T1U, ovs, &(xo[2])); Chris@10: STN2(&(xo[4]), T1R, T1U, ovs); Chris@10: } Chris@10: { Chris@10: V TX, TY, T1V, T1W; Chris@10: TX = VADD(TR, TS); Chris@10: TY = VBYI(VSUB(TV, TU)); Chris@10: T1V = VSUB(TX, TY); Chris@10: STM2(&(xo[22]), T1V, ovs, &(xo[2])); Chris@10: STN2(&(xo[20]), T1O, T1V, ovs); Chris@10: T1W = VADD(TX, TY); Chris@10: STM2(&(xo[18]), T1W, ovs, &(xo[2])); Chris@10: STN2(&(xo[16]), T1K, T1W, ovs); Chris@10: } Chris@10: { Chris@10: V TP, TQ, T1X, T1Y; Chris@10: TP = VADD(Tl, TA); Chris@10: TQ = VBYI(VADD(TN, TK)); Chris@10: T1X = VSUB(TP, TQ); Chris@10: STM2(&(xo[26]), T1X, ovs, &(xo[2])); Chris@10: STN2(&(xo[24]), T1L, T1X, ovs); Chris@10: T1Y = VADD(TP, TQ); Chris@10: STM2(&(xo[14]), T1Y, ovs, &(xo[2])); Chris@10: STN2(&(xo[12]), T1P, T1Y, ovs); Chris@10: } Chris@10: { Chris@10: V TT, TW, T1Z, T20; Chris@10: TT = VSUB(TR, TS); Chris@10: TW = VBYI(VADD(TU, TV)); Chris@10: T1Z = VSUB(TT, TW); Chris@10: STM2(&(xo[38]), T1Z, ovs, &(xo[2])); Chris@10: STN2(&(xo[36]), T1S, T1Z, ovs); Chris@10: T20 = VADD(TT, TW); Chris@10: STM2(&(xo[2]), T20, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), T1J, T20, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), {92, 12, 12, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_20) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_20, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */