Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:04:53 EDT 2018 */ Chris@82: Chris@82: #include "dft/codelet-dft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n1fv_20 -include dft/simd/n1f.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 104 FP additions, 50 FP multiplications, Chris@82: * (or, 58 additions, 4 multiplications, 46 fused multiply/add), Chris@82: * 53 stack variables, 4 constants, and 40 memory accesses Chris@82: */ Chris@82: #include "dft/simd/n1f.h" Chris@82: Chris@82: static void n1fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: { Chris@82: INT i; Chris@82: const R *xi; Chris@82: R *xo; Chris@82: xi = ri; Chris@82: xo = ro; Chris@82: 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@82: V T3, T1r, Tm, T13, TG, TN, TO, TH, T16, T19, T1a, T1v, T1w, T1x, T1s; Chris@82: V T1t, T1u, T1d, T1g, T1h, Ti, TE, TB, TL, Tj, TC; Chris@82: { Chris@82: V T1, T2, T11, Tk, Tl, T12; Chris@82: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@82: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@82: T11 = VADD(T1, T2); Chris@82: Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@82: T12 = VADD(Tk, Tl); Chris@82: T3 = VSUB(T1, T2); Chris@82: T1r = VADD(T11, T12); Chris@82: Tm = VSUB(Tk, Tl); Chris@82: T13 = VSUB(T11, T12); Chris@82: } Chris@82: { Chris@82: V T6, T14, Tw, T1c, Tz, T1f, T9, T17, Td, T1b, Tp, T15, Ts, T18, Tg; Chris@82: V T1e; Chris@82: { Chris@82: V T4, T5, Tu, Tv; Chris@82: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@82: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@82: T6 = VSUB(T4, T5); Chris@82: T14 = VADD(T4, T5); Chris@82: Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tw = VSUB(Tu, Tv); Chris@82: T1c = VADD(Tu, Tv); Chris@82: } Chris@82: { Chris@82: V Tx, Ty, T7, T8; Chris@82: Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@82: Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tz = VSUB(Tx, Ty); Chris@82: T1f = VADD(Tx, Ty); Chris@82: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@82: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@82: T9 = VSUB(T7, T8); Chris@82: T17 = VADD(T7, T8); Chris@82: } Chris@82: { Chris@82: V Tb, Tc, Tn, To; Chris@82: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@82: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@82: Td = VSUB(Tb, Tc); Chris@82: T1b = VADD(Tb, Tc); Chris@82: Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@82: To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tp = VSUB(Tn, To); Chris@82: T15 = VADD(Tn, To); Chris@82: } Chris@82: { Chris@82: V Tq, Tr, Te, Tf; Chris@82: Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@82: Ts = VSUB(Tq, Tr); Chris@82: T18 = VADD(Tq, Tr); Chris@82: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@82: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@82: Tg = VSUB(Te, Tf); Chris@82: T1e = VADD(Te, Tf); Chris@82: } Chris@82: TG = VSUB(Ts, Tp); Chris@82: TN = VSUB(T6, T9); Chris@82: TO = VSUB(Td, Tg); Chris@82: TH = VSUB(Tz, Tw); Chris@82: T16 = VSUB(T14, T15); Chris@82: T19 = VSUB(T17, T18); Chris@82: T1a = VADD(T16, T19); Chris@82: T1v = VADD(T1b, T1c); Chris@82: T1w = VADD(T1e, T1f); Chris@82: T1x = VADD(T1v, T1w); Chris@82: T1s = VADD(T14, T15); Chris@82: T1t = VADD(T17, T18); Chris@82: T1u = VADD(T1s, T1t); Chris@82: T1d = VSUB(T1b, T1c); Chris@82: T1g = VSUB(T1e, T1f); Chris@82: T1h = VADD(T1d, T1g); Chris@82: { Chris@82: V Ta, Th, Tt, TA; Chris@82: Ta = VADD(T6, T9); Chris@82: Th = VADD(Td, Tg); Chris@82: Ti = VADD(Ta, Th); Chris@82: TE = VSUB(Ta, Th); Chris@82: Tt = VADD(Tp, Ts); Chris@82: TA = VADD(Tw, Tz); Chris@82: TB = VADD(Tt, TA); Chris@82: TL = VSUB(TA, Tt); Chris@82: } Chris@82: } Chris@82: Tj = VADD(T3, Ti); Chris@82: TC = VADD(Tm, TB); Chris@82: ST(&(xo[WS(os, 5)]), VFNMSI(TC, Tj), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 15)]), VFMAI(TC, Tj), ovs, &(xo[WS(os, 1)])); Chris@82: { Chris@82: V T1A, T1y, T1z, T1E, T1G, T1C, T1D, T1F, T1B; Chris@82: T1A = VSUB(T1u, T1x); Chris@82: T1y = VADD(T1u, T1x); Chris@82: T1z = VFNMS(LDK(KP250000000), T1y, T1r); Chris@82: T1C = VSUB(T1s, T1t); Chris@82: T1D = VSUB(T1v, T1w); Chris@82: T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C)); Chris@82: T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D)); Chris@82: ST(&(xo[0]), VADD(T1r, T1y), ovs, &(xo[0])); Chris@82: T1F = VFNMS(LDK(KP559016994), T1A, T1z); Chris@82: ST(&(xo[WS(os, 8)]), VFNMSI(T1G, T1F), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 12)]), VFMAI(T1G, T1F), ovs, &(xo[0])); Chris@82: T1B = VFMA(LDK(KP559016994), T1A, T1z); Chris@82: ST(&(xo[WS(os, 4)]), VFMAI(T1E, T1B), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 16)]), VFNMSI(T1E, T1B), ovs, &(xo[0])); Chris@82: } Chris@82: { Chris@82: V T1k, T1i, T1j, T1o, T1q, T1m, T1n, T1p, T1l; Chris@82: T1k = VSUB(T1a, T1h); Chris@82: T1i = VADD(T1a, T1h); Chris@82: T1j = VFNMS(LDK(KP250000000), T1i, T13); Chris@82: T1m = VSUB(T1d, T1g); Chris@82: T1n = VSUB(T16, T19); Chris@82: T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m)); Chris@82: T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n)); Chris@82: ST(&(xo[WS(os, 10)]), VADD(T13, T1i), ovs, &(xo[0])); Chris@82: T1p = VFMA(LDK(KP559016994), T1k, T1j); Chris@82: ST(&(xo[WS(os, 6)]), VFNMSI(T1q, T1p), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 14)]), VFMAI(T1q, T1p), ovs, &(xo[0])); Chris@82: T1l = VFNMS(LDK(KP559016994), T1k, T1j); Chris@82: ST(&(xo[WS(os, 2)]), VFMAI(T1o, T1l), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 18)]), VFNMSI(T1o, T1l), ovs, &(xo[0])); Chris@82: } Chris@82: { Chris@82: V TI, TP, TX, TU, TM, TW, TF, TT, TK, TD; Chris@82: TI = VFMA(LDK(KP618033988), TH, TG); Chris@82: TP = VFMA(LDK(KP618033988), TO, TN); Chris@82: TX = VFNMS(LDK(KP618033988), TN, TO); Chris@82: TU = VFNMS(LDK(KP618033988), TG, TH); Chris@82: TK = VFNMS(LDK(KP250000000), TB, Tm); Chris@82: TM = VFNMS(LDK(KP559016994), TL, TK); Chris@82: TW = VFMA(LDK(KP559016994), TL, TK); Chris@82: TD = VFNMS(LDK(KP250000000), Ti, T3); Chris@82: TF = VFMA(LDK(KP559016994), TE, TD); Chris@82: TT = VFNMS(LDK(KP559016994), TE, TD); Chris@82: { Chris@82: V TJ, TQ, TZ, T10; Chris@82: TJ = VFMA(LDK(KP951056516), TI, TF); Chris@82: TQ = VFMA(LDK(KP951056516), TP, TM); Chris@82: ST(&(xo[WS(os, 1)]), VFNMSI(TQ, TJ), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 19)]), VFMAI(TQ, TJ), ovs, &(xo[WS(os, 1)])); Chris@82: TZ = VFMA(LDK(KP951056516), TU, TT); Chris@82: T10 = VFMA(LDK(KP951056516), TX, TW); Chris@82: ST(&(xo[WS(os, 13)]), VFNMSI(T10, TZ), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 7)]), VFMAI(T10, TZ), ovs, &(xo[WS(os, 1)])); Chris@82: } Chris@82: { Chris@82: V TR, TS, TV, TY; Chris@82: TR = VFNMS(LDK(KP951056516), TI, TF); Chris@82: TS = VFNMS(LDK(KP951056516), TP, TM); Chris@82: ST(&(xo[WS(os, 9)]), VFNMSI(TS, TR), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 11)]), VFMAI(TS, TR), ovs, &(xo[WS(os, 1)])); Chris@82: TV = VFNMS(LDK(KP951056516), TU, TT); Chris@82: TY = VFNMS(LDK(KP951056516), TX, TW); Chris@82: ST(&(xo[WS(os, 17)]), VFNMSI(TY, TV), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 3)]), VFMAI(TY, TV), ovs, &(xo[WS(os, 1)])); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 20, XSIMD_STRING("n1fv_20"), {58, 4, 46, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_n1fv_20) (planner *p) { Chris@82: X(kdft_register) (p, n1fv_20, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n1fv_20 -include dft/simd/n1f.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 104 FP additions, 24 FP multiplications, Chris@82: * (or, 92 additions, 12 multiplications, 12 fused multiply/add), Chris@82: * 53 stack variables, 4 constants, and 40 memory accesses Chris@82: */ Chris@82: #include "dft/simd/n1f.h" Chris@82: Chris@82: static void n1fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@82: { Chris@82: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: { Chris@82: INT i; Chris@82: const R *xi; Chris@82: R *xo; Chris@82: xi = ri; Chris@82: xo = ro; Chris@82: 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@82: V T3, T1B, Tm, T1i, TG, TN, TO, TH, T13, T16, T1k, T1u, T1v, T1z, T1r; Chris@82: V T1s, T1y, T1a, T1d, T1j, Ti, TD, TB, TL, Tj, TC; Chris@82: { Chris@82: V T1, T2, T1g, Tk, Tl, T1h; Chris@82: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@82: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@82: T1g = VADD(T1, T2); Chris@82: Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@82: T1h = VADD(Tk, Tl); Chris@82: T3 = VSUB(T1, T2); Chris@82: T1B = VADD(T1g, T1h); Chris@82: Tm = VSUB(Tk, Tl); Chris@82: T1i = VSUB(T1g, T1h); Chris@82: } Chris@82: { Chris@82: V T6, T18, Tw, T12, Tz, T15, T9, T1b, Td, T11, Tp, T19, Ts, T1c, Tg; Chris@82: V T14; Chris@82: { Chris@82: V T4, T5, Tu, Tv; Chris@82: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@82: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@82: T6 = VSUB(T4, T5); Chris@82: T18 = VADD(T4, T5); Chris@82: Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tw = VSUB(Tu, Tv); Chris@82: T12 = VADD(Tu, Tv); Chris@82: } Chris@82: { Chris@82: V Tx, Ty, T7, T8; Chris@82: Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@82: Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tz = VSUB(Tx, Ty); Chris@82: T15 = VADD(Tx, Ty); Chris@82: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@82: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@82: T9 = VSUB(T7, T8); Chris@82: T1b = VADD(T7, T8); Chris@82: } Chris@82: { Chris@82: V Tb, Tc, Tn, To; Chris@82: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@82: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@82: Td = VSUB(Tb, Tc); Chris@82: T11 = VADD(Tb, Tc); Chris@82: Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@82: To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tp = VSUB(Tn, To); Chris@82: T19 = VADD(Tn, To); Chris@82: } Chris@82: { Chris@82: V Tq, Tr, Te, Tf; Chris@82: Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@82: Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@82: Ts = VSUB(Tq, Tr); Chris@82: T1c = VADD(Tq, Tr); Chris@82: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@82: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@82: Tg = VSUB(Te, Tf); Chris@82: T14 = VADD(Te, Tf); Chris@82: } Chris@82: TG = VSUB(Ts, Tp); Chris@82: TN = VSUB(T6, T9); Chris@82: TO = VSUB(Td, Tg); Chris@82: TH = VSUB(Tz, Tw); Chris@82: T13 = VSUB(T11, T12); Chris@82: T16 = VSUB(T14, T15); Chris@82: T1k = VADD(T13, T16); Chris@82: T1u = VADD(T11, T12); Chris@82: T1v = VADD(T14, T15); Chris@82: T1z = VADD(T1u, T1v); Chris@82: T1r = VADD(T18, T19); Chris@82: T1s = VADD(T1b, T1c); Chris@82: T1y = VADD(T1r, T1s); Chris@82: T1a = VSUB(T18, T19); Chris@82: T1d = VSUB(T1b, T1c); Chris@82: T1j = VADD(T1a, T1d); Chris@82: { Chris@82: V Ta, Th, Tt, TA; Chris@82: Ta = VADD(T6, T9); Chris@82: Th = VADD(Td, Tg); Chris@82: Ti = VADD(Ta, Th); Chris@82: TD = VMUL(LDK(KP559016994), VSUB(Ta, Th)); Chris@82: Tt = VADD(Tp, Ts); Chris@82: TA = VADD(Tw, Tz); Chris@82: TB = VADD(Tt, TA); Chris@82: TL = VMUL(LDK(KP559016994), VSUB(TA, Tt)); Chris@82: } Chris@82: } Chris@82: Tj = VADD(T3, Ti); Chris@82: TC = VBYI(VADD(Tm, TB)); Chris@82: ST(&(xo[WS(os, 5)]), VSUB(Tj, TC), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 15)]), VADD(Tj, TC), ovs, &(xo[WS(os, 1)])); Chris@82: { Chris@82: V T1A, T1C, T1D, T1x, T1G, T1t, T1w, T1F, T1E; Chris@82: T1A = VMUL(LDK(KP559016994), VSUB(T1y, T1z)); Chris@82: T1C = VADD(T1y, T1z); Chris@82: T1D = VFNMS(LDK(KP250000000), T1C, T1B); Chris@82: T1t = VSUB(T1r, T1s); Chris@82: T1w = VSUB(T1u, T1v); Chris@82: T1x = VBYI(VFMA(LDK(KP951056516), T1t, VMUL(LDK(KP587785252), T1w))); Chris@82: T1G = VBYI(VFNMS(LDK(KP587785252), T1t, VMUL(LDK(KP951056516), T1w))); Chris@82: ST(&(xo[0]), VADD(T1B, T1C), ovs, &(xo[0])); Chris@82: T1F = VSUB(T1D, T1A); Chris@82: ST(&(xo[WS(os, 8)]), VSUB(T1F, T1G), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 12)]), VADD(T1G, T1F), ovs, &(xo[0])); Chris@82: T1E = VADD(T1A, T1D); Chris@82: ST(&(xo[WS(os, 4)]), VADD(T1x, T1E), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 16)]), VSUB(T1E, T1x), ovs, &(xo[0])); Chris@82: } Chris@82: { Chris@82: V T1n, T1l, T1m, T1f, T1q, T17, T1e, T1p, T1o; Chris@82: T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k)); Chris@82: T1l = VADD(T1j, T1k); Chris@82: T1m = VFNMS(LDK(KP250000000), T1l, T1i); Chris@82: T17 = VSUB(T13, T16); Chris@82: T1e = VSUB(T1a, T1d); Chris@82: T1f = VBYI(VFNMS(LDK(KP587785252), T1e, VMUL(LDK(KP951056516), T17))); Chris@82: T1q = VBYI(VFMA(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17))); Chris@82: ST(&(xo[WS(os, 10)]), VADD(T1i, T1l), ovs, &(xo[0])); Chris@82: T1p = VADD(T1n, T1m); Chris@82: ST(&(xo[WS(os, 6)]), VSUB(T1p, T1q), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 14)]), VADD(T1q, T1p), ovs, &(xo[0])); Chris@82: T1o = VSUB(T1m, T1n); Chris@82: ST(&(xo[WS(os, 2)]), VADD(T1f, T1o), ovs, &(xo[0])); Chris@82: ST(&(xo[WS(os, 18)]), VSUB(T1o, T1f), ovs, &(xo[0])); Chris@82: } Chris@82: { Chris@82: V TI, TP, TX, TU, TM, TW, TF, TT, TK, TE; Chris@82: TI = VFMA(LDK(KP951056516), TG, VMUL(LDK(KP587785252), TH)); Chris@82: TP = VFMA(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TO)); Chris@82: TX = VFNMS(LDK(KP587785252), TN, VMUL(LDK(KP951056516), TO)); Chris@82: TU = VFNMS(LDK(KP587785252), TG, VMUL(LDK(KP951056516), TH)); Chris@82: TK = VFMS(LDK(KP250000000), TB, Tm); Chris@82: TM = VADD(TK, TL); Chris@82: TW = VSUB(TL, TK); Chris@82: TE = VFNMS(LDK(KP250000000), Ti, T3); Chris@82: TF = VADD(TD, TE); Chris@82: TT = VSUB(TE, TD); Chris@82: { Chris@82: V TJ, TQ, TZ, T10; Chris@82: TJ = VADD(TF, TI); Chris@82: TQ = VBYI(VSUB(TM, TP)); Chris@82: ST(&(xo[WS(os, 19)]), VSUB(TJ, TQ), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 1)]), VADD(TJ, TQ), ovs, &(xo[WS(os, 1)])); Chris@82: TZ = VADD(TT, TU); Chris@82: T10 = VBYI(VADD(TX, TW)); Chris@82: ST(&(xo[WS(os, 13)]), VSUB(TZ, T10), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 7)]), VADD(TZ, T10), ovs, &(xo[WS(os, 1)])); Chris@82: } Chris@82: { Chris@82: V TR, TS, TV, TY; Chris@82: TR = VSUB(TF, TI); Chris@82: TS = VBYI(VADD(TP, TM)); Chris@82: ST(&(xo[WS(os, 11)]), VSUB(TR, TS), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 9)]), VADD(TR, TS), ovs, &(xo[WS(os, 1)])); Chris@82: TV = VSUB(TT, TU); Chris@82: TY = VBYI(VSUB(TW, TX)); Chris@82: ST(&(xo[WS(os, 17)]), VSUB(TV, TY), ovs, &(xo[WS(os, 1)])); Chris@82: ST(&(xo[WS(os, 3)]), VADD(TV, TY), ovs, &(xo[WS(os, 1)])); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const kdft_desc desc = { 20, XSIMD_STRING("n1fv_20"), {92, 12, 12, 0}, &GENUS, 0, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_n1fv_20) (planner *p) { Chris@82: X(kdft_register) (p, n1fv_20, &desc); Chris@82: } Chris@82: Chris@82: #endif