cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:05:11 EDT 2018 */ cannam@167: cannam@167: #include "dft/codelet-dft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n2fv_20 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */ cannam@167: cannam@167: /* cannam@167: * This function contains 104 FP additions, 50 FP multiplications, cannam@167: * (or, 58 additions, 4 multiplications, 46 fused multiply/add), cannam@167: * 57 stack variables, 4 constants, and 50 memory accesses cannam@167: */ cannam@167: #include "dft/simd/n2f.h" cannam@167: cannam@167: static void n2fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@167: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT i; cannam@167: const R *xi; cannam@167: R *xo; cannam@167: xi = ri; cannam@167: xo = ro; cannam@167: 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)) { cannam@167: V T3, T1r, Tm, T13, TG, TN, TO, TH, T16, T19, T1a, T1v, T1w, T1x, T1s; cannam@167: V T1t, T1u, T1d, T1g, T1h, Ti, TE, TB, TL; cannam@167: { cannam@167: V T1, T2, T11, Tk, Tl, T12; cannam@167: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@167: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); cannam@167: T11 = VADD(T1, T2); cannam@167: Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); cannam@167: T12 = VADD(Tk, Tl); cannam@167: T3 = VSUB(T1, T2); cannam@167: T1r = VADD(T11, T12); cannam@167: Tm = VSUB(Tk, Tl); cannam@167: T13 = VSUB(T11, T12); cannam@167: } cannam@167: { cannam@167: V T6, T14, Tw, T1c, Tz, T1f, T9, T17, Td, T1b, Tp, T15, Ts, T18, Tg; cannam@167: V T1e; cannam@167: { cannam@167: V T4, T5, Tu, Tv; cannam@167: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@167: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); cannam@167: T6 = VSUB(T4, T5); cannam@167: T14 = VADD(T4, T5); cannam@167: Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tw = VSUB(Tu, Tv); cannam@167: T1c = VADD(Tu, Tv); cannam@167: } cannam@167: { cannam@167: V Tx, Ty, T7, T8; cannam@167: Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); cannam@167: Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tz = VSUB(Tx, Ty); cannam@167: T1f = VADD(Tx, Ty); cannam@167: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); cannam@167: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@167: T9 = VSUB(T7, T8); cannam@167: T17 = VADD(T7, T8); cannam@167: } cannam@167: { cannam@167: V Tb, Tc, Tn, To; cannam@167: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@167: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); cannam@167: Td = VSUB(Tb, Tc); cannam@167: T1b = VADD(Tb, Tc); cannam@167: Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@167: To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tp = VSUB(Tn, To); cannam@167: T15 = VADD(Tn, To); cannam@167: } cannam@167: { cannam@167: V Tq, Tr, Te, Tf; cannam@167: Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); cannam@167: Ts = VSUB(Tq, Tr); cannam@167: T18 = VADD(Tq, Tr); cannam@167: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); cannam@167: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@167: Tg = VSUB(Te, Tf); cannam@167: T1e = VADD(Te, Tf); cannam@167: } cannam@167: TG = VSUB(Ts, Tp); cannam@167: TN = VSUB(T6, T9); cannam@167: TO = VSUB(Td, Tg); cannam@167: TH = VSUB(Tz, Tw); cannam@167: T16 = VSUB(T14, T15); cannam@167: T19 = VSUB(T17, T18); cannam@167: T1a = VADD(T16, T19); cannam@167: T1v = VADD(T1b, T1c); cannam@167: T1w = VADD(T1e, T1f); cannam@167: T1x = VADD(T1v, T1w); cannam@167: T1s = VADD(T14, T15); cannam@167: T1t = VADD(T17, T18); cannam@167: T1u = VADD(T1s, T1t); cannam@167: T1d = VSUB(T1b, T1c); cannam@167: T1g = VSUB(T1e, T1f); cannam@167: T1h = VADD(T1d, T1g); cannam@167: { cannam@167: V Ta, Th, Tt, TA; cannam@167: Ta = VADD(T6, T9); cannam@167: Th = VADD(Td, Tg); cannam@167: Ti = VADD(Ta, Th); cannam@167: TE = VSUB(Ta, Th); cannam@167: Tt = VADD(Tp, Ts); cannam@167: TA = VADD(Tw, Tz); cannam@167: TB = VADD(Tt, TA); cannam@167: TL = VSUB(TA, Tt); cannam@167: } cannam@167: } cannam@167: { cannam@167: V T1I, T1J, T1K, T1L, T1N, T1H, Tj, TC; cannam@167: Tj = VADD(T3, Ti); cannam@167: TC = VADD(Tm, TB); cannam@167: T1H = VFNMSI(TC, Tj); cannam@167: STM2(&(xo[10]), T1H, ovs, &(xo[2])); cannam@167: T1I = VFMAI(TC, Tj); cannam@167: STM2(&(xo[30]), T1I, ovs, &(xo[2])); cannam@167: { cannam@167: V T1A, T1y, T1z, T1E, T1G, T1C, T1D, T1F, T1B, T1M; cannam@167: T1A = VSUB(T1u, T1x); cannam@167: T1y = VADD(T1u, T1x); cannam@167: T1z = VFNMS(LDK(KP250000000), T1y, T1r); cannam@167: T1C = VSUB(T1s, T1t); cannam@167: T1D = VSUB(T1v, T1w); cannam@167: T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C)); cannam@167: T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D)); cannam@167: T1J = VADD(T1r, T1y); cannam@167: STM2(&(xo[0]), T1J, ovs, &(xo[0])); cannam@167: T1F = VFNMS(LDK(KP559016994), T1A, T1z); cannam@167: T1K = VFNMSI(T1G, T1F); cannam@167: STM2(&(xo[16]), T1K, ovs, &(xo[0])); cannam@167: T1L = VFMAI(T1G, T1F); cannam@167: STM2(&(xo[24]), T1L, ovs, &(xo[0])); cannam@167: T1B = VFMA(LDK(KP559016994), T1A, T1z); cannam@167: T1M = VFMAI(T1E, T1B); cannam@167: STM2(&(xo[8]), T1M, ovs, &(xo[0])); cannam@167: STN2(&(xo[8]), T1M, T1H, ovs); cannam@167: T1N = VFNMSI(T1E, T1B); cannam@167: STM2(&(xo[32]), T1N, ovs, &(xo[0])); cannam@167: } cannam@167: { cannam@167: V T1O, T1P, T1R, T1S; cannam@167: { cannam@167: V T1k, T1i, T1j, T1o, T1q, T1m, T1n, T1p, T1Q, T1l; cannam@167: T1k = VSUB(T1a, T1h); cannam@167: T1i = VADD(T1a, T1h); cannam@167: T1j = VFNMS(LDK(KP250000000), T1i, T13); cannam@167: T1m = VSUB(T1d, T1g); cannam@167: T1n = VSUB(T16, T19); cannam@167: T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m)); cannam@167: T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n)); cannam@167: T1O = VADD(T13, T1i); cannam@167: STM2(&(xo[20]), T1O, ovs, &(xo[0])); cannam@167: T1p = VFMA(LDK(KP559016994), T1k, T1j); cannam@167: T1P = VFNMSI(T1q, T1p); cannam@167: STM2(&(xo[12]), T1P, ovs, &(xo[0])); cannam@167: T1Q = VFMAI(T1q, T1p); cannam@167: STM2(&(xo[28]), T1Q, ovs, &(xo[0])); cannam@167: STN2(&(xo[28]), T1Q, T1I, ovs); cannam@167: T1l = VFNMS(LDK(KP559016994), T1k, T1j); cannam@167: T1R = VFMAI(T1o, T1l); cannam@167: STM2(&(xo[4]), T1R, ovs, &(xo[0])); cannam@167: T1S = VFNMSI(T1o, T1l); cannam@167: STM2(&(xo[36]), T1S, ovs, &(xo[0])); cannam@167: } cannam@167: { cannam@167: V TI, TP, TX, TU, TM, TW, TF, TT, TK, TD; cannam@167: TI = VFMA(LDK(KP618033988), TH, TG); cannam@167: TP = VFMA(LDK(KP618033988), TO, TN); cannam@167: TX = VFNMS(LDK(KP618033988), TN, TO); cannam@167: TU = VFNMS(LDK(KP618033988), TG, TH); cannam@167: TK = VFNMS(LDK(KP250000000), TB, Tm); cannam@167: TM = VFNMS(LDK(KP559016994), TL, TK); cannam@167: TW = VFMA(LDK(KP559016994), TL, TK); cannam@167: TD = VFNMS(LDK(KP250000000), Ti, T3); cannam@167: TF = VFMA(LDK(KP559016994), TE, TD); cannam@167: TT = VFNMS(LDK(KP559016994), TE, TD); cannam@167: { cannam@167: V TJ, TQ, T1T, T1U; cannam@167: TJ = VFMA(LDK(KP951056516), TI, TF); cannam@167: TQ = VFMA(LDK(KP951056516), TP, TM); cannam@167: T1T = VFNMSI(TQ, TJ); cannam@167: STM2(&(xo[2]), T1T, ovs, &(xo[2])); cannam@167: STN2(&(xo[0]), T1J, T1T, ovs); cannam@167: T1U = VFMAI(TQ, TJ); cannam@167: STM2(&(xo[38]), T1U, ovs, &(xo[2])); cannam@167: STN2(&(xo[36]), T1S, T1U, ovs); cannam@167: } cannam@167: { cannam@167: V TZ, T10, T1V, T1W; cannam@167: TZ = VFMA(LDK(KP951056516), TU, TT); cannam@167: T10 = VFMA(LDK(KP951056516), TX, TW); cannam@167: T1V = VFNMSI(T10, TZ); cannam@167: STM2(&(xo[26]), T1V, ovs, &(xo[2])); cannam@167: STN2(&(xo[24]), T1L, T1V, ovs); cannam@167: T1W = VFMAI(T10, TZ); cannam@167: STM2(&(xo[14]), T1W, ovs, &(xo[2])); cannam@167: STN2(&(xo[12]), T1P, T1W, ovs); cannam@167: } cannam@167: { cannam@167: V TR, TS, T1X, T1Y; cannam@167: TR = VFNMS(LDK(KP951056516), TI, TF); cannam@167: TS = VFNMS(LDK(KP951056516), TP, TM); cannam@167: T1X = VFNMSI(TS, TR); cannam@167: STM2(&(xo[18]), T1X, ovs, &(xo[2])); cannam@167: STN2(&(xo[16]), T1K, T1X, ovs); cannam@167: T1Y = VFMAI(TS, TR); cannam@167: STM2(&(xo[22]), T1Y, ovs, &(xo[2])); cannam@167: STN2(&(xo[20]), T1O, T1Y, ovs); cannam@167: } cannam@167: { cannam@167: V TV, TY, T1Z, T20; cannam@167: TV = VFNMS(LDK(KP951056516), TU, TT); cannam@167: TY = VFNMS(LDK(KP951056516), TX, TW); cannam@167: T1Z = VFNMSI(TY, TV); cannam@167: STM2(&(xo[34]), T1Z, ovs, &(xo[2])); cannam@167: STN2(&(xo[32]), T1N, T1Z, ovs); cannam@167: T20 = VFMAI(TY, TV); cannam@167: STM2(&(xo[6]), T20, ovs, &(xo[2])); cannam@167: STN2(&(xo[4]), T1R, T20, ovs); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 20, XSIMD_STRING("n2fv_20"), {58, 4, 46, 0}, &GENUS, 0, 2, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_n2fv_20) (planner *p) { cannam@167: X(kdft_register) (p, n2fv_20, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n2fv_20 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */ cannam@167: cannam@167: /* cannam@167: * This function contains 104 FP additions, 24 FP multiplications, cannam@167: * (or, 92 additions, 12 multiplications, 12 fused multiply/add), cannam@167: * 57 stack variables, 4 constants, and 50 memory accesses cannam@167: */ cannam@167: #include "dft/simd/n2f.h" cannam@167: cannam@167: static void n2fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@167: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: { cannam@167: INT i; cannam@167: const R *xi; cannam@167: R *xo; cannam@167: xi = ri; cannam@167: xo = ro; cannam@167: 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)) { cannam@167: V T3, T1B, Tm, T1i, TG, TN, TO, TH, T13, T16, T1k, T1u, T1v, T1z, T1r; cannam@167: V T1s, T1y, T1a, T1d, T1j, Ti, TD, TB, TL; cannam@167: { cannam@167: V T1, T2, T1g, Tk, Tl, T1h; cannam@167: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@167: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); cannam@167: T1g = VADD(T1, T2); cannam@167: Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); cannam@167: T1h = VADD(Tk, Tl); cannam@167: T3 = VSUB(T1, T2); cannam@167: T1B = VADD(T1g, T1h); cannam@167: Tm = VSUB(Tk, Tl); cannam@167: T1i = VSUB(T1g, T1h); cannam@167: } cannam@167: { cannam@167: V T6, T18, Tw, T12, Tz, T15, T9, T1b, Td, T11, Tp, T19, Ts, T1c, Tg; cannam@167: V T14; cannam@167: { cannam@167: V T4, T5, Tu, Tv; cannam@167: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@167: T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); cannam@167: T6 = VSUB(T4, T5); cannam@167: T18 = VADD(T4, T5); cannam@167: Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tw = VSUB(Tu, Tv); cannam@167: T12 = VADD(Tu, Tv); cannam@167: } cannam@167: { cannam@167: V Tx, Ty, T7, T8; cannam@167: Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); cannam@167: Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tz = VSUB(Tx, Ty); cannam@167: T15 = VADD(Tx, Ty); cannam@167: T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); cannam@167: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@167: T9 = VSUB(T7, T8); cannam@167: T1b = VADD(T7, T8); cannam@167: } cannam@167: { cannam@167: V Tb, Tc, Tn, To; cannam@167: Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@167: Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); cannam@167: Td = VSUB(Tb, Tc); cannam@167: T11 = VADD(Tb, Tc); cannam@167: Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@167: To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tp = VSUB(Tn, To); cannam@167: T19 = VADD(Tn, To); cannam@167: } cannam@167: { cannam@167: V Tq, Tr, Te, Tf; cannam@167: Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@167: Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); cannam@167: Ts = VSUB(Tq, Tr); cannam@167: T1c = VADD(Tq, Tr); cannam@167: Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); cannam@167: Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@167: Tg = VSUB(Te, Tf); cannam@167: T14 = VADD(Te, Tf); cannam@167: } cannam@167: TG = VSUB(Ts, Tp); cannam@167: TN = VSUB(T6, T9); cannam@167: TO = VSUB(Td, Tg); cannam@167: TH = VSUB(Tz, Tw); cannam@167: T13 = VSUB(T11, T12); cannam@167: T16 = VSUB(T14, T15); cannam@167: T1k = VADD(T13, T16); cannam@167: T1u = VADD(T11, T12); cannam@167: T1v = VADD(T14, T15); cannam@167: T1z = VADD(T1u, T1v); cannam@167: T1r = VADD(T18, T19); cannam@167: T1s = VADD(T1b, T1c); cannam@167: T1y = VADD(T1r, T1s); cannam@167: T1a = VSUB(T18, T19); cannam@167: T1d = VSUB(T1b, T1c); cannam@167: T1j = VADD(T1a, T1d); cannam@167: { cannam@167: V Ta, Th, Tt, TA; cannam@167: Ta = VADD(T6, T9); cannam@167: Th = VADD(Td, Tg); cannam@167: Ti = VADD(Ta, Th); cannam@167: TD = VMUL(LDK(KP559016994), VSUB(Ta, Th)); cannam@167: Tt = VADD(Tp, Ts); cannam@167: TA = VADD(Tw, Tz); cannam@167: TB = VADD(Tt, TA); cannam@167: TL = VMUL(LDK(KP559016994), VSUB(TA, Tt)); cannam@167: } cannam@167: } cannam@167: { cannam@167: V T1I, T1J, T1K, T1L, T1N, T1H, Tj, TC; cannam@167: Tj = VADD(T3, Ti); cannam@167: TC = VBYI(VADD(Tm, TB)); cannam@167: T1H = VSUB(Tj, TC); cannam@167: STM2(&(xo[10]), T1H, ovs, &(xo[2])); cannam@167: T1I = VADD(Tj, TC); cannam@167: STM2(&(xo[30]), T1I, ovs, &(xo[2])); cannam@167: { cannam@167: V T1A, T1C, T1D, T1x, T1G, T1t, T1w, T1F, T1E, T1M; cannam@167: T1A = VMUL(LDK(KP559016994), VSUB(T1y, T1z)); cannam@167: T1C = VADD(T1y, T1z); cannam@167: T1D = VFNMS(LDK(KP250000000), T1C, T1B); cannam@167: T1t = VSUB(T1r, T1s); cannam@167: T1w = VSUB(T1u, T1v); cannam@167: T1x = VBYI(VFMA(LDK(KP951056516), T1t, VMUL(LDK(KP587785252), T1w))); cannam@167: T1G = VBYI(VFNMS(LDK(KP587785252), T1t, VMUL(LDK(KP951056516), T1w))); cannam@167: T1J = VADD(T1B, T1C); cannam@167: STM2(&(xo[0]), T1J, ovs, &(xo[0])); cannam@167: T1F = VSUB(T1D, T1A); cannam@167: T1K = VSUB(T1F, T1G); cannam@167: STM2(&(xo[16]), T1K, ovs, &(xo[0])); cannam@167: T1L = VADD(T1G, T1F); cannam@167: STM2(&(xo[24]), T1L, ovs, &(xo[0])); cannam@167: T1E = VADD(T1A, T1D); cannam@167: T1M = VADD(T1x, T1E); cannam@167: STM2(&(xo[8]), T1M, ovs, &(xo[0])); cannam@167: STN2(&(xo[8]), T1M, T1H, ovs); cannam@167: T1N = VSUB(T1E, T1x); cannam@167: STM2(&(xo[32]), T1N, ovs, &(xo[0])); cannam@167: } cannam@167: { cannam@167: V T1O, T1P, T1R, T1S; cannam@167: { cannam@167: V T1n, T1l, T1m, T1f, T1q, T17, T1e, T1p, T1Q, T1o; cannam@167: T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k)); cannam@167: T1l = VADD(T1j, T1k); cannam@167: T1m = VFNMS(LDK(KP250000000), T1l, T1i); cannam@167: T17 = VSUB(T13, T16); cannam@167: T1e = VSUB(T1a, T1d); cannam@167: T1f = VBYI(VFNMS(LDK(KP587785252), T1e, VMUL(LDK(KP951056516), T17))); cannam@167: T1q = VBYI(VFMA(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17))); cannam@167: T1O = VADD(T1i, T1l); cannam@167: STM2(&(xo[20]), T1O, ovs, &(xo[0])); cannam@167: T1p = VADD(T1n, T1m); cannam@167: T1P = VSUB(T1p, T1q); cannam@167: STM2(&(xo[12]), T1P, ovs, &(xo[0])); cannam@167: T1Q = VADD(T1q, T1p); cannam@167: STM2(&(xo[28]), T1Q, ovs, &(xo[0])); cannam@167: STN2(&(xo[28]), T1Q, T1I, ovs); cannam@167: T1o = VSUB(T1m, T1n); cannam@167: T1R = VADD(T1f, T1o); cannam@167: STM2(&(xo[4]), T1R, ovs, &(xo[0])); cannam@167: T1S = VSUB(T1o, T1f); cannam@167: STM2(&(xo[36]), T1S, ovs, &(xo[0])); cannam@167: } cannam@167: { cannam@167: V TI, TP, TX, TU, TM, TW, TF, TT, TK, TE; cannam@167: TI = VFMA(LDK(KP951056516), TG, VMUL(LDK(KP587785252), TH)); cannam@167: TP = VFMA(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TO)); cannam@167: TX = VFNMS(LDK(KP587785252), TN, VMUL(LDK(KP951056516), TO)); cannam@167: TU = VFNMS(LDK(KP587785252), TG, VMUL(LDK(KP951056516), TH)); cannam@167: TK = VFMS(LDK(KP250000000), TB, Tm); cannam@167: TM = VADD(TK, TL); cannam@167: TW = VSUB(TL, TK); cannam@167: TE = VFNMS(LDK(KP250000000), Ti, T3); cannam@167: TF = VADD(TD, TE); cannam@167: TT = VSUB(TE, TD); cannam@167: { cannam@167: V TJ, TQ, T1T, T1U; cannam@167: TJ = VADD(TF, TI); cannam@167: TQ = VBYI(VSUB(TM, TP)); cannam@167: T1T = VSUB(TJ, TQ); cannam@167: STM2(&(xo[38]), T1T, ovs, &(xo[2])); cannam@167: STN2(&(xo[36]), T1S, T1T, ovs); cannam@167: T1U = VADD(TJ, TQ); cannam@167: STM2(&(xo[2]), T1U, ovs, &(xo[2])); cannam@167: STN2(&(xo[0]), T1J, T1U, ovs); cannam@167: } cannam@167: { cannam@167: V TZ, T10, T1V, T1W; cannam@167: TZ = VADD(TT, TU); cannam@167: T10 = VBYI(VADD(TX, TW)); cannam@167: T1V = VSUB(TZ, T10); cannam@167: STM2(&(xo[26]), T1V, ovs, &(xo[2])); cannam@167: STN2(&(xo[24]), T1L, T1V, ovs); cannam@167: T1W = VADD(TZ, T10); cannam@167: STM2(&(xo[14]), T1W, ovs, &(xo[2])); cannam@167: STN2(&(xo[12]), T1P, T1W, ovs); cannam@167: } cannam@167: { cannam@167: V TR, TS, T1X, T1Y; cannam@167: TR = VSUB(TF, TI); cannam@167: TS = VBYI(VADD(TP, TM)); cannam@167: T1X = VSUB(TR, TS); cannam@167: STM2(&(xo[22]), T1X, ovs, &(xo[2])); cannam@167: STN2(&(xo[20]), T1O, T1X, ovs); cannam@167: T1Y = VADD(TR, TS); cannam@167: STM2(&(xo[18]), T1Y, ovs, &(xo[2])); cannam@167: STN2(&(xo[16]), T1K, T1Y, ovs); cannam@167: } cannam@167: { cannam@167: V TV, TY, T1Z, T20; cannam@167: TV = VSUB(TT, TU); cannam@167: TY = VBYI(VSUB(TW, TX)); cannam@167: T1Z = VSUB(TV, TY); cannam@167: STM2(&(xo[34]), T1Z, ovs, &(xo[2])); cannam@167: STN2(&(xo[32]), T1N, T1Z, ovs); cannam@167: T20 = VADD(TV, TY); cannam@167: STM2(&(xo[6]), T20, ovs, &(xo[2])); cannam@167: STN2(&(xo[4]), T1R, T20, ovs); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 20, XSIMD_STRING("n2fv_20"), {92, 12, 12, 0}, &GENUS, 0, 2, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_n2fv_20) (planner *p) { cannam@167: X(kdft_register) (p, n2fv_20, &desc); cannam@167: } cannam@167: cannam@167: #endif