/* * Copyright (c) 2003, 2007-8 Matteo Frigo * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* This file was automatically generated --- DO NOT EDIT */ /* Generated on Sun Jul 12 06:40:36 EDT 2009 */ #include "codelet-dft.h" #ifdef HAVE_FMA /* Generated by: ../../../genfft/gen_notw_c -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 */ /* * This function contains 104 FP additions, 50 FP multiplications, * (or, 58 additions, 4 multiplications, 46 fused multiply/add), * 79 stack variables, 4 constants, and 50 memory accesses */ #include "n2b.h" static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) { DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP618033988, +0.618033988749894848204586834365638117720309180); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); INT i; const R *xi; R *xo; xi = ii; xo = io; for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) { V T1H, T1I, TS, TA, TN, TV, T1M, T1N, T1O, T1P, T1R, T1S, TK, TU, TR; V Tl; { V T3, TE, T1r, T13, Ta, TL, Tz, TG, Ts, TF, Th, TM, T1u, T1C, T1n; V T1a, T1m, T1h, T1x, T1D, Tk, Ti; { V T1, T2, TC, TD; T1 = LD(&(xi[0]), ivs, &(xi[0])); T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); TC = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); TD = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); { V T14, T6, T1c, Tv, Tm, T1f, Ty, T17, T9, Tn, Tp, T1b, Td, Tq, Te; V Tf, T15, To; { V Tw, Tx, T7, T8, Tb, Tc; { V T4, T5, Tt, Tu, T11, T12; T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); T3 = VSUB(T1, T2); T11 = VADD(T1, T2); TE = VSUB(TC, TD); T12 = VADD(TC, TD); T14 = VADD(T4, T5); T6 = VSUB(T4, T5); T1c = VADD(Tt, Tu); Tv = VSUB(Tt, Tu); Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); T1r = VADD(T11, T12); T13 = VSUB(T11, T12); } Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); T1f = VADD(Tw, Tx); Ty = VSUB(Tw, Tx); T17 = VADD(T7, T8); T9 = VSUB(T7, T8); Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); T1b = VADD(Tb, Tc); Td = VSUB(Tb, Tc); Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); } Ta = VADD(T6, T9); TL = VSUB(T6, T9); T15 = VADD(Tm, Tn); To = VSUB(Tm, Tn); Tz = VSUB(Tv, Ty); TG = VADD(Tv, Ty); { V T1d, T1v, T18, Tr, T1e, Tg, T16, T1s; T1d = VSUB(T1b, T1c); T1v = VADD(T1b, T1c); T18 = VADD(Tp, Tq); Tr = VSUB(Tp, Tq); T1e = VADD(Te, Tf); Tg = VSUB(Te, Tf); T16 = VSUB(T14, T15); T1s = VADD(T14, T15); { V T1t, T19, T1w, T1g; T1t = VADD(T17, T18); T19 = VSUB(T17, T18); Ts = VSUB(To, Tr); TF = VADD(To, Tr); T1w = VADD(T1e, T1f); T1g = VSUB(T1e, T1f); Th = VADD(Td, Tg); TM = VSUB(Td, Tg); T1u = VADD(T1s, T1t); T1C = VSUB(T1s, T1t); T1n = VSUB(T16, T19); T1a = VADD(T16, T19); T1m = VSUB(T1d, T1g); T1h = VADD(T1d, T1g); T1x = VADD(T1v, T1w); T1D = VSUB(T1v, T1w); } } } } Tk = VSUB(Ta, Th); Ti = VADD(Ta, Th); { V TJ, T1k, T1A, TZ, Tj, T1E, T1G, TI, T10, T1j, T1z, T1i, T1y, TH; TJ = VSUB(TF, TG); TH = VADD(TF, TG); T1i = VADD(T1a, T1h); T1k = VSUB(T1a, T1h); T1y = VADD(T1u, T1x); T1A = VSUB(T1u, T1x); TZ = VADD(T3, Ti); Tj = VFNMS(LDK(KP250000000), Ti, T3); T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C)); T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D)); TI = VFNMS(LDK(KP250000000), TH, TE); T10 = VADD(TE, TH); T1j = VFNMS(LDK(KP250000000), T1i, T13); T1H = VADD(T1r, T1y); STM2(&(xo[0]), T1H, ovs, &(xo[0])); T1z = VFNMS(LDK(KP250000000), T1y, T1r); T1I = VADD(T13, T1i); STM2(&(xo[20]), T1I, ovs, &(xo[0])); { V T1J, T1K, T1p, T1l, T1o, T1q, T1F, T1B, T1L, T1Q; TS = VFNMS(LDK(KP618033988), Ts, Tz); TA = VFMA(LDK(KP618033988), Tz, Ts); TN = VFMA(LDK(KP618033988), TM, TL); TV = VFNMS(LDK(KP618033988), TL, TM); T1J = VFMAI(T10, TZ); STM2(&(xo[10]), T1J, ovs, &(xo[2])); T1K = VFNMSI(T10, TZ); STM2(&(xo[30]), T1K, ovs, &(xo[2])); T1p = VFMA(LDK(KP559016994), T1k, T1j); T1l = VFNMS(LDK(KP559016994), T1k, T1j); T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m)); T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n)); T1F = VFNMS(LDK(KP559016994), T1A, T1z); T1B = VFMA(LDK(KP559016994), T1A, T1z); T1L = VFNMSI(T1q, T1p); STM2(&(xo[28]), T1L, ovs, &(xo[0])); STN2(&(xo[28]), T1L, T1K, ovs); T1M = VFMAI(T1q, T1p); STM2(&(xo[12]), T1M, ovs, &(xo[0])); T1N = VFMAI(T1o, T1l); STM2(&(xo[36]), T1N, ovs, &(xo[0])); T1O = VFNMSI(T1o, T1l); STM2(&(xo[4]), T1O, ovs, &(xo[0])); T1P = VFMAI(T1E, T1B); STM2(&(xo[32]), T1P, ovs, &(xo[0])); T1Q = VFNMSI(T1E, T1B); STM2(&(xo[8]), T1Q, ovs, &(xo[0])); STN2(&(xo[8]), T1Q, T1J, ovs); T1R = VFNMSI(T1G, T1F); STM2(&(xo[24]), T1R, ovs, &(xo[0])); T1S = VFMAI(T1G, T1F); STM2(&(xo[16]), T1S, ovs, &(xo[0])); TK = VFMA(LDK(KP559016994), TJ, TI); TU = VFNMS(LDK(KP559016994), TJ, TI); TR = VFNMS(LDK(KP559016994), Tk, Tj); Tl = VFMA(LDK(KP559016994), Tk, Tj); } } } { V TY, TW, TO, TQ, TB, TP, TX, TT; TY = VFMA(LDK(KP951056516), TV, TU); TW = VFNMS(LDK(KP951056516), TV, TU); TO = VFMA(LDK(KP951056516), TN, TK); TQ = VFNMS(LDK(KP951056516), TN, TK); TB = VFNMS(LDK(KP951056516), TA, Tl); TP = VFMA(LDK(KP951056516), TA, Tl); TX = VFNMS(LDK(KP951056516), TS, TR); TT = VFMA(LDK(KP951056516), TS, TR); { V T1T, T1U, T1V, T1W; T1T = VFMAI(TQ, TP); STM2(&(xo[18]), T1T, ovs, &(xo[2])); STN2(&(xo[16]), T1S, T1T, ovs); T1U = VFNMSI(TQ, TP); STM2(&(xo[22]), T1U, ovs, &(xo[2])); STN2(&(xo[20]), T1I, T1U, ovs); T1V = VFMAI(TO, TB); STM2(&(xo[2]), T1V, ovs, &(xo[2])); STN2(&(xo[0]), T1H, T1V, ovs); T1W = VFNMSI(TO, TB); STM2(&(xo[38]), T1W, ovs, &(xo[2])); STN2(&(xo[36]), T1N, T1W, ovs); { V T1X, T1Y, T1Z, T20; T1X = VFMAI(TW, TT); STM2(&(xo[34]), T1X, ovs, &(xo[2])); STN2(&(xo[32]), T1P, T1X, ovs); T1Y = VFNMSI(TW, TT); STM2(&(xo[6]), T1Y, ovs, &(xo[2])); STN2(&(xo[4]), T1O, T1Y, ovs); T1Z = VFMAI(TY, TX); STM2(&(xo[26]), T1Z, ovs, &(xo[2])); STN2(&(xo[24]), T1R, T1Z, ovs); T20 = VFNMSI(TY, TX); STM2(&(xo[14]), T20, ovs, &(xo[2])); STN2(&(xo[12]), T1M, T20, ovs); } } } } } static const kdft_desc desc = { 20, "n2bv_20", {58, 4, 46, 0}, &GENUS, 0, 2, 0, 0 }; void X(codelet_n2bv_20) (planner *p) { X(kdft_register) (p, n2bv_20, &desc); } #else /* HAVE_FMA */ /* Generated by: ../../../genfft/gen_notw_c -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include n2b.h -store-multiple 2 */ /* * This function contains 104 FP additions, 24 FP multiplications, * (or, 92 additions, 12 multiplications, 12 fused multiply/add), * 57 stack variables, 4 constants, and 50 memory accesses */ #include "n2b.h" static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) { DVK(KP587785252, +0.587785252292473129168705954639072768597652438); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); INT i; const R *xi; R *xo; xi = ii; xo = io; for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) { V T3, T1y, TH, T1i, Ts, TL, TM, Tz, T13, T16, T1j, T1u, T1v, T1w, T1r; V T1s, T1t, T1a, T1d, T1k, Ti, Tk, TE, TI; { V T1, T2, T1g, TF, TG, T1h; T1 = LD(&(xi[0]), ivs, &(xi[0])); T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); T1g = VADD(T1, T2); TF = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); TG = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); T1h = VADD(TF, TG); T3 = VSUB(T1, T2); T1y = VADD(T1g, T1h); TH = VSUB(TF, TG); T1i = VSUB(T1g, T1h); } { V T6, T11, Tv, T19, Ty, T1c, T9, T14, Td, T18, To, T12, Tr, T15, Tg; V T1b; { V T4, T5, Tt, Tu; T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); T6 = VSUB(T4, T5); T11 = VADD(T4, T5); Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Tv = VSUB(Tt, Tu); T19 = VADD(Tt, Tu); } { V Tw, Tx, T7, T8; Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Ty = VSUB(Tw, Tx); T1c = VADD(Tw, Tx); T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); T9 = VSUB(T7, T8); T14 = VADD(T7, T8); } { V Tb, Tc, Tm, Tn; Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Td = VSUB(Tb, Tc); T18 = VADD(Tb, Tc); Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); To = VSUB(Tm, Tn); T12 = VADD(Tm, Tn); } { V Tp, Tq, Te, Tf; Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Tr = VSUB(Tp, Tq); T15 = VADD(Tp, Tq); Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Tg = VSUB(Te, Tf); T1b = VADD(Te, Tf); } Ts = VSUB(To, Tr); TL = VSUB(T6, T9); TM = VSUB(Td, Tg); Tz = VSUB(Tv, Ty); T13 = VSUB(T11, T12); T16 = VSUB(T14, T15); T1j = VADD(T13, T16); T1u = VADD(T18, T19); T1v = VADD(T1b, T1c); T1w = VADD(T1u, T1v); T1r = VADD(T11, T12); T1s = VADD(T14, T15); T1t = VADD(T1r, T1s); T1a = VSUB(T18, T19); T1d = VSUB(T1b, T1c); T1k = VADD(T1a, T1d); { V Ta, Th, TC, TD; Ta = VADD(T6, T9); Th = VADD(Td, Tg); Ti = VADD(Ta, Th); Tk = VMUL(LDK(KP559016994), VSUB(Ta, Th)); TC = VADD(To, Tr); TD = VADD(Tv, Ty); TE = VMUL(LDK(KP559016994), VSUB(TC, TD)); TI = VADD(TC, TD); } } { V T1H, T1J, T1K, T1L, T1N, T1I, TZ, T10; TZ = VADD(T3, Ti); T10 = VBYI(VADD(TH, TI)); T1H = VSUB(TZ, T10); STM2(&(xo[30]), T1H, ovs, &(xo[2])); T1I = VADD(TZ, T10); STM2(&(xo[10]), T1I, ovs, &(xo[2])); { V T1x, T1z, T1A, T1E, T1G, T1C, T1D, T1F, T1B, T1M; T1x = VMUL(LDK(KP559016994), VSUB(T1t, T1w)); T1z = VADD(T1t, T1w); T1A = VFNMS(LDK(KP250000000), T1z, T1y); T1C = VSUB(T1r, T1s); T1D = VSUB(T1u, T1v); T1E = VBYI(VFMA(LDK(KP951056516), T1C, VMUL(LDK(KP587785252), T1D))); T1G = VBYI(VFNMS(LDK(KP951056516), T1D, VMUL(LDK(KP587785252), T1C))); T1J = VADD(T1y, T1z); STM2(&(xo[0]), T1J, ovs, &(xo[0])); T1F = VSUB(T1A, T1x); T1K = VSUB(T1F, T1G); STM2(&(xo[16]), T1K, ovs, &(xo[0])); T1L = VADD(T1G, T1F); STM2(&(xo[24]), T1L, ovs, &(xo[0])); T1B = VADD(T1x, T1A); T1M = VSUB(T1B, T1E); STM2(&(xo[8]), T1M, ovs, &(xo[0])); STN2(&(xo[8]), T1M, T1I, ovs); T1N = VADD(T1E, T1B); STM2(&(xo[32]), T1N, ovs, &(xo[0])); } { V T1O, T1P, T1R, T1S; { V T1n, T1l, T1m, T1f, T1p, T17, T1e, T1q, T1Q, T1o; T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k)); T1l = VADD(T1j, T1k); T1m = VFNMS(LDK(KP250000000), T1l, T1i); T17 = VSUB(T13, T16); T1e = VSUB(T1a, T1d); T1f = VBYI(VFNMS(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17))); T1p = VBYI(VFMA(LDK(KP951056516), T17, VMUL(LDK(KP587785252), T1e))); T1O = VADD(T1i, T1l); STM2(&(xo[20]), T1O, ovs, &(xo[0])); T1q = VADD(T1n, T1m); T1P = VADD(T1p, T1q); STM2(&(xo[12]), T1P, ovs, &(xo[0])); T1Q = VSUB(T1q, T1p); STM2(&(xo[28]), T1Q, ovs, &(xo[0])); STN2(&(xo[28]), T1Q, T1H, ovs); T1o = VSUB(T1m, T1n); T1R = VADD(T1f, T1o); STM2(&(xo[4]), T1R, ovs, &(xo[0])); T1S = VSUB(T1o, T1f); STM2(&(xo[36]), T1S, ovs, &(xo[0])); } { V TA, TN, TU, TS, TK, TV, Tl, TR, TJ, Tj; TA = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts)); TN = VFNMS(LDK(KP951056516), TM, VMUL(LDK(KP587785252), TL)); TU = VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TM)); TS = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz)); TJ = VFNMS(LDK(KP250000000), TI, TH); TK = VSUB(TE, TJ); TV = VADD(TE, TJ); Tj = VFNMS(LDK(KP250000000), Ti, T3); Tl = VSUB(Tj, Tk); TR = VADD(Tk, Tj); { V TB, TO, T1T, T1U; TB = VSUB(Tl, TA); TO = VBYI(VSUB(TK, TN)); T1T = VSUB(TB, TO); STM2(&(xo[34]), T1T, ovs, &(xo[2])); STN2(&(xo[32]), T1N, T1T, ovs); T1U = VADD(TB, TO); STM2(&(xo[6]), T1U, ovs, &(xo[2])); STN2(&(xo[4]), T1R, T1U, ovs); } { V TX, TY, T1V, T1W; TX = VADD(TR, TS); TY = VBYI(VSUB(TV, TU)); T1V = VSUB(TX, TY); STM2(&(xo[22]), T1V, ovs, &(xo[2])); STN2(&(xo[20]), T1O, T1V, ovs); T1W = VADD(TX, TY); STM2(&(xo[18]), T1W, ovs, &(xo[2])); STN2(&(xo[16]), T1K, T1W, ovs); } { V TP, TQ, T1X, T1Y; TP = VADD(Tl, TA); TQ = VBYI(VADD(TN, TK)); T1X = VSUB(TP, TQ); STM2(&(xo[26]), T1X, ovs, &(xo[2])); STN2(&(xo[24]), T1L, T1X, ovs); T1Y = VADD(TP, TQ); STM2(&(xo[14]), T1Y, ovs, &(xo[2])); STN2(&(xo[12]), T1P, T1Y, ovs); } { V TT, TW, T1Z, T20; TT = VSUB(TR, TS); TW = VBYI(VADD(TU, TV)); T1Z = VSUB(TT, TW); STM2(&(xo[38]), T1Z, ovs, &(xo[2])); STN2(&(xo[36]), T1S, T1Z, ovs); T20 = VADD(TT, TW); STM2(&(xo[2]), T20, ovs, &(xo[2])); STN2(&(xo[0]), T1J, T20, ovs); } } } } } } static const kdft_desc desc = { 20, "n2bv_20", {92, 12, 12, 0}, &GENUS, 0, 2, 0, 0 }; void X(codelet_n2bv_20) (planner *p) { X(kdft_register) (p, n2bv_20, &desc); } #endif /* HAVE_FMA */