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 07:37:04 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 32 -name n1bv_32 -include n1b.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 186 FP additions, 98 FP multiplications, Chris@10: * (or, 88 additions, 0 multiplications, 98 fused multiply/add), Chris@10: * 104 stack variables, 7 constants, and 64 memory accesses Chris@10: */ Chris@10: #include "n1b.h" Chris@10: Chris@10: static void n1bv_32(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(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@10: DVK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@10: DVK(KP198912367, +0.198912367379658006911597622644676228597850501); Chris@10: DVK(KP668178637, +0.668178637919298919997757686523080761552472251); Chris@10: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@10: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); Chris@10: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); 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(64, is), MAKE_VOLATILE_STRIDE(64, os)) { Chris@10: V T1h, Tr, T1a, T1k, TI, T1b, T1L, T1P, T1I, T1G, T1O, T1Q, T1H, T1z, T1c; Chris@10: V TZ; Chris@10: { Chris@10: V T2x, T1T, T2K, T1W, T1p, Tb, T1A, T16, Tu, TF, T2O, T2H, T2b, T2t, TY; Chris@10: V T1w, TT, T1v, T20, T2C, Tj, Te, T2e, To, T2i, T23, T2D, TB, TG, Th; Chris@10: V T2f, Tk; Chris@10: { Chris@10: V TL, TW, TP, TQ, T2F, T27, T28, TO; Chris@10: { Chris@10: V T1, T2, T12, T13, T4, T5, T7, T8; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@10: T12 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: T13 = LD(&(xi[WS(is, 24)]), ivs, &(xi[0])); Chris@10: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 20)]), ivs, &(xi[0])); Chris@10: T7 = LD(&(xi[WS(is, 28)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: { Chris@10: V TM, T25, T26, TN; Chris@10: { Chris@10: V TJ, T3, T14, T1U, T6, T1V, T9, TK, TU, TV, T1R, T1S, Ta, T15; Chris@10: TJ = LD(&(xi[WS(is, 31)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1R = VADD(T1, T2); Chris@10: T3 = VSUB(T1, T2); Chris@10: T1S = VADD(T12, T13); Chris@10: T14 = VSUB(T12, T13); Chris@10: T1U = VADD(T4, T5); Chris@10: T6 = VSUB(T4, T5); Chris@10: T1V = VADD(T7, T8); Chris@10: T9 = VSUB(T7, T8); Chris@10: TK = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@10: TU = LD(&(xi[WS(is, 23)]), ivs, &(xi[WS(is, 1)])); Chris@10: T2x = VSUB(T1R, T1S); Chris@10: T1T = VADD(T1R, T1S); Chris@10: TV = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: TM = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: T2K = VSUB(T1U, T1V); Chris@10: T1W = VADD(T1U, T1V); Chris@10: Ta = VADD(T6, T9); Chris@10: T15 = VSUB(T6, T9); Chris@10: T25 = VADD(TJ, TK); Chris@10: TL = VSUB(TJ, TK); Chris@10: T26 = VADD(TV, TU); Chris@10: TW = VSUB(TU, TV); Chris@10: TN = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@10: TP = LD(&(xi[WS(is, 27)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1p = VFNMS(LDK(KP707106781), Ta, T3); Chris@10: Tb = VFMA(LDK(KP707106781), Ta, T3); Chris@10: T1A = VFNMS(LDK(KP707106781), T15, T14); Chris@10: T16 = VFMA(LDK(KP707106781), T15, T14); Chris@10: TQ = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: } Chris@10: T2F = VSUB(T25, T26); Chris@10: T27 = VADD(T25, T26); Chris@10: T28 = VADD(TM, TN); Chris@10: TO = VSUB(TM, TN); Chris@10: } Chris@10: } Chris@10: { Chris@10: V Ty, T21, Tx, Tz, T1Y, T1Z; Chris@10: { Chris@10: V Ts, Tt, TD, T29, TR, TE, Tv, Tw; Chris@10: Ts = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tt = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@10: TD = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T29 = VADD(TP, TQ); Chris@10: TR = VSUB(TP, TQ); Chris@10: TE = LD(&(xi[WS(is, 25)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tv = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tw = LD(&(xi[WS(is, 21)]), ivs, &(xi[WS(is, 1)])); Chris@10: Ty = LD(&(xi[WS(is, 29)]), ivs, &(xi[WS(is, 1)])); Chris@10: T1Y = VADD(Ts, Tt); Chris@10: Tu = VSUB(Ts, Tt); Chris@10: { Chris@10: V T2G, T2a, TX, TS; Chris@10: T2G = VSUB(T29, T28); Chris@10: T2a = VADD(T28, T29); Chris@10: TX = VSUB(TR, TO); Chris@10: TS = VADD(TO, TR); Chris@10: T1Z = VADD(TD, TE); Chris@10: TF = VSUB(TD, TE); Chris@10: T21 = VADD(Tv, Tw); Chris@10: Tx = VSUB(Tv, Tw); Chris@10: T2O = VFMA(LDK(KP414213562), T2F, T2G); Chris@10: T2H = VFNMS(LDK(KP414213562), T2G, T2F); Chris@10: T2b = VSUB(T27, T2a); Chris@10: T2t = VADD(T27, T2a); Chris@10: TY = VFMA(LDK(KP707106781), TX, TW); Chris@10: T1w = VFNMS(LDK(KP707106781), TX, TW); Chris@10: TT = VFMA(LDK(KP707106781), TS, TL); Chris@10: T1v = VFNMS(LDK(KP707106781), TS, TL); Chris@10: Tz = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: } Chris@10: } Chris@10: T20 = VADD(T1Y, T1Z); Chris@10: T2C = VSUB(T1Y, T1Z); Chris@10: { Chris@10: V Tc, Td, Tm, Tn; Chris@10: Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: Td = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@10: Tm = LD(&(xi[WS(is, 22)]), ivs, &(xi[0])); Chris@10: Tn = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: { Chris@10: V Tf, TA, T22, Tg; Chris@10: Tf = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: TA = VSUB(Ty, Tz); Chris@10: T22 = VADD(Ty, Tz); Chris@10: Tg = LD(&(xi[WS(is, 26)]), ivs, &(xi[0])); Chris@10: Tj = LD(&(xi[WS(is, 30)]), ivs, &(xi[0])); Chris@10: Te = VSUB(Tc, Td); Chris@10: T2e = VADD(Tc, Td); Chris@10: To = VSUB(Tm, Tn); Chris@10: T2i = VADD(Tn, Tm); Chris@10: T23 = VADD(T21, T22); Chris@10: T2D = VSUB(T21, T22); Chris@10: TB = VADD(Tx, TA); Chris@10: TG = VSUB(Tx, TA); Chris@10: Th = VSUB(Tf, Tg); Chris@10: T2f = VADD(Tf, Tg); Chris@10: Tk = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1t, TH, T1s, TC, T2P, T2U, T2n, T2d, T2w, T2u, T1q, T19, T1B, Tq, T2W; Chris@10: V T2M, T2B, T2T, T2v, T2r, T2o, T2m, T2X, T2I; Chris@10: { Chris@10: V T1X, T2p, T2E, T2N, T2s, T2y, T2g, T17, Ti, T2h, Tl, T2c, T2l, T24; Chris@10: T1X = VSUB(T1T, T1W); Chris@10: T2p = VADD(T1T, T1W); Chris@10: T2E = VFNMS(LDK(KP414213562), T2D, T2C); Chris@10: T2N = VFMA(LDK(KP414213562), T2C, T2D); Chris@10: T2s = VADD(T20, T23); Chris@10: T24 = VSUB(T20, T23); Chris@10: T1t = VFNMS(LDK(KP707106781), TG, TF); Chris@10: TH = VFMA(LDK(KP707106781), TG, TF); Chris@10: T1s = VFNMS(LDK(KP707106781), TB, Tu); Chris@10: TC = VFMA(LDK(KP707106781), TB, Tu); Chris@10: T2y = VSUB(T2e, T2f); Chris@10: T2g = VADD(T2e, T2f); Chris@10: T17 = VFMA(LDK(KP414213562), Te, Th); Chris@10: Ti = VFNMS(LDK(KP414213562), Th, Te); Chris@10: T2h = VADD(Tj, Tk); Chris@10: Tl = VSUB(Tj, Tk); Chris@10: T2c = VADD(T24, T2b); Chris@10: T2l = VSUB(T24, T2b); Chris@10: { Chris@10: V T2L, T2A, T2q, T2k; Chris@10: T2P = VSUB(T2N, T2O); Chris@10: T2U = VADD(T2N, T2O); Chris@10: { Chris@10: V T2z, T2j, T18, Tp; Chris@10: T2z = VSUB(T2h, T2i); Chris@10: T2j = VADD(T2h, T2i); Chris@10: T18 = VFMA(LDK(KP414213562), Tl, To); Chris@10: Tp = VFNMS(LDK(KP414213562), To, Tl); Chris@10: T2n = VFMA(LDK(KP707106781), T2c, T1X); Chris@10: T2d = VFNMS(LDK(KP707106781), T2c, T1X); Chris@10: T2w = VADD(T2s, T2t); Chris@10: T2u = VSUB(T2s, T2t); Chris@10: T2L = VSUB(T2y, T2z); Chris@10: T2A = VADD(T2y, T2z); Chris@10: T2q = VADD(T2g, T2j); Chris@10: T2k = VSUB(T2g, T2j); Chris@10: T1q = VADD(T17, T18); Chris@10: T19 = VSUB(T17, T18); Chris@10: T1B = VSUB(Ti, Tp); Chris@10: Tq = VADD(Ti, Tp); Chris@10: } Chris@10: T2W = VFNMS(LDK(KP707106781), T2L, T2K); Chris@10: T2M = VFMA(LDK(KP707106781), T2L, T2K); Chris@10: T2B = VFMA(LDK(KP707106781), T2A, T2x); Chris@10: T2T = VFNMS(LDK(KP707106781), T2A, T2x); Chris@10: T2v = VADD(T2p, T2q); Chris@10: T2r = VSUB(T2p, T2q); Chris@10: T2o = VFMA(LDK(KP707106781), T2l, T2k); Chris@10: T2m = VFNMS(LDK(KP707106781), T2l, T2k); Chris@10: T2X = VSUB(T2E, T2H); Chris@10: T2I = VADD(T2E, T2H); Chris@10: } Chris@10: } Chris@10: { Chris@10: V T2V, T2Z, T2Y, T30, T2R, T2J; Chris@10: T2V = VFNMS(LDK(KP923879532), T2U, T2T); Chris@10: T2Z = VFMA(LDK(KP923879532), T2U, T2T); Chris@10: ST(&(xo[WS(os, 16)]), VSUB(T2v, T2w), ovs, &(xo[0])); Chris@10: ST(&(xo[0]), VADD(T2v, T2w), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 8)]), VFMAI(T2u, T2r), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 24)]), VFNMSI(T2u, T2r), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 4)]), VFMAI(T2o, T2n), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 28)]), VFNMSI(T2o, T2n), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 20)]), VFMAI(T2m, T2d), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 12)]), VFNMSI(T2m, T2d), ovs, &(xo[0])); Chris@10: T2Y = VFMA(LDK(KP923879532), T2X, T2W); Chris@10: T30 = VFNMS(LDK(KP923879532), T2X, T2W); Chris@10: T2R = VFMA(LDK(KP923879532), T2I, T2B); Chris@10: T2J = VFNMS(LDK(KP923879532), T2I, T2B); Chris@10: { Chris@10: V T1J, T1r, T1C, T1M, T2S, T2Q, T1u, T1D, T1E, T1x; Chris@10: T1J = VFNMS(LDK(KP923879532), T1q, T1p); Chris@10: T1r = VFMA(LDK(KP923879532), T1q, T1p); Chris@10: T1C = VFNMS(LDK(KP923879532), T1B, T1A); Chris@10: T1M = VFMA(LDK(KP923879532), T1B, T1A); Chris@10: ST(&(xo[WS(os, 6)]), VFNMSI(T30, T2Z), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 26)]), VFMAI(T30, T2Z), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 22)]), VFNMSI(T2Y, T2V), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 10)]), VFMAI(T2Y, T2V), ovs, &(xo[0])); Chris@10: T2S = VFMA(LDK(KP923879532), T2P, T2M); Chris@10: T2Q = VFNMS(LDK(KP923879532), T2P, T2M); Chris@10: T1u = VFMA(LDK(KP668178637), T1t, T1s); Chris@10: T1D = VFNMS(LDK(KP668178637), T1s, T1t); Chris@10: T1E = VFNMS(LDK(KP668178637), T1v, T1w); Chris@10: T1x = VFMA(LDK(KP668178637), T1w, T1v); Chris@10: { Chris@10: V T1K, T1F, T1N, T1y; Chris@10: T1h = VFNMS(LDK(KP923879532), Tq, Tb); Chris@10: Tr = VFMA(LDK(KP923879532), Tq, Tb); Chris@10: ST(&(xo[WS(os, 30)]), VFNMSI(T2S, T2R), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 2)]), VFMAI(T2S, T2R), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 18)]), VFMAI(T2Q, T2J), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 14)]), VFNMSI(T2Q, T2J), ovs, &(xo[0])); Chris@10: T1K = VADD(T1D, T1E); Chris@10: T1F = VSUB(T1D, T1E); Chris@10: T1N = VSUB(T1u, T1x); Chris@10: T1y = VADD(T1u, T1x); Chris@10: T1a = VFMA(LDK(KP923879532), T19, T16); Chris@10: T1k = VFNMS(LDK(KP923879532), T19, T16); Chris@10: TI = VFNMS(LDK(KP198912367), TH, TC); Chris@10: T1b = VFMA(LDK(KP198912367), TC, TH); Chris@10: T1L = VFMA(LDK(KP831469612), T1K, T1J); Chris@10: T1P = VFNMS(LDK(KP831469612), T1K, T1J); Chris@10: T1I = VFMA(LDK(KP831469612), T1F, T1C); Chris@10: T1G = VFNMS(LDK(KP831469612), T1F, T1C); Chris@10: T1O = VFNMS(LDK(KP831469612), T1N, T1M); Chris@10: T1Q = VFMA(LDK(KP831469612), T1N, T1M); Chris@10: T1H = VFMA(LDK(KP831469612), T1y, T1r); Chris@10: T1z = VFNMS(LDK(KP831469612), T1y, T1r); Chris@10: T1c = VFMA(LDK(KP198912367), TT, TY); Chris@10: TZ = VFNMS(LDK(KP198912367), TY, TT); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: V T1d, T1i, T10, T1l; Chris@10: ST(&(xo[WS(os, 21)]), VFMAI(T1O, T1L), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 11)]), VFNMSI(T1O, T1L), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 27)]), VFNMSI(T1Q, T1P), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 5)]), VFMAI(T1Q, T1P), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 29)]), VFMAI(T1I, T1H), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 3)]), VFNMSI(T1I, T1H), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 13)]), VFMAI(T1G, T1z), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 19)]), VFNMSI(T1G, T1z), ovs, &(xo[WS(os, 1)])); Chris@10: T1d = VSUB(T1b, T1c); Chris@10: T1i = VADD(T1b, T1c); Chris@10: T10 = VADD(TI, TZ); Chris@10: T1l = VSUB(TI, TZ); Chris@10: { Chris@10: V T1n, T1j, T1e, T1g, T1o, T1m, T11, T1f; Chris@10: T1n = VFMA(LDK(KP980785280), T1i, T1h); Chris@10: T1j = VFNMS(LDK(KP980785280), T1i, T1h); Chris@10: T1e = VFNMS(LDK(KP980785280), T1d, T1a); Chris@10: T1g = VFMA(LDK(KP980785280), T1d, T1a); Chris@10: T1o = VFNMS(LDK(KP980785280), T1l, T1k); Chris@10: T1m = VFMA(LDK(KP980785280), T1l, T1k); Chris@10: T11 = VFNMS(LDK(KP980785280), T10, Tr); Chris@10: T1f = VFMA(LDK(KP980785280), T10, Tr); Chris@10: ST(&(xo[WS(os, 23)]), VFNMSI(T1m, T1j), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 9)]), VFMAI(T1m, T1j), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 25)]), VFMAI(T1o, T1n), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 7)]), VFNMSI(T1o, T1n), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 1)]), VFMAI(T1g, T1f), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 31)]), VFNMSI(T1g, T1f), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 17)]), VFMAI(T1e, T11), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 15)]), VFNMSI(T1e, T11), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 32, XSIMD_STRING("n1bv_32"), {88, 0, 98, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1bv_32) (planner *p) { Chris@10: X(kdft_register) (p, n1bv_32, &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 32 -name n1bv_32 -include n1b.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 186 FP additions, 42 FP multiplications, Chris@10: * (or, 170 additions, 26 multiplications, 16 fused multiply/add), Chris@10: * 58 stack variables, 7 constants, and 64 memory accesses Chris@10: */ Chris@10: #include "n1b.h" Chris@10: Chris@10: static void n1bv_32(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(KP555570233, +0.555570233019602224742830813948532874374937191); Chris@10: DVK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@10: DVK(KP195090322, +0.195090322016128267848284868477022240927691618); Chris@10: DVK(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@10: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); Chris@10: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@10: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); 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(64, is), MAKE_VOLATILE_STRIDE(64, os)) { Chris@10: V T2f, T2k, T2N, T2M, T19, T1B, Tb, T1p, TT, T1v, TY, T1w, T2E, T2F, T2G; Chris@10: V T24, T2o, TC, T1s, TH, T1t, T2B, T2C, T2D, T1X, T2n, T2I, T2J, Tq, T1A; Chris@10: V T14, T1q, T2c, T2l; Chris@10: { Chris@10: V T3, T2i, T18, T2j, T6, T2d, T9, T2e, T15, Ta; Chris@10: { Chris@10: V T1, T2, T16, T17; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0])); Chris@10: T3 = VSUB(T1, T2); Chris@10: T2i = VADD(T1, T2); Chris@10: T16 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: T17 = LD(&(xi[WS(is, 24)]), ivs, &(xi[0])); Chris@10: T18 = VSUB(T16, T17); Chris@10: T2j = VADD(T16, T17); Chris@10: } Chris@10: { Chris@10: V T4, T5, T7, T8; Chris@10: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 20)]), ivs, &(xi[0])); Chris@10: T6 = VSUB(T4, T5); Chris@10: T2d = VADD(T4, T5); Chris@10: T7 = LD(&(xi[WS(is, 28)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0])); Chris@10: T9 = VSUB(T7, T8); Chris@10: T2e = VADD(T7, T8); Chris@10: } Chris@10: T2f = VSUB(T2d, T2e); Chris@10: T2k = VSUB(T2i, T2j); Chris@10: T2N = VADD(T2d, T2e); Chris@10: T2M = VADD(T2i, T2j); Chris@10: T15 = VMUL(LDK(KP707106781), VSUB(T6, T9)); Chris@10: T19 = VSUB(T15, T18); Chris@10: T1B = VADD(T18, T15); Chris@10: Ta = VMUL(LDK(KP707106781), VADD(T6, T9)); Chris@10: Tb = VSUB(T3, Ta); Chris@10: T1p = VADD(T3, Ta); Chris@10: } Chris@10: { Chris@10: V TL, T21, TW, T1Y, TO, T22, TS, T1Z; Chris@10: { Chris@10: V TJ, TK, TU, TV; Chris@10: TJ = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: TK = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)])); Chris@10: TL = VSUB(TJ, TK); Chris@10: T21 = VADD(TJ, TK); Chris@10: TU = LD(&(xi[WS(is, 31)]), ivs, &(xi[WS(is, 1)])); Chris@10: TV = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)])); Chris@10: TW = VSUB(TU, TV); Chris@10: T1Y = VADD(TU, TV); Chris@10: } Chris@10: { Chris@10: V TM, TN, TQ, TR; Chris@10: TM = LD(&(xi[WS(is, 27)]), ivs, &(xi[WS(is, 1)])); Chris@10: TN = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)])); Chris@10: TO = VSUB(TM, TN); Chris@10: T22 = VADD(TM, TN); Chris@10: TQ = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: TR = LD(&(xi[WS(is, 23)]), ivs, &(xi[WS(is, 1)])); Chris@10: TS = VSUB(TQ, TR); Chris@10: T1Z = VADD(TQ, TR); Chris@10: } Chris@10: { Chris@10: V TP, TX, T20, T23; Chris@10: TP = VMUL(LDK(KP707106781), VSUB(TL, TO)); Chris@10: TT = VSUB(TP, TS); Chris@10: T1v = VADD(TS, TP); Chris@10: TX = VMUL(LDK(KP707106781), VADD(TL, TO)); Chris@10: TY = VSUB(TW, TX); Chris@10: T1w = VADD(TW, TX); Chris@10: T2E = VADD(T1Y, T1Z); Chris@10: T2F = VADD(T21, T22); Chris@10: T2G = VSUB(T2E, T2F); Chris@10: T20 = VSUB(T1Y, T1Z); Chris@10: T23 = VSUB(T21, T22); Chris@10: T24 = VFMA(LDK(KP923879532), T20, VMUL(LDK(KP382683432), T23)); Chris@10: T2o = VFNMS(LDK(KP382683432), T20, VMUL(LDK(KP923879532), T23)); Chris@10: } Chris@10: } Chris@10: { Chris@10: V Tu, T1U, TF, T1R, Tx, T1V, TB, T1S; Chris@10: { Chris@10: V Ts, Tt, TD, TE; Chris@10: Ts = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tt = LD(&(xi[WS(is, 21)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tu = VSUB(Ts, Tt); Chris@10: T1U = VADD(Ts, Tt); Chris@10: TD = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: TE = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)])); Chris@10: TF = VSUB(TD, TE); Chris@10: T1R = VADD(TD, TE); Chris@10: } Chris@10: { Chris@10: V Tv, Tw, Tz, TA; Chris@10: Tv = LD(&(xi[WS(is, 29)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tw = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tx = VSUB(Tv, Tw); Chris@10: T1V = VADD(Tv, Tw); Chris@10: Tz = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: TA = LD(&(xi[WS(is, 25)]), ivs, &(xi[WS(is, 1)])); Chris@10: TB = VSUB(Tz, TA); Chris@10: T1S = VADD(Tz, TA); Chris@10: } Chris@10: { Chris@10: V Ty, TG, T1T, T1W; Chris@10: Ty = VMUL(LDK(KP707106781), VSUB(Tu, Tx)); Chris@10: TC = VSUB(Ty, TB); Chris@10: T1s = VADD(TB, Ty); Chris@10: TG = VMUL(LDK(KP707106781), VADD(Tu, Tx)); Chris@10: TH = VSUB(TF, TG); Chris@10: T1t = VADD(TF, TG); Chris@10: T2B = VADD(T1R, T1S); Chris@10: T2C = VADD(T1U, T1V); Chris@10: T2D = VSUB(T2B, T2C); Chris@10: T1T = VSUB(T1R, T1S); Chris@10: T1W = VSUB(T1U, T1V); Chris@10: T1X = VFNMS(LDK(KP382683432), T1W, VMUL(LDK(KP923879532), T1T)); Chris@10: T2n = VFMA(LDK(KP382683432), T1T, VMUL(LDK(KP923879532), T1W)); Chris@10: } Chris@10: } Chris@10: { Chris@10: V Te, T26, To, T29, Th, T27, Tl, T2a, Ti, Tp; Chris@10: { Chris@10: V Tc, Td, Tm, Tn; Chris@10: Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: Td = LD(&(xi[WS(is, 18)]), ivs, &(xi[0])); Chris@10: Te = VSUB(Tc, Td); Chris@10: T26 = VADD(Tc, Td); Chris@10: Tm = LD(&(xi[WS(is, 30)]), ivs, &(xi[0])); Chris@10: Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0])); Chris@10: To = VSUB(Tm, Tn); Chris@10: T29 = VADD(Tm, Tn); Chris@10: } Chris@10: { Chris@10: V Tf, Tg, Tj, Tk; Chris@10: Tf = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: Tg = LD(&(xi[WS(is, 26)]), ivs, &(xi[0])); Chris@10: Th = VSUB(Tf, Tg); Chris@10: T27 = VADD(Tf, Tg); Chris@10: Tj = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: Tk = LD(&(xi[WS(is, 22)]), ivs, &(xi[0])); Chris@10: Tl = VSUB(Tj, Tk); Chris@10: T2a = VADD(Tj, Tk); Chris@10: } Chris@10: T2I = VADD(T26, T27); Chris@10: T2J = VADD(T29, T2a); Chris@10: Ti = VFMA(LDK(KP382683432), Te, VMUL(LDK(KP923879532), Th)); Chris@10: Tp = VFNMS(LDK(KP382683432), To, VMUL(LDK(KP923879532), Tl)); Chris@10: Tq = VSUB(Ti, Tp); Chris@10: T1A = VADD(Ti, Tp); Chris@10: { Chris@10: V T12, T13, T28, T2b; Chris@10: T12 = VFNMS(LDK(KP382683432), Th, VMUL(LDK(KP923879532), Te)); Chris@10: T13 = VFMA(LDK(KP923879532), To, VMUL(LDK(KP382683432), Tl)); Chris@10: T14 = VSUB(T12, T13); Chris@10: T1q = VADD(T12, T13); Chris@10: T28 = VSUB(T26, T27); Chris@10: T2b = VSUB(T29, T2a); Chris@10: T2c = VMUL(LDK(KP707106781), VSUB(T28, T2b)); Chris@10: T2l = VMUL(LDK(KP707106781), VADD(T28, T2b)); Chris@10: } Chris@10: } Chris@10: { Chris@10: V T2L, T2R, T2Q, T2S; Chris@10: { Chris@10: V T2H, T2K, T2O, T2P; Chris@10: T2H = VMUL(LDK(KP707106781), VSUB(T2D, T2G)); Chris@10: T2K = VSUB(T2I, T2J); Chris@10: T2L = VBYI(VSUB(T2H, T2K)); Chris@10: T2R = VBYI(VADD(T2K, T2H)); Chris@10: T2O = VSUB(T2M, T2N); Chris@10: T2P = VMUL(LDK(KP707106781), VADD(T2D, T2G)); Chris@10: T2Q = VSUB(T2O, T2P); Chris@10: T2S = VADD(T2O, T2P); Chris@10: } Chris@10: ST(&(xo[WS(os, 12)]), VADD(T2L, T2Q), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 28)]), VSUB(T2S, T2R), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 20)]), VSUB(T2Q, T2L), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 4)]), VADD(T2R, T2S), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V T2h, T2r, T2q, T2s; Chris@10: { Chris@10: V T25, T2g, T2m, T2p; Chris@10: T25 = VSUB(T1X, T24); Chris@10: T2g = VSUB(T2c, T2f); Chris@10: T2h = VBYI(VSUB(T25, T2g)); Chris@10: T2r = VBYI(VADD(T2g, T25)); Chris@10: T2m = VSUB(T2k, T2l); Chris@10: T2p = VSUB(T2n, T2o); Chris@10: T2q = VSUB(T2m, T2p); Chris@10: T2s = VADD(T2m, T2p); Chris@10: } Chris@10: ST(&(xo[WS(os, 10)]), VADD(T2h, T2q), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 26)]), VSUB(T2s, T2r), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 22)]), VSUB(T2q, T2h), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 6)]), VADD(T2r, T2s), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V T2V, T2Z, T2Y, T30; Chris@10: { Chris@10: V T2T, T2U, T2W, T2X; Chris@10: T2T = VADD(T2M, T2N); Chris@10: T2U = VADD(T2I, T2J); Chris@10: T2V = VSUB(T2T, T2U); Chris@10: T2Z = VADD(T2T, T2U); Chris@10: T2W = VADD(T2B, T2C); Chris@10: T2X = VADD(T2E, T2F); Chris@10: T2Y = VBYI(VSUB(T2W, T2X)); Chris@10: T30 = VADD(T2W, T2X); Chris@10: } Chris@10: ST(&(xo[WS(os, 24)]), VSUB(T2V, T2Y), ovs, &(xo[0])); Chris@10: ST(&(xo[0]), VADD(T2Z, T30), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 8)]), VADD(T2V, T2Y), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 16)]), VSUB(T2Z, T30), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V T2v, T2z, T2y, T2A; Chris@10: { Chris@10: V T2t, T2u, T2w, T2x; Chris@10: T2t = VADD(T2k, T2l); Chris@10: T2u = VADD(T1X, T24); Chris@10: T2v = VADD(T2t, T2u); Chris@10: T2z = VSUB(T2t, T2u); Chris@10: T2w = VADD(T2f, T2c); Chris@10: T2x = VADD(T2n, T2o); Chris@10: T2y = VBYI(VADD(T2w, T2x)); Chris@10: T2A = VBYI(VSUB(T2x, T2w)); Chris@10: } Chris@10: ST(&(xo[WS(os, 30)]), VSUB(T2v, T2y), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 14)]), VADD(T2z, T2A), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 2)]), VADD(T2v, T2y), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 18)]), VSUB(T2z, T2A), ovs, &(xo[0])); Chris@10: } Chris@10: { Chris@10: V T1r, T1C, T1M, T1K, T1F, T1N, T1y, T1J; Chris@10: T1r = VSUB(T1p, T1q); Chris@10: T1C = VSUB(T1A, T1B); Chris@10: T1M = VADD(T1p, T1q); Chris@10: T1K = VADD(T1B, T1A); Chris@10: { Chris@10: V T1D, T1E, T1u, T1x; Chris@10: T1D = VFNMS(LDK(KP195090322), T1s, VMUL(LDK(KP980785280), T1t)); Chris@10: T1E = VFMA(LDK(KP195090322), T1v, VMUL(LDK(KP980785280), T1w)); Chris@10: T1F = VSUB(T1D, T1E); Chris@10: T1N = VADD(T1D, T1E); Chris@10: T1u = VFMA(LDK(KP980785280), T1s, VMUL(LDK(KP195090322), T1t)); Chris@10: T1x = VFNMS(LDK(KP195090322), T1w, VMUL(LDK(KP980785280), T1v)); Chris@10: T1y = VSUB(T1u, T1x); Chris@10: T1J = VADD(T1u, T1x); Chris@10: } Chris@10: { Chris@10: V T1z, T1G, T1P, T1Q; Chris@10: T1z = VADD(T1r, T1y); Chris@10: T1G = VBYI(VADD(T1C, T1F)); Chris@10: ST(&(xo[WS(os, 25)]), VSUB(T1z, T1G), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 7)]), VADD(T1z, T1G), ovs, &(xo[WS(os, 1)])); Chris@10: T1P = VBYI(VADD(T1K, T1J)); Chris@10: T1Q = VADD(T1M, T1N); Chris@10: ST(&(xo[WS(os, 1)]), VADD(T1P, T1Q), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 31)]), VSUB(T1Q, T1P), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: { Chris@10: V T1H, T1I, T1L, T1O; Chris@10: T1H = VSUB(T1r, T1y); Chris@10: T1I = VBYI(VSUB(T1F, T1C)); Chris@10: ST(&(xo[WS(os, 23)]), VSUB(T1H, T1I), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 9)]), VADD(T1H, T1I), ovs, &(xo[WS(os, 1)])); Chris@10: T1L = VBYI(VSUB(T1J, T1K)); Chris@10: T1O = VSUB(T1M, T1N); Chris@10: ST(&(xo[WS(os, 15)]), VADD(T1L, T1O), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 17)]), VSUB(T1O, T1L), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: { Chris@10: V Tr, T1a, T1k, T1i, T1d, T1l, T10, T1h; Chris@10: Tr = VSUB(Tb, Tq); Chris@10: T1a = VSUB(T14, T19); Chris@10: T1k = VADD(Tb, Tq); Chris@10: T1i = VADD(T19, T14); Chris@10: { Chris@10: V T1b, T1c, TI, TZ; Chris@10: T1b = VFNMS(LDK(KP555570233), TC, VMUL(LDK(KP831469612), TH)); Chris@10: T1c = VFMA(LDK(KP555570233), TT, VMUL(LDK(KP831469612), TY)); Chris@10: T1d = VSUB(T1b, T1c); Chris@10: T1l = VADD(T1b, T1c); Chris@10: TI = VFMA(LDK(KP831469612), TC, VMUL(LDK(KP555570233), TH)); Chris@10: TZ = VFNMS(LDK(KP555570233), TY, VMUL(LDK(KP831469612), TT)); Chris@10: T10 = VSUB(TI, TZ); Chris@10: T1h = VADD(TI, TZ); Chris@10: } Chris@10: { Chris@10: V T11, T1e, T1n, T1o; Chris@10: T11 = VADD(Tr, T10); Chris@10: T1e = VBYI(VADD(T1a, T1d)); Chris@10: ST(&(xo[WS(os, 27)]), VSUB(T11, T1e), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 5)]), VADD(T11, T1e), ovs, &(xo[WS(os, 1)])); Chris@10: T1n = VBYI(VADD(T1i, T1h)); Chris@10: T1o = VADD(T1k, T1l); Chris@10: ST(&(xo[WS(os, 3)]), VADD(T1n, T1o), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 29)]), VSUB(T1o, T1n), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: { Chris@10: V T1f, T1g, T1j, T1m; Chris@10: T1f = VSUB(Tr, T10); Chris@10: T1g = VBYI(VSUB(T1d, T1a)); Chris@10: ST(&(xo[WS(os, 21)]), VSUB(T1f, T1g), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 11)]), VADD(T1f, T1g), ovs, &(xo[WS(os, 1)])); Chris@10: T1j = VBYI(VSUB(T1h, T1i)); Chris@10: T1m = VSUB(T1k, T1l); Chris@10: ST(&(xo[WS(os, 13)]), VADD(T1j, T1m), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 19)]), VSUB(T1m, T1j), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 32, XSIMD_STRING("n1bv_32"), {170, 26, 16, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1bv_32) (planner *p) { Chris@10: X(kdft_register) (p, n1bv_32, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */