cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:45:06 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1sv_16 -include ts.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 174 FP additions, 100 FP multiplications, cannam@127: * (or, 104 additions, 30 multiplications, 70 fused multiply/add), cannam@127: * 113 stack variables, 3 constants, and 64 memory accesses cannam@127: */ cannam@127: #include "ts.h" cannam@127: cannam@127: static void t1sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@127: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT m; cannam@127: for (m = mb, W = W + (mb * 30); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 30), MAKE_VOLATILE_STRIDE(32, rs)) { cannam@127: V T2S, T2O, T2B, T2j, T2A, T24, T3J, T3L, T2Q, T2I, T2R, T2L, T2C, T2y, T3D; cannam@127: V T3F; cannam@127: { cannam@127: V T3o, T3z, T1I, T8, T35, T2o, T1s, T2r, T36, T2w, T1F, T2p, T1N, T3k, Tl; cannam@127: V T3A, T2V, T1T, Tz, T1U, T30, T29, T11, T2c, TH, TK, TJ, T31, T2h, T1e; cannam@127: V T2a, T1Z, TI, T1Y, TF; cannam@127: { cannam@127: V Ta, Td, Tg, Tj, T2t, T1y, Tf, T1J, Tb, Tc, T2v, T1E, Ti; cannam@127: { cannam@127: V T1, T3n, T3, T6, T5, T1h, T1k, T1n, T1q, T1m, T3l, T4, T1j, T1p, T2k; cannam@127: V T1i, T2, T1g; cannam@127: T1 = LD(&(ri[0]), ms, &(ri[0])); cannam@127: T3n = LD(&(ii[0]), ms, &(ii[0])); cannam@127: T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0])); cannam@127: T6 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0])); cannam@127: T2 = LDW(&(W[TWVL * 14])); cannam@127: T5 = LDW(&(W[TWVL * 15])); cannam@127: T1h = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1k = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1g = LDW(&(W[TWVL * 28])); cannam@127: T1n = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1q = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1m = LDW(&(W[TWVL * 12])); cannam@127: T3l = VMUL(T2, T6); cannam@127: T4 = VMUL(T2, T3); cannam@127: T1j = LDW(&(W[TWVL * 29])); cannam@127: T1p = LDW(&(W[TWVL * 13])); cannam@127: T2k = VMUL(T1g, T1k); cannam@127: T1i = VMUL(T1g, T1h); cannam@127: { cannam@127: V T1u, T1x, T1A, T2s, T1v, T1D, T1z, T1w, T1C, T2u, T1B, T9; cannam@127: { cannam@127: V T2l, T1l, T1t, T2n, T1r; cannam@127: { cannam@127: V T2m, T1o, T3m, T7; cannam@127: T1u = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)])); cannam@127: T2m = VMUL(T1m, T1q); cannam@127: T1o = VMUL(T1m, T1n); cannam@127: T3m = VFNMS(T5, T3, T3l); cannam@127: T7 = VFMA(T5, T6, T4); cannam@127: T1x = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)])); cannam@127: T2l = VFNMS(T1j, T1h, T2k); cannam@127: T1l = VFMA(T1j, T1k, T1i); cannam@127: T1t = LDW(&(W[TWVL * 4])); cannam@127: T2n = VFNMS(T1p, T1n, T2m); cannam@127: T1r = VFMA(T1p, T1q, T1o); cannam@127: T3o = VADD(T3m, T3n); cannam@127: T3z = VSUB(T3n, T3m); cannam@127: T1I = VSUB(T1, T7); cannam@127: T8 = VADD(T1, T7); cannam@127: } cannam@127: T1A = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)])); cannam@127: T2s = VMUL(T1t, T1x); cannam@127: T1v = VMUL(T1t, T1u); cannam@127: T35 = VADD(T2l, T2n); cannam@127: T2o = VSUB(T2l, T2n); cannam@127: T1s = VADD(T1l, T1r); cannam@127: T2r = VSUB(T1l, T1r); cannam@127: T1D = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1z = LDW(&(W[TWVL * 20])); cannam@127: } cannam@127: T1w = LDW(&(W[TWVL * 5])); cannam@127: T1C = LDW(&(W[TWVL * 21])); cannam@127: Ta = LD(&(ri[WS(rs, 4)]), ms, &(ri[0])); cannam@127: Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0])); cannam@127: T9 = LDW(&(W[TWVL * 6])); cannam@127: Tg = LD(&(ri[WS(rs, 12)]), ms, &(ri[0])); cannam@127: Tj = LD(&(ii[WS(rs, 12)]), ms, &(ii[0])); cannam@127: T2u = VMUL(T1z, T1D); cannam@127: T1B = VMUL(T1z, T1A); cannam@127: T2t = VFNMS(T1w, T1u, T2s); cannam@127: T1y = VFMA(T1w, T1x, T1v); cannam@127: Tf = LDW(&(W[TWVL * 22])); cannam@127: T1J = VMUL(T9, Td); cannam@127: Tb = VMUL(T9, Ta); cannam@127: Tc = LDW(&(W[TWVL * 7])); cannam@127: T2v = VFNMS(T1C, T1A, T2u); cannam@127: T1E = VFMA(T1C, T1D, T1B); cannam@127: Ti = LDW(&(W[TWVL * 23])); cannam@127: } cannam@127: } cannam@127: { cannam@127: V TW, TZ, TY, T27, TX, T26, TU; cannam@127: { cannam@127: V To, Tr, Tu, Tx, Tq, Tw, T1P, Tp, T1R, Tv; cannam@127: { cannam@127: V T1K, Te, T1M, Tk, Tn, Tt, T1L, Th; cannam@127: To = LD(&(ri[WS(rs, 2)]), ms, &(ri[0])); cannam@127: T1L = VMUL(Tf, Tj); cannam@127: Th = VMUL(Tf, Tg); cannam@127: Tr = LD(&(ii[WS(rs, 2)]), ms, &(ii[0])); cannam@127: T1K = VFNMS(Tc, Ta, T1J); cannam@127: Te = VFMA(Tc, Td, Tb); cannam@127: T36 = VADD(T2t, T2v); cannam@127: T2w = VSUB(T2t, T2v); cannam@127: T1F = VADD(T1y, T1E); cannam@127: T2p = VSUB(T1y, T1E); cannam@127: T1M = VFNMS(Ti, Tg, T1L); cannam@127: Tk = VFMA(Ti, Tj, Th); cannam@127: Tn = LDW(&(W[TWVL * 2])); cannam@127: Tu = LD(&(ri[WS(rs, 10)]), ms, &(ri[0])); cannam@127: Tx = LD(&(ii[WS(rs, 10)]), ms, &(ii[0])); cannam@127: Tt = LDW(&(W[TWVL * 18])); cannam@127: Tq = LDW(&(W[TWVL * 3])); cannam@127: Tw = LDW(&(W[TWVL * 19])); cannam@127: T1N = VSUB(T1K, T1M); cannam@127: T3k = VADD(T1K, T1M); cannam@127: Tl = VADD(Te, Tk); cannam@127: T3A = VSUB(Te, Tk); cannam@127: T1P = VMUL(Tn, Tr); cannam@127: Tp = VMUL(Tn, To); cannam@127: T1R = VMUL(Tt, Tx); cannam@127: Tv = VMUL(Tt, Tu); cannam@127: } cannam@127: { cannam@127: V TQ, TT, T1Q, Ts, T1S, Ty, TV, T25, TR, TP, TS; cannam@127: TQ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)])); cannam@127: TT = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)])); cannam@127: TP = LDW(&(W[0])); cannam@127: TW = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1Q = VFNMS(Tq, To, T1P); cannam@127: Ts = VFMA(Tq, Tr, Tp); cannam@127: T1S = VFNMS(Tw, Tu, T1R); cannam@127: Ty = VFMA(Tw, Tx, Tv); cannam@127: TZ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)])); cannam@127: TV = LDW(&(W[TWVL * 16])); cannam@127: T25 = VMUL(TP, TT); cannam@127: TR = VMUL(TP, TQ); cannam@127: TS = LDW(&(W[TWVL * 1])); cannam@127: TY = LDW(&(W[TWVL * 17])); cannam@127: T2V = VADD(T1Q, T1S); cannam@127: T1T = VSUB(T1Q, T1S); cannam@127: Tz = VADD(Ts, Ty); cannam@127: T1U = VSUB(Ts, Ty); cannam@127: T27 = VMUL(TV, TZ); cannam@127: TX = VMUL(TV, TW); cannam@127: T26 = VFNMS(TS, TQ, T25); cannam@127: TU = VFMA(TS, TT, TR); cannam@127: } cannam@127: } cannam@127: { cannam@127: V T19, T1c, T1b, T2f, T1a, T2e, T17; cannam@127: { cannam@127: V T13, T16, T12, T28, T10, T18, T15, T2d, T14; cannam@127: T13 = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)])); cannam@127: T16 = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)])); cannam@127: T12 = LDW(&(W[TWVL * 8])); cannam@127: T19 = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)])); cannam@127: T28 = VFNMS(TY, TW, T27); cannam@127: T10 = VFMA(TY, TZ, TX); cannam@127: T1c = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)])); cannam@127: T18 = LDW(&(W[TWVL * 24])); cannam@127: T15 = LDW(&(W[TWVL * 9])); cannam@127: T1b = LDW(&(W[TWVL * 25])); cannam@127: T2d = VMUL(T12, T16); cannam@127: T14 = VMUL(T12, T13); cannam@127: T30 = VADD(T26, T28); cannam@127: T29 = VSUB(T26, T28); cannam@127: T11 = VADD(TU, T10); cannam@127: T2c = VSUB(TU, T10); cannam@127: T2f = VMUL(T18, T1c); cannam@127: T1a = VMUL(T18, T19); cannam@127: T2e = VFNMS(T15, T13, T2d); cannam@127: T17 = VFMA(T15, T16, T14); cannam@127: } cannam@127: { cannam@127: V TB, TE, TA, T2g, T1d, TG, TD, T1X, TC; cannam@127: TB = LD(&(ri[WS(rs, 14)]), ms, &(ri[0])); cannam@127: TE = LD(&(ii[WS(rs, 14)]), ms, &(ii[0])); cannam@127: TA = LDW(&(W[TWVL * 26])); cannam@127: TH = LD(&(ri[WS(rs, 6)]), ms, &(ri[0])); cannam@127: T2g = VFNMS(T1b, T19, T2f); cannam@127: T1d = VFMA(T1b, T1c, T1a); cannam@127: TK = LD(&(ii[WS(rs, 6)]), ms, &(ii[0])); cannam@127: TG = LDW(&(W[TWVL * 10])); cannam@127: TD = LDW(&(W[TWVL * 27])); cannam@127: TJ = LDW(&(W[TWVL * 11])); cannam@127: T1X = VMUL(TA, TE); cannam@127: TC = VMUL(TA, TB); cannam@127: T31 = VADD(T2e, T2g); cannam@127: T2h = VSUB(T2e, T2g); cannam@127: T1e = VADD(T17, T1d); cannam@127: T2a = VSUB(T17, T1d); cannam@127: T1Z = VMUL(TG, TK); cannam@127: TI = VMUL(TG, TH); cannam@127: T1Y = VFNMS(TD, TB, T1X); cannam@127: TF = VFMA(TD, TE, TC); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: V T2U, Tm, T3p, T3u, T34, T1G, T1f, T2Z, T20, TL, T32, T3f, T3g, T37; cannam@127: T2U = VSUB(T8, Tl); cannam@127: Tm = VADD(T8, Tl); cannam@127: T3p = VADD(T3k, T3o); cannam@127: T3u = VSUB(T3o, T3k); cannam@127: T34 = VSUB(T1s, T1F); cannam@127: T1G = VADD(T1s, T1F); cannam@127: T1f = VADD(T11, T1e); cannam@127: T2Z = VSUB(T11, T1e); cannam@127: T20 = VFNMS(TJ, TH, T1Z); cannam@127: TL = VFMA(TJ, TK, TI); cannam@127: T32 = VSUB(T30, T31); cannam@127: T3f = VADD(T30, T31); cannam@127: T3g = VADD(T35, T36); cannam@127: T37 = VSUB(T35, T36); cannam@127: { cannam@127: V T3r, T1H, T21, T1W, T3i, T3h, T3j, T2X, TN, T3t, T2W, TM; cannam@127: T3r = VSUB(T1G, T1f); cannam@127: T1H = VADD(T1f, T1G); cannam@127: T21 = VSUB(T1Y, T20); cannam@127: T2W = VADD(T1Y, T20); cannam@127: T1W = VSUB(TF, TL); cannam@127: TM = VADD(TF, TL); cannam@127: T3i = VADD(T3f, T3g); cannam@127: T3h = VSUB(T3f, T3g); cannam@127: T3j = VADD(T2V, T2W); cannam@127: T2X = VSUB(T2V, T2W); cannam@127: TN = VADD(Tz, TM); cannam@127: T3t = VSUB(TM, Tz); cannam@127: { cannam@127: V T2E, T1O, T3B, T3H, T2x, T2q, T2K, T2J, T3C, T23, T3I, T2H; cannam@127: { cannam@127: V T2F, T1V, T22, T2G; cannam@127: T2E = VADD(T1I, T1N); cannam@127: T1O = VSUB(T1I, T1N); cannam@127: { cannam@127: V T3b, T33, T3c, T38; cannam@127: T3b = VSUB(T32, T2Z); cannam@127: T33 = VADD(T2Z, T32); cannam@127: T3c = VADD(T34, T37); cannam@127: T38 = VSUB(T34, T37); cannam@127: { cannam@127: V T3a, T2Y, T3s, T3q; cannam@127: T3a = VSUB(T2U, T2X); cannam@127: T2Y = VADD(T2U, T2X); cannam@127: T3s = VSUB(T3p, T3j); cannam@127: T3q = VADD(T3j, T3p); cannam@127: { cannam@127: V T3x, T3v, T3e, TO; cannam@127: T3x = VSUB(T3u, T3t); cannam@127: T3v = VADD(T3t, T3u); cannam@127: T3e = VSUB(Tm, TN); cannam@127: TO = VADD(Tm, TN); cannam@127: { cannam@127: V T3d, T3w, T3y, T39; cannam@127: T3d = VSUB(T3b, T3c); cannam@127: T3w = VADD(T3b, T3c); cannam@127: T3y = VSUB(T38, T33); cannam@127: T39 = VADD(T33, T38); cannam@127: ST(&(ii[WS(rs, 4)]), VADD(T3r, T3s), ms, &(ii[0])); cannam@127: ST(&(ii[WS(rs, 12)]), VSUB(T3s, T3r), ms, &(ii[0])); cannam@127: ST(&(ii[0]), VADD(T3i, T3q), ms, &(ii[0])); cannam@127: ST(&(ii[WS(rs, 8)]), VSUB(T3q, T3i), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 4)]), VADD(T3e, T3h), ms, &(ri[0])); cannam@127: ST(&(ri[WS(rs, 12)]), VSUB(T3e, T3h), ms, &(ri[0])); cannam@127: ST(&(ri[0]), VADD(TO, T1H), ms, &(ri[0])); cannam@127: ST(&(ri[WS(rs, 8)]), VSUB(TO, T1H), ms, &(ri[0])); cannam@127: ST(&(ri[WS(rs, 6)]), VFMA(LDK(KP707106781), T3d, T3a), ms, &(ri[0])); cannam@127: ST(&(ri[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3d, T3a), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3w, T3v), ms, &(ii[0])); cannam@127: ST(&(ii[WS(rs, 2)]), VFMA(LDK(KP707106781), T3w, T3v), ms, &(ii[0])); cannam@127: ST(&(ii[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3y, T3x), ms, &(ii[0])); cannam@127: ST(&(ii[WS(rs, 6)]), VFMA(LDK(KP707106781), T3y, T3x), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 2)]), VFMA(LDK(KP707106781), T39, T2Y), ms, &(ri[0])); cannam@127: ST(&(ri[WS(rs, 10)]), VFNMS(LDK(KP707106781), T39, T2Y), ms, &(ri[0])); cannam@127: T3B = VSUB(T3z, T3A); cannam@127: T3H = VADD(T3A, T3z); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: T2F = VADD(T1U, T1T); cannam@127: T1V = VSUB(T1T, T1U); cannam@127: T22 = VADD(T1W, T21); cannam@127: T2G = VSUB(T1W, T21); cannam@127: { cannam@127: V T2M, T2N, T2b, T2i; cannam@127: T2x = VSUB(T2r, T2w); cannam@127: T2M = VADD(T2r, T2w); cannam@127: T2N = VSUB(T2o, T2p); cannam@127: T2q = VADD(T2o, T2p); cannam@127: T2K = VSUB(T29, T2a); cannam@127: T2b = VADD(T29, T2a); cannam@127: T2i = VSUB(T2c, T2h); cannam@127: T2J = VADD(T2c, T2h); cannam@127: T3C = VADD(T1V, T22); cannam@127: T23 = VSUB(T1V, T22); cannam@127: T2S = VFMA(LDK(KP414213562), T2M, T2N); cannam@127: T2O = VFNMS(LDK(KP414213562), T2N, T2M); cannam@127: T3I = VSUB(T2G, T2F); cannam@127: T2H = VADD(T2F, T2G); cannam@127: T2B = VFNMS(LDK(KP414213562), T2b, T2i); cannam@127: T2j = VFMA(LDK(KP414213562), T2i, T2b); cannam@127: } cannam@127: } cannam@127: T2A = VFNMS(LDK(KP707106781), T23, T1O); cannam@127: T24 = VFMA(LDK(KP707106781), T23, T1O); cannam@127: T3J = VFMA(LDK(KP707106781), T3I, T3H); cannam@127: T3L = VFNMS(LDK(KP707106781), T3I, T3H); cannam@127: T2Q = VFNMS(LDK(KP707106781), T2H, T2E); cannam@127: T2I = VFMA(LDK(KP707106781), T2H, T2E); cannam@127: T2R = VFNMS(LDK(KP414213562), T2J, T2K); cannam@127: T2L = VFMA(LDK(KP414213562), T2K, T2J); cannam@127: T2C = VFMA(LDK(KP414213562), T2q, T2x); cannam@127: T2y = VFNMS(LDK(KP414213562), T2x, T2q); cannam@127: T3D = VFMA(LDK(KP707106781), T3C, T3B); cannam@127: T3F = VFNMS(LDK(KP707106781), T3C, T3B); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: V T3E, T2T, T2P, T3G; cannam@127: T3E = VADD(T2R, T2S); cannam@127: T2T = VSUB(T2R, T2S); cannam@127: T2P = VADD(T2L, T2O); cannam@127: T3G = VSUB(T2O, T2L); cannam@127: { cannam@127: V T3K, T2D, T2z, T3M; cannam@127: T3K = VSUB(T2C, T2B); cannam@127: T2D = VADD(T2B, T2C); cannam@127: T2z = VSUB(T2j, T2y); cannam@127: T3M = VADD(T2j, T2y); cannam@127: ST(&(ri[WS(rs, 5)]), VFMA(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 13)]), VFNMS(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 9)]), VFNMS(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 13)]), VFNMS(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 5)]), VFMA(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 9)]), VFNMS(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 15)]), VFMA(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 11)]), VFNMS(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 15)]), VFMA(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 11)]), VFNMS(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 2), cannam@127: VTW(0, 3), cannam@127: VTW(0, 4), cannam@127: VTW(0, 5), cannam@127: VTW(0, 6), cannam@127: VTW(0, 7), cannam@127: VTW(0, 8), cannam@127: VTW(0, 9), cannam@127: VTW(0, 10), cannam@127: VTW(0, 11), cannam@127: VTW(0, 12), cannam@127: VTW(0, 13), cannam@127: VTW(0, 14), cannam@127: VTW(0, 15), cannam@127: {TW_NEXT, (2 * VL), 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {104, 30, 70, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t1sv_16) (planner *p) { cannam@127: X(kdft_dit_register) (p, t1sv_16, &desc); cannam@127: } cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1sv_16 -include ts.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 174 FP additions, 84 FP multiplications, cannam@127: * (or, 136 additions, 46 multiplications, 38 fused multiply/add), cannam@127: * 52 stack variables, 3 constants, and 64 memory accesses cannam@127: */ cannam@127: #include "ts.h" cannam@127: cannam@127: static void t1sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); cannam@127: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT m; cannam@127: for (m = mb, W = W + (mb * 30); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 30), MAKE_VOLATILE_STRIDE(32, rs)) { cannam@127: V T7, T37, T1t, T2U, Ti, T38, T1w, T2R, Tu, T2s, T1C, T2c, TF, T2t, T1H; cannam@127: V T2d, T1f, T1q, T2B, T2C, T2D, T2E, T1Z, T2j, T24, T2k, TS, T13, T2w, T2x; cannam@127: V T2y, T2z, T1O, T2g, T1T, T2h; cannam@127: { cannam@127: V T1, T2T, T6, T2S; cannam@127: T1 = LD(&(ri[0]), ms, &(ri[0])); cannam@127: T2T = LD(&(ii[0]), ms, &(ii[0])); cannam@127: { cannam@127: V T3, T5, T2, T4; cannam@127: T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0])); cannam@127: T5 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0])); cannam@127: T2 = LDW(&(W[TWVL * 14])); cannam@127: T4 = LDW(&(W[TWVL * 15])); cannam@127: T6 = VFMA(T2, T3, VMUL(T4, T5)); cannam@127: T2S = VFNMS(T4, T3, VMUL(T2, T5)); cannam@127: } cannam@127: T7 = VADD(T1, T6); cannam@127: T37 = VSUB(T2T, T2S); cannam@127: T1t = VSUB(T1, T6); cannam@127: T2U = VADD(T2S, T2T); cannam@127: } cannam@127: { cannam@127: V Tc, T1u, Th, T1v; cannam@127: { cannam@127: V T9, Tb, T8, Ta; cannam@127: T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0])); cannam@127: Tb = LD(&(ii[WS(rs, 4)]), ms, &(ii[0])); cannam@127: T8 = LDW(&(W[TWVL * 6])); cannam@127: Ta = LDW(&(W[TWVL * 7])); cannam@127: Tc = VFMA(T8, T9, VMUL(Ta, Tb)); cannam@127: T1u = VFNMS(Ta, T9, VMUL(T8, Tb)); cannam@127: } cannam@127: { cannam@127: V Te, Tg, Td, Tf; cannam@127: Te = LD(&(ri[WS(rs, 12)]), ms, &(ri[0])); cannam@127: Tg = LD(&(ii[WS(rs, 12)]), ms, &(ii[0])); cannam@127: Td = LDW(&(W[TWVL * 22])); cannam@127: Tf = LDW(&(W[TWVL * 23])); cannam@127: Th = VFMA(Td, Te, VMUL(Tf, Tg)); cannam@127: T1v = VFNMS(Tf, Te, VMUL(Td, Tg)); cannam@127: } cannam@127: Ti = VADD(Tc, Th); cannam@127: T38 = VSUB(Tc, Th); cannam@127: T1w = VSUB(T1u, T1v); cannam@127: T2R = VADD(T1u, T1v); cannam@127: } cannam@127: { cannam@127: V To, T1y, Tt, T1z, T1A, T1B; cannam@127: { cannam@127: V Tl, Tn, Tk, Tm; cannam@127: Tl = LD(&(ri[WS(rs, 2)]), ms, &(ri[0])); cannam@127: Tn = LD(&(ii[WS(rs, 2)]), ms, &(ii[0])); cannam@127: Tk = LDW(&(W[TWVL * 2])); cannam@127: Tm = LDW(&(W[TWVL * 3])); cannam@127: To = VFMA(Tk, Tl, VMUL(Tm, Tn)); cannam@127: T1y = VFNMS(Tm, Tl, VMUL(Tk, Tn)); cannam@127: } cannam@127: { cannam@127: V Tq, Ts, Tp, Tr; cannam@127: Tq = LD(&(ri[WS(rs, 10)]), ms, &(ri[0])); cannam@127: Ts = LD(&(ii[WS(rs, 10)]), ms, &(ii[0])); cannam@127: Tp = LDW(&(W[TWVL * 18])); cannam@127: Tr = LDW(&(W[TWVL * 19])); cannam@127: Tt = VFMA(Tp, Tq, VMUL(Tr, Ts)); cannam@127: T1z = VFNMS(Tr, Tq, VMUL(Tp, Ts)); cannam@127: } cannam@127: Tu = VADD(To, Tt); cannam@127: T2s = VADD(T1y, T1z); cannam@127: T1A = VSUB(T1y, T1z); cannam@127: T1B = VSUB(To, Tt); cannam@127: T1C = VSUB(T1A, T1B); cannam@127: T2c = VADD(T1B, T1A); cannam@127: } cannam@127: { cannam@127: V Tz, T1E, TE, T1F, T1D, T1G; cannam@127: { cannam@127: V Tw, Ty, Tv, Tx; cannam@127: Tw = LD(&(ri[WS(rs, 14)]), ms, &(ri[0])); cannam@127: Ty = LD(&(ii[WS(rs, 14)]), ms, &(ii[0])); cannam@127: Tv = LDW(&(W[TWVL * 26])); cannam@127: Tx = LDW(&(W[TWVL * 27])); cannam@127: Tz = VFMA(Tv, Tw, VMUL(Tx, Ty)); cannam@127: T1E = VFNMS(Tx, Tw, VMUL(Tv, Ty)); cannam@127: } cannam@127: { cannam@127: V TB, TD, TA, TC; cannam@127: TB = LD(&(ri[WS(rs, 6)]), ms, &(ri[0])); cannam@127: TD = LD(&(ii[WS(rs, 6)]), ms, &(ii[0])); cannam@127: TA = LDW(&(W[TWVL * 10])); cannam@127: TC = LDW(&(W[TWVL * 11])); cannam@127: TE = VFMA(TA, TB, VMUL(TC, TD)); cannam@127: T1F = VFNMS(TC, TB, VMUL(TA, TD)); cannam@127: } cannam@127: TF = VADD(Tz, TE); cannam@127: T2t = VADD(T1E, T1F); cannam@127: T1D = VSUB(Tz, TE); cannam@127: T1G = VSUB(T1E, T1F); cannam@127: T1H = VADD(T1D, T1G); cannam@127: T2d = VSUB(T1D, T1G); cannam@127: } cannam@127: { cannam@127: V T19, T20, T1p, T1X, T1e, T21, T1k, T1W; cannam@127: { cannam@127: V T16, T18, T15, T17; cannam@127: T16 = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)])); cannam@127: T18 = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)])); cannam@127: T15 = LDW(&(W[TWVL * 28])); cannam@127: T17 = LDW(&(W[TWVL * 29])); cannam@127: T19 = VFMA(T15, T16, VMUL(T17, T18)); cannam@127: T20 = VFNMS(T17, T16, VMUL(T15, T18)); cannam@127: } cannam@127: { cannam@127: V T1m, T1o, T1l, T1n; cannam@127: T1m = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1o = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1l = LDW(&(W[TWVL * 20])); cannam@127: T1n = LDW(&(W[TWVL * 21])); cannam@127: T1p = VFMA(T1l, T1m, VMUL(T1n, T1o)); cannam@127: T1X = VFNMS(T1n, T1m, VMUL(T1l, T1o)); cannam@127: } cannam@127: { cannam@127: V T1b, T1d, T1a, T1c; cannam@127: T1b = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1d = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1a = LDW(&(W[TWVL * 12])); cannam@127: T1c = LDW(&(W[TWVL * 13])); cannam@127: T1e = VFMA(T1a, T1b, VMUL(T1c, T1d)); cannam@127: T21 = VFNMS(T1c, T1b, VMUL(T1a, T1d)); cannam@127: } cannam@127: { cannam@127: V T1h, T1j, T1g, T1i; cannam@127: T1h = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)])); cannam@127: T1j = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)])); cannam@127: T1g = LDW(&(W[TWVL * 4])); cannam@127: T1i = LDW(&(W[TWVL * 5])); cannam@127: T1k = VFMA(T1g, T1h, VMUL(T1i, T1j)); cannam@127: T1W = VFNMS(T1i, T1h, VMUL(T1g, T1j)); cannam@127: } cannam@127: T1f = VADD(T19, T1e); cannam@127: T1q = VADD(T1k, T1p); cannam@127: T2B = VSUB(T1f, T1q); cannam@127: T2C = VADD(T20, T21); cannam@127: T2D = VADD(T1W, T1X); cannam@127: T2E = VSUB(T2C, T2D); cannam@127: { cannam@127: V T1V, T1Y, T22, T23; cannam@127: T1V = VSUB(T19, T1e); cannam@127: T1Y = VSUB(T1W, T1X); cannam@127: T1Z = VSUB(T1V, T1Y); cannam@127: T2j = VADD(T1V, T1Y); cannam@127: T22 = VSUB(T20, T21); cannam@127: T23 = VSUB(T1k, T1p); cannam@127: T24 = VADD(T22, T23); cannam@127: T2k = VSUB(T22, T23); cannam@127: } cannam@127: } cannam@127: { cannam@127: V TM, T1K, T12, T1R, TR, T1L, TX, T1Q; cannam@127: { cannam@127: V TJ, TL, TI, TK; cannam@127: TJ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)])); cannam@127: TL = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)])); cannam@127: TI = LDW(&(W[0])); cannam@127: TK = LDW(&(W[TWVL * 1])); cannam@127: TM = VFMA(TI, TJ, VMUL(TK, TL)); cannam@127: T1K = VFNMS(TK, TJ, VMUL(TI, TL)); cannam@127: } cannam@127: { cannam@127: V TZ, T11, TY, T10; cannam@127: TZ = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)])); cannam@127: T11 = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)])); cannam@127: TY = LDW(&(W[TWVL * 24])); cannam@127: T10 = LDW(&(W[TWVL * 25])); cannam@127: T12 = VFMA(TY, TZ, VMUL(T10, T11)); cannam@127: T1R = VFNMS(T10, TZ, VMUL(TY, T11)); cannam@127: } cannam@127: { cannam@127: V TO, TQ, TN, TP; cannam@127: TO = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)])); cannam@127: TQ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)])); cannam@127: TN = LDW(&(W[TWVL * 16])); cannam@127: TP = LDW(&(W[TWVL * 17])); cannam@127: TR = VFMA(TN, TO, VMUL(TP, TQ)); cannam@127: T1L = VFNMS(TP, TO, VMUL(TN, TQ)); cannam@127: } cannam@127: { cannam@127: V TU, TW, TT, TV; cannam@127: TU = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)])); cannam@127: TW = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)])); cannam@127: TT = LDW(&(W[TWVL * 8])); cannam@127: TV = LDW(&(W[TWVL * 9])); cannam@127: TX = VFMA(TT, TU, VMUL(TV, TW)); cannam@127: T1Q = VFNMS(TV, TU, VMUL(TT, TW)); cannam@127: } cannam@127: TS = VADD(TM, TR); cannam@127: T13 = VADD(TX, T12); cannam@127: T2w = VSUB(TS, T13); cannam@127: T2x = VADD(T1K, T1L); cannam@127: T2y = VADD(T1Q, T1R); cannam@127: T2z = VSUB(T2x, T2y); cannam@127: { cannam@127: V T1M, T1N, T1P, T1S; cannam@127: T1M = VSUB(T1K, T1L); cannam@127: T1N = VSUB(TX, T12); cannam@127: T1O = VADD(T1M, T1N); cannam@127: T2g = VSUB(T1M, T1N); cannam@127: T1P = VSUB(TM, TR); cannam@127: T1S = VSUB(T1Q, T1R); cannam@127: T1T = VSUB(T1P, T1S); cannam@127: T2h = VADD(T1P, T1S); cannam@127: } cannam@127: } cannam@127: { cannam@127: V T1J, T27, T3g, T3i, T26, T3h, T2a, T3d; cannam@127: { cannam@127: V T1x, T1I, T3e, T3f; cannam@127: T1x = VSUB(T1t, T1w); cannam@127: T1I = VMUL(LDK(KP707106781), VSUB(T1C, T1H)); cannam@127: T1J = VADD(T1x, T1I); cannam@127: T27 = VSUB(T1x, T1I); cannam@127: T3e = VMUL(LDK(KP707106781), VSUB(T2d, T2c)); cannam@127: T3f = VADD(T38, T37); cannam@127: T3g = VADD(T3e, T3f); cannam@127: T3i = VSUB(T3f, T3e); cannam@127: } cannam@127: { cannam@127: V T1U, T25, T28, T29; cannam@127: T1U = VFMA(LDK(KP923879532), T1O, VMUL(LDK(KP382683432), T1T)); cannam@127: T25 = VFNMS(LDK(KP923879532), T24, VMUL(LDK(KP382683432), T1Z)); cannam@127: T26 = VADD(T1U, T25); cannam@127: T3h = VSUB(T25, T1U); cannam@127: T28 = VFNMS(LDK(KP923879532), T1T, VMUL(LDK(KP382683432), T1O)); cannam@127: T29 = VFMA(LDK(KP382683432), T24, VMUL(LDK(KP923879532), T1Z)); cannam@127: T2a = VSUB(T28, T29); cannam@127: T3d = VADD(T28, T29); cannam@127: } cannam@127: ST(&(ri[WS(rs, 11)]), VSUB(T1J, T26), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 11)]), VSUB(T3g, T3d), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 3)]), VADD(T1J, T26), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 3)]), VADD(T3d, T3g), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 15)]), VSUB(T27, T2a), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 15)]), VSUB(T3i, T3h), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 7)]), VADD(T27, T2a), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 7)]), VADD(T3h, T3i), ms, &(ii[WS(rs, 1)])); cannam@127: } cannam@127: { cannam@127: V T2v, T2H, T32, T34, T2G, T33, T2K, T2Z; cannam@127: { cannam@127: V T2r, T2u, T30, T31; cannam@127: T2r = VSUB(T7, Ti); cannam@127: T2u = VSUB(T2s, T2t); cannam@127: T2v = VADD(T2r, T2u); cannam@127: T2H = VSUB(T2r, T2u); cannam@127: T30 = VSUB(TF, Tu); cannam@127: T31 = VSUB(T2U, T2R); cannam@127: T32 = VADD(T30, T31); cannam@127: T34 = VSUB(T31, T30); cannam@127: } cannam@127: { cannam@127: V T2A, T2F, T2I, T2J; cannam@127: T2A = VADD(T2w, T2z); cannam@127: T2F = VSUB(T2B, T2E); cannam@127: T2G = VMUL(LDK(KP707106781), VADD(T2A, T2F)); cannam@127: T33 = VMUL(LDK(KP707106781), VSUB(T2F, T2A)); cannam@127: T2I = VSUB(T2z, T2w); cannam@127: T2J = VADD(T2B, T2E); cannam@127: T2K = VMUL(LDK(KP707106781), VSUB(T2I, T2J)); cannam@127: T2Z = VMUL(LDK(KP707106781), VADD(T2I, T2J)); cannam@127: } cannam@127: ST(&(ri[WS(rs, 10)]), VSUB(T2v, T2G), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 10)]), VSUB(T32, T2Z), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 2)]), VADD(T2v, T2G), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 2)]), VADD(T2Z, T32), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 14)]), VSUB(T2H, T2K), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 14)]), VSUB(T34, T33), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 6)]), VADD(T2H, T2K), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 6)]), VADD(T33, T34), ms, &(ii[0])); cannam@127: } cannam@127: { cannam@127: V T2f, T2n, T3a, T3c, T2m, T3b, T2q, T35; cannam@127: { cannam@127: V T2b, T2e, T36, T39; cannam@127: T2b = VADD(T1t, T1w); cannam@127: T2e = VMUL(LDK(KP707106781), VADD(T2c, T2d)); cannam@127: T2f = VADD(T2b, T2e); cannam@127: T2n = VSUB(T2b, T2e); cannam@127: T36 = VMUL(LDK(KP707106781), VADD(T1C, T1H)); cannam@127: T39 = VSUB(T37, T38); cannam@127: T3a = VADD(T36, T39); cannam@127: T3c = VSUB(T39, T36); cannam@127: } cannam@127: { cannam@127: V T2i, T2l, T2o, T2p; cannam@127: T2i = VFMA(LDK(KP382683432), T2g, VMUL(LDK(KP923879532), T2h)); cannam@127: T2l = VFNMS(LDK(KP382683432), T2k, VMUL(LDK(KP923879532), T2j)); cannam@127: T2m = VADD(T2i, T2l); cannam@127: T3b = VSUB(T2l, T2i); cannam@127: T2o = VFNMS(LDK(KP382683432), T2h, VMUL(LDK(KP923879532), T2g)); cannam@127: T2p = VFMA(LDK(KP923879532), T2k, VMUL(LDK(KP382683432), T2j)); cannam@127: T2q = VSUB(T2o, T2p); cannam@127: T35 = VADD(T2o, T2p); cannam@127: } cannam@127: ST(&(ri[WS(rs, 9)]), VSUB(T2f, T2m), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 9)]), VSUB(T3a, T35), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 1)]), VADD(T2f, T2m), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 1)]), VADD(T35, T3a), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 13)]), VSUB(T2n, T2q), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 13)]), VSUB(T3c, T3b), ms, &(ii[WS(rs, 1)])); cannam@127: ST(&(ri[WS(rs, 5)]), VADD(T2n, T2q), ms, &(ri[WS(rs, 1)])); cannam@127: ST(&(ii[WS(rs, 5)]), VADD(T3b, T3c), ms, &(ii[WS(rs, 1)])); cannam@127: } cannam@127: { cannam@127: V TH, T2L, T2W, T2Y, T1s, T2X, T2O, T2P; cannam@127: { cannam@127: V Tj, TG, T2Q, T2V; cannam@127: Tj = VADD(T7, Ti); cannam@127: TG = VADD(Tu, TF); cannam@127: TH = VADD(Tj, TG); cannam@127: T2L = VSUB(Tj, TG); cannam@127: T2Q = VADD(T2s, T2t); cannam@127: T2V = VADD(T2R, T2U); cannam@127: T2W = VADD(T2Q, T2V); cannam@127: T2Y = VSUB(T2V, T2Q); cannam@127: } cannam@127: { cannam@127: V T14, T1r, T2M, T2N; cannam@127: T14 = VADD(TS, T13); cannam@127: T1r = VADD(T1f, T1q); cannam@127: T1s = VADD(T14, T1r); cannam@127: T2X = VSUB(T1r, T14); cannam@127: T2M = VADD(T2x, T2y); cannam@127: T2N = VADD(T2C, T2D); cannam@127: T2O = VSUB(T2M, T2N); cannam@127: T2P = VADD(T2M, T2N); cannam@127: } cannam@127: ST(&(ri[WS(rs, 8)]), VSUB(TH, T1s), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 8)]), VSUB(T2W, T2P), ms, &(ii[0])); cannam@127: ST(&(ri[0]), VADD(TH, T1s), ms, &(ri[0])); cannam@127: ST(&(ii[0]), VADD(T2P, T2W), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 12)]), VSUB(T2L, T2O), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 12)]), VSUB(T2Y, T2X), ms, &(ii[0])); cannam@127: ST(&(ri[WS(rs, 4)]), VADD(T2L, T2O), ms, &(ri[0])); cannam@127: ST(&(ii[WS(rs, 4)]), VADD(T2X, T2Y), ms, &(ii[0])); cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 2), cannam@127: VTW(0, 3), cannam@127: VTW(0, 4), cannam@127: VTW(0, 5), cannam@127: VTW(0, 6), cannam@127: VTW(0, 7), cannam@127: VTW(0, 8), cannam@127: VTW(0, 9), cannam@127: VTW(0, 10), cannam@127: VTW(0, 11), cannam@127: VTW(0, 12), cannam@127: VTW(0, 13), cannam@127: VTW(0, 14), cannam@127: VTW(0, 15), cannam@127: {TW_NEXT, (2 * VL), 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {136, 46, 38, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t1sv_16) (planner *p) { cannam@127: X(kdft_dit_register) (p, t1sv_16, &desc); cannam@127: } cannam@127: #endif /* HAVE_FMA */