Chris@42: /* Chris@42: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@42: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@42: * Chris@42: * This program is free software; you can redistribute it and/or modify Chris@42: * it under the terms of the GNU General Public License as published by Chris@42: * the Free Software Foundation; either version 2 of the License, or Chris@42: * (at your option) any later version. Chris@42: * Chris@42: * This program is distributed in the hope that it will be useful, Chris@42: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@42: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@42: * GNU General Public License for more details. Chris@42: * Chris@42: * You should have received a copy of the GNU General Public License Chris@42: * along with this program; if not, write to the Free Software Chris@42: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@42: * Chris@42: */ Chris@42: Chris@42: /* This file was automatically generated --- DO NOT EDIT */ Chris@42: /* Generated on Sat Jul 30 16:52:45 EDT 2016 */ Chris@42: Chris@42: #include "codelet-rdft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 32 -dif -sign 1 -name hc2cbdftv_32 -include hc2cbv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 249 FP additions, 192 FP multiplications, Chris@42: * (or, 119 additions, 62 multiplications, 130 fused multiply/add), Chris@42: * 166 stack variables, 7 constants, and 64 memory accesses Chris@42: */ Chris@42: #include "hc2cbv.h" Chris@42: Chris@42: static void hc2cbdftv_32(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@42: DVK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@42: DVK(KP668178637, +0.668178637919298919997757686523080761552472251); Chris@42: DVK(KP198912367, +0.198912367379658006911597622644676228597850501); Chris@42: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@42: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); Chris@42: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 62)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(128, rs)) { Chris@42: V T3a, T3N; Chris@42: { Chris@42: V T2G, T1o, T2o, T2Y, T1b, T1V, Ts, T1S, T3A, T48, T3p, T45, T31, T2z, T2H; Chris@42: V T1L, Tv, TG, TM, T3q, T1r, TX, TN, T1s, Ty, T1t, TB, TO, TQ, T1y; Chris@42: V T3t, TR, T1H, T1K, TV, T1p, T1q, T1w, TW, Tt, Tu, TE, TF, TK, TL; Chris@42: V Tw, Tx, Tz, TA, T1x; Chris@42: { Chris@42: V T1i, T4, T1j, T15, T1l, T1m, Tb, T16, Tf, T1G, Ti, T1F, Tm, T1J, T1I; Chris@42: V Tp, T2, T3, T13, T14, T5, T6, T8, T9, Td, T7, Ta, Te, Tg, Th; Chris@42: V Tk, Tl, Tn, To, T2m, Tc, T3l, T1k, T3m, T18, Tj, T3y, T1n, Tq, T19; Chris@42: V T3n, T17, T2x, T1a, T2n, T2y, Tr, T3z, T3o; Chris@42: T2 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: T3 = LD(&(Rm[WS(rs, 15)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T13 = LD(&(Rp[WS(rs, 8)]), ms, &(Rp[0])); Chris@42: T14 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T5 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); Chris@42: T6 = LD(&(Rm[WS(rs, 11)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T8 = LD(&(Rp[WS(rs, 12)]), ms, &(Rp[0])); Chris@42: T9 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Td = LD(&(Rp[WS(rs, 10)]), ms, &(Rp[0])); Chris@42: T1i = VFMACONJ(T3, T2); Chris@42: T4 = VFNMSCONJ(T3, T2); Chris@42: T1j = VFMACONJ(T14, T13); Chris@42: T15 = VFNMSCONJ(T14, T13); Chris@42: T1l = VFMACONJ(T6, T5); Chris@42: T7 = VFNMSCONJ(T6, T5); Chris@42: T1m = VFMACONJ(T9, T8); Chris@42: Ta = VFMSCONJ(T9, T8); Chris@42: Te = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tg = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); Chris@42: Th = LD(&(Rm[WS(rs, 13)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tk = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); Chris@42: Tl = LD(&(Rm[WS(rs, 9)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tn = LD(&(Rp[WS(rs, 14)]), ms, &(Rp[0])); Chris@42: To = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tb = VADD(T7, Ta); Chris@42: T16 = VSUB(T7, Ta); Chris@42: Tf = VFNMSCONJ(Te, Td); Chris@42: T1G = VFMACONJ(Te, Td); Chris@42: Ti = VFNMSCONJ(Th, Tg); Chris@42: T1F = VFMACONJ(Th, Tg); Chris@42: Tm = VFNMSCONJ(Tl, Tk); Chris@42: T1J = VFMACONJ(Tl, Tk); Chris@42: T1I = VFMACONJ(To, Tn); Chris@42: Tp = VFMSCONJ(To, Tn); Chris@42: T2m = VFMA(LDK(KP707106781), Tb, T4); Chris@42: Tc = VFNMS(LDK(KP707106781), Tb, T4); Chris@42: T3l = VSUB(T1i, T1j); Chris@42: T1k = VADD(T1i, T1j); Chris@42: T1H = VADD(T1F, T1G); Chris@42: T3m = VSUB(T1F, T1G); Chris@42: T18 = VFNMS(LDK(KP414213562), Tf, Ti); Chris@42: Tj = VFMA(LDK(KP414213562), Ti, Tf); Chris@42: T3y = VSUB(T1l, T1m); Chris@42: T1n = VADD(T1l, T1m); Chris@42: Tq = VFNMS(LDK(KP414213562), Tp, Tm); Chris@42: T19 = VFMA(LDK(KP414213562), Tm, Tp); Chris@42: T1K = VADD(T1I, T1J); Chris@42: T3n = VSUB(T1I, T1J); Chris@42: T17 = VFNMS(LDK(KP707106781), T16, T15); Chris@42: T2x = VFMA(LDK(KP707106781), T16, T15); Chris@42: T1a = VSUB(T18, T19); Chris@42: T2n = VADD(T18, T19); Chris@42: T2y = VADD(Tj, Tq); Chris@42: Tr = VSUB(Tj, Tq); Chris@42: T3z = VSUB(T3m, T3n); Chris@42: T3o = VADD(T3m, T3n); Chris@42: T2G = VADD(T1k, T1n); Chris@42: T1o = VSUB(T1k, T1n); Chris@42: T2o = VFNMS(LDK(KP923879532), T2n, T2m); Chris@42: T2Y = VFMA(LDK(KP923879532), T2n, T2m); Chris@42: T1b = VFNMS(LDK(KP923879532), T1a, T17); Chris@42: T1V = VFMA(LDK(KP923879532), T1a, T17); Chris@42: Ts = VFMA(LDK(KP923879532), Tr, Tc); Chris@42: T1S = VFNMS(LDK(KP923879532), Tr, Tc); Chris@42: T3A = VFMA(LDK(KP707106781), T3z, T3y); Chris@42: T48 = VFNMS(LDK(KP707106781), T3z, T3y); Chris@42: T3p = VFMA(LDK(KP707106781), T3o, T3l); Chris@42: T45 = VFNMS(LDK(KP707106781), T3o, T3l); Chris@42: T31 = VFMA(LDK(KP923879532), T2y, T2x); Chris@42: T2z = VFNMS(LDK(KP923879532), T2y, T2x); Chris@42: } Chris@42: Tt = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tu = LD(&(Rm[WS(rs, 14)]), -ms, &(Rm[0])); Chris@42: TE = LD(&(Rp[WS(rs, 9)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TF = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); Chris@42: TK = LD(&(Rp[WS(rs, 15)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TL = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: TV = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T2H = VADD(T1H, T1K); Chris@42: T1L = VSUB(T1H, T1K); Chris@42: Tv = VFNMSCONJ(Tu, Tt); Chris@42: T1p = VFMACONJ(Tu, Tt); Chris@42: TG = VFNMSCONJ(TF, TE); Chris@42: T1q = VFMACONJ(TF, TE); Chris@42: T1w = VFMACONJ(TL, TK); Chris@42: TM = VFMSCONJ(TL, TK); Chris@42: TW = LD(&(Rm[WS(rs, 8)]), -ms, &(Rm[0])); Chris@42: Tw = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tx = LD(&(Rm[WS(rs, 10)]), -ms, &(Rm[0])); Chris@42: Tz = LD(&(Rp[WS(rs, 13)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TA = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); Chris@42: T3q = VSUB(T1p, T1q); Chris@42: T1r = VADD(T1p, T1q); Chris@42: T1x = VFMACONJ(TW, TV); Chris@42: TX = VFNMSCONJ(TW, TV); Chris@42: TN = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T1s = VFMACONJ(Tx, Tw); Chris@42: Ty = VFNMSCONJ(Tx, Tw); Chris@42: T1t = VFMACONJ(TA, Tz); Chris@42: TB = VFMSCONJ(TA, Tz); Chris@42: TO = LD(&(Rm[WS(rs, 12)]), -ms, &(Rm[0])); Chris@42: TQ = LD(&(Rp[WS(rs, 11)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T1y = VADD(T1w, T1x); Chris@42: T3t = VSUB(T1w, T1x); Chris@42: TR = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); Chris@42: { Chris@42: V T38, T3f, T4p, T4v, T3T, T3Z, T2a, T2i, T4b, T4h, T1O, T20, T2M, T2U, T3F; Chris@42: V T3L, T2g, T3X, T3J, T1g, T4f, T2S, T4l, T2E, T2X, T3O, T3b, T3i, T26, T4t; Chris@42: V T43, T1Y, T3c, T30, T3d, T33; Chris@42: { Chris@42: V T2I, T2A, T2r, T1c, TJ, T2L, T2u, T2B, T10, T1d, T3x, T3E, T1E, T1N, T1h; Chris@42: V T1Z, T4m, T1M, T1D, T4a, T4o, T4n, T47, T4u, T3R, T3S, T3Q, T3Y, T28, T29; Chris@42: V T27, T2h, T44, T4g; Chris@42: { Chris@42: V T36, T1v, T2J, T3s, T3B, T2p, TI, T2q, TD, T1B, T3u, TY, TT, T35, T1u; Chris@42: V T3r, TH, TC, T1z, TP, T1A, TS, T3w, T3D, T1C, T2K, T3v, T3C, T2s, TZ; Chris@42: V T2t, TU, T37, T49, T46; Chris@42: T2I = VSUB(T2G, T2H); Chris@42: T36 = VADD(T2G, T2H); Chris@42: T1u = VADD(T1s, T1t); Chris@42: T3r = VSUB(T1s, T1t); Chris@42: TH = VSUB(Ty, TB); Chris@42: TC = VADD(Ty, TB); Chris@42: T1z = VFMACONJ(TO, TN); Chris@42: TP = VFNMSCONJ(TO, TN); Chris@42: T1A = VFMACONJ(TR, TQ); Chris@42: TS = VFMSCONJ(TR, TQ); Chris@42: T1v = VSUB(T1r, T1u); Chris@42: T2J = VADD(T1r, T1u); Chris@42: T3s = VFNMS(LDK(KP414213562), T3r, T3q); Chris@42: T3B = VFMA(LDK(KP414213562), T3q, T3r); Chris@42: T2p = VFMA(LDK(KP707106781), TH, TG); Chris@42: TI = VFNMS(LDK(KP707106781), TH, TG); Chris@42: T2q = VFMA(LDK(KP707106781), TC, Tv); Chris@42: TD = VFNMS(LDK(KP707106781), TC, Tv); Chris@42: T1B = VADD(T1z, T1A); Chris@42: T3u = VSUB(T1A, T1z); Chris@42: TY = VSUB(TS, TP); Chris@42: TT = VADD(TP, TS); Chris@42: T35 = LDW(&(W[TWVL * 30])); Chris@42: T4m = LDW(&(W[TWVL * 10])); Chris@42: T2A = VFNMS(LDK(KP198912367), T2p, T2q); Chris@42: T2r = VFMA(LDK(KP198912367), T2q, T2p); Chris@42: T1c = VFNMS(LDK(KP668178637), TD, TI); Chris@42: TJ = VFMA(LDK(KP668178637), TI, TD); Chris@42: T1C = VSUB(T1y, T1B); Chris@42: T2K = VADD(T1y, T1B); Chris@42: T3v = VFNMS(LDK(KP414213562), T3u, T3t); Chris@42: T3C = VFMA(LDK(KP414213562), T3t, T3u); Chris@42: T2s = VFNMS(LDK(KP707106781), TY, TX); Chris@42: TZ = VFMA(LDK(KP707106781), TY, TX); Chris@42: T2t = VFMA(LDK(KP707106781), TT, TM); Chris@42: TU = VFNMS(LDK(KP707106781), TT, TM); Chris@42: T1M = VSUB(T1v, T1C); Chris@42: T1D = VADD(T1v, T1C); Chris@42: T37 = VADD(T2J, T2K); Chris@42: T2L = VSUB(T2J, T2K); Chris@42: T3w = VADD(T3s, T3v); Chris@42: T49 = VSUB(T3s, T3v); Chris@42: T3D = VSUB(T3B, T3C); Chris@42: T46 = VADD(T3B, T3C); Chris@42: T2u = VFNMS(LDK(KP198912367), T2t, T2s); Chris@42: T2B = VFMA(LDK(KP198912367), T2s, T2t); Chris@42: T10 = VFNMS(LDK(KP668178637), TZ, TU); Chris@42: T1d = VFMA(LDK(KP668178637), TU, TZ); Chris@42: T38 = VZMUL(T35, VSUB(T36, T37)); Chris@42: T3f = VADD(T36, T37); Chris@42: T4a = VFMA(LDK(KP923879532), T49, T48); Chris@42: T4o = VFNMS(LDK(KP923879532), T49, T48); Chris@42: T4n = VFMA(LDK(KP923879532), T46, T45); Chris@42: T47 = VFNMS(LDK(KP923879532), T46, T45); Chris@42: T4u = LDW(&(W[TWVL * 50])); Chris@42: T3R = VFMA(LDK(KP923879532), T3w, T3p); Chris@42: T3x = VFNMS(LDK(KP923879532), T3w, T3p); Chris@42: T3E = VFNMS(LDK(KP923879532), T3D, T3A); Chris@42: T3S = VFMA(LDK(KP923879532), T3D, T3A); Chris@42: T3Q = LDW(&(W[TWVL * 58])); Chris@42: T3Y = LDW(&(W[TWVL * 2])); Chris@42: } Chris@42: T28 = VFMA(LDK(KP707106781), T1D, T1o); Chris@42: T1E = VFNMS(LDK(KP707106781), T1D, T1o); Chris@42: T1N = VFNMS(LDK(KP707106781), T1M, T1L); Chris@42: T29 = VFMA(LDK(KP707106781), T1M, T1L); Chris@42: T4p = VZMUL(T4m, VFNMSI(T4o, T4n)); Chris@42: T4v = VZMUL(T4u, VFMAI(T4o, T4n)); Chris@42: T27 = LDW(&(W[TWVL * 6])); Chris@42: T2h = LDW(&(W[TWVL * 54])); Chris@42: T3T = VZMUL(T3Q, VFNMSI(T3S, T3R)); Chris@42: T3Z = VZMUL(T3Y, VFMAI(T3S, T3R)); Chris@42: T44 = LDW(&(W[TWVL * 18])); Chris@42: T4g = LDW(&(W[TWVL * 42])); Chris@42: T2a = VZMUL(T27, VFMAI(T29, T28)); Chris@42: T2i = VZMUL(T2h, VFNMSI(T29, T28)); Chris@42: T1h = LDW(&(W[TWVL * 22])); Chris@42: T1Z = LDW(&(W[TWVL * 38])); Chris@42: T4b = VZMUL(T44, VFMAI(T4a, T47)); Chris@42: T4h = VZMUL(T4g, VFNMSI(T4a, T47)); Chris@42: { Chris@42: V T1W, T1T, T1, T3W, T2d, T3I, T2e, T12, T2f, T1f, T2F, T2T, T3k, T3K, T11; Chris@42: V T1e, T32, T2Z, T2l, T4k, T2P, T4e, T2Q, T2w, T2R, T2D, T2v, T2C, T1R, T4s; Chris@42: V T23, T42, T24, T1U, T25, T1X; Chris@42: T2F = LDW(&(W[TWVL * 46])); Chris@42: T2T = LDW(&(W[TWVL * 14])); Chris@42: T1O = VZMUL(T1h, VFNMSI(T1N, T1E)); Chris@42: T20 = VZMUL(T1Z, VFMAI(T1N, T1E)); Chris@42: T3k = LDW(&(W[TWVL * 26])); Chris@42: T3K = LDW(&(W[TWVL * 34])); Chris@42: T2M = VZMUL(T2F, VFNMSI(T2L, T2I)); Chris@42: T2U = VZMUL(T2T, VFMAI(T2L, T2I)); Chris@42: T11 = VADD(TJ, T10); Chris@42: T1W = VSUB(TJ, T10); Chris@42: T1T = VSUB(T1d, T1c); Chris@42: T1e = VADD(T1c, T1d); Chris@42: T1 = LDW(&(W[TWVL * 24])); Chris@42: T3W = LDW(&(W[TWVL * 4])); Chris@42: T3F = VZMUL(T3k, VFNMSI(T3E, T3x)); Chris@42: T3L = VZMUL(T3K, VFMAI(T3E, T3x)); Chris@42: T2d = LDW(&(W[TWVL * 56])); Chris@42: T3I = LDW(&(W[TWVL * 36])); Chris@42: T2e = VFMA(LDK(KP831469612), T11, Ts); Chris@42: T12 = VFNMS(LDK(KP831469612), T11, Ts); Chris@42: T2f = VFMA(LDK(KP831469612), T1e, T1b); Chris@42: T1f = VFNMS(LDK(KP831469612), T1e, T1b); Chris@42: T2v = VSUB(T2r, T2u); Chris@42: T32 = VADD(T2r, T2u); Chris@42: T2Z = VADD(T2A, T2B); Chris@42: T2C = VSUB(T2A, T2B); Chris@42: T2l = LDW(&(W[TWVL * 48])); Chris@42: T4k = LDW(&(W[TWVL * 12])); Chris@42: T2P = LDW(&(W[TWVL * 16])); Chris@42: T4e = LDW(&(W[TWVL * 44])); Chris@42: T2g = VZMULI(T2d, VFMAI(T2f, T2e)); Chris@42: T3X = VZMULI(T3W, VFNMSI(T2f, T2e)); Chris@42: T3J = VZMULI(T3I, VFNMSI(T1f, T12)); Chris@42: T1g = VZMULI(T1, VFMAI(T1f, T12)); Chris@42: T2Q = VFNMS(LDK(KP980785280), T2v, T2o); Chris@42: T2w = VFMA(LDK(KP980785280), T2v, T2o); Chris@42: T2R = VFMA(LDK(KP980785280), T2C, T2z); Chris@42: T2D = VFNMS(LDK(KP980785280), T2C, T2z); Chris@42: T1R = LDW(&(W[TWVL * 40])); Chris@42: T4s = LDW(&(W[TWVL * 52])); Chris@42: T23 = LDW(&(W[TWVL * 8])); Chris@42: T42 = LDW(&(W[TWVL * 20])); Chris@42: T4f = VZMULI(T4e, VFNMSI(T2R, T2Q)); Chris@42: T2S = VZMULI(T2P, VFMAI(T2R, T2Q)); Chris@42: T4l = VZMULI(T4k, VFNMSI(T2D, T2w)); Chris@42: T2E = VZMULI(T2l, VFMAI(T2D, T2w)); Chris@42: T24 = VFMA(LDK(KP831469612), T1T, T1S); Chris@42: T1U = VFNMS(LDK(KP831469612), T1T, T1S); Chris@42: T25 = VFMA(LDK(KP831469612), T1W, T1V); Chris@42: T1X = VFNMS(LDK(KP831469612), T1W, T1V); Chris@42: T2X = LDW(&(W[TWVL * 32])); Chris@42: T3O = LDW(&(W[TWVL * 60])); Chris@42: T3b = LDW(&(W[0])); Chris@42: T3i = LDW(&(W[TWVL * 28])); Chris@42: T26 = VZMULI(T23, VFMAI(T25, T24)); Chris@42: T4t = VZMULI(T4s, VFNMSI(T25, T24)); Chris@42: T43 = VZMULI(T42, VFNMSI(T1X, T1U)); Chris@42: T1Y = VZMULI(T1R, VFMAI(T1X, T1U)); Chris@42: T3c = VFMA(LDK(KP980785280), T2Z, T2Y); Chris@42: T30 = VFNMS(LDK(KP980785280), T2Z, T2Y); Chris@42: T3d = VFMA(LDK(KP980785280), T32, T31); Chris@42: T33 = VFNMS(LDK(KP980785280), T32, T31); Chris@42: } Chris@42: } Chris@42: { Chris@42: V T3e, T3P, T3j, T34, T2c, T4j, T2k, T4d, T1P, T1Q, T4x, T4w, T2j, T4c, T21; Chris@42: V T22, T4r, T4q, T2b, T4i, T3h, T3H, T2N, T2O, T41, T40, T3g, T3G, T2V, T2W; Chris@42: V T3V, T3U, T39, T3M; Chris@42: T1P = VADD(T1g, T1O); Chris@42: T1Q = VCONJ(VSUB(T1O, T1g)); Chris@42: T4x = VCONJ(VSUB(T4v, T4t)); Chris@42: T4w = VADD(T4t, T4v); Chris@42: T2j = VADD(T2g, T2i); Chris@42: T2k = VCONJ(VSUB(T2i, T2g)); Chris@42: T4d = VCONJ(VSUB(T4b, T43)); Chris@42: T4c = VADD(T43, T4b); Chris@42: T3e = VZMULI(T3b, VFMAI(T3d, T3c)); Chris@42: T3P = VZMULI(T3O, VFNMSI(T3d, T3c)); Chris@42: T3j = VZMULI(T3i, VFNMSI(T33, T30)); Chris@42: T34 = VZMULI(T2X, VFMAI(T33, T30)); Chris@42: ST(&(Rp[WS(rs, 6)]), T1P, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 13)]), T4w, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rp[WS(rs, 14)]), T2j, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 5)]), T4c, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 13)]), T4x, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 6)]), T1Q, -ms, &(Rm[0])); Chris@42: T21 = VADD(T1Y, T20); Chris@42: T22 = VCONJ(VSUB(T20, T1Y)); Chris@42: T4r = VCONJ(VSUB(T4p, T4l)); Chris@42: T4q = VADD(T4l, T4p); Chris@42: T2b = VADD(T26, T2a); Chris@42: T2c = VCONJ(VSUB(T2a, T26)); Chris@42: T4j = VCONJ(VSUB(T4h, T4f)); Chris@42: T4i = VADD(T4f, T4h); Chris@42: ST(&(Rm[WS(rs, 5)]), T4d, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 14)]), T2k, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 10)]), T21, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 3)]), T4q, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rp[WS(rs, 2)]), T2b, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 11)]), T4i, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 3)]), T4r, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 10)]), T22, -ms, &(Rm[0])); Chris@42: T2N = VADD(T2E, T2M); Chris@42: T2O = VCONJ(VSUB(T2M, T2E)); Chris@42: T41 = VCONJ(VSUB(T3Z, T3X)); Chris@42: T40 = VADD(T3X, T3Z); Chris@42: T3g = VADD(T3e, T3f); Chris@42: T3h = VCONJ(VSUB(T3f, T3e)); Chris@42: T3H = VCONJ(VSUB(T3F, T3j)); Chris@42: T3G = VADD(T3j, T3F); Chris@42: ST(&(Rm[WS(rs, 11)]), T4j, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 2)]), T2c, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 12)]), T2N, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 1)]), T40, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rp[0]), T3g, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 7)]), T3G, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 1)]), T41, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 12)]), T2O, -ms, &(Rm[0])); Chris@42: T2V = VADD(T2S, T2U); Chris@42: T2W = VCONJ(VSUB(T2U, T2S)); Chris@42: T3V = VCONJ(VSUB(T3T, T3P)); Chris@42: T3U = VADD(T3P, T3T); Chris@42: T39 = VADD(T34, T38); Chris@42: T3a = VCONJ(VSUB(T38, T34)); Chris@42: T3N = VCONJ(VSUB(T3L, T3J)); Chris@42: T3M = VADD(T3J, T3L); Chris@42: ST(&(Rm[WS(rs, 7)]), T3H, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[0]), T3h, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 4)]), T2V, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 15)]), T3U, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rp[WS(rs, 8)]), T39, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 9)]), T3M, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 15)]), T3V, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 4)]), T2W, -ms, &(Rm[0])); Chris@42: } Chris@42: } Chris@42: } Chris@42: ST(&(Rm[WS(rs, 9)]), T3N, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 8)]), T3a, -ms, &(Rm[0])); Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(1, 1), Chris@42: VTW(1, 2), Chris@42: VTW(1, 3), Chris@42: VTW(1, 4), Chris@42: VTW(1, 5), Chris@42: VTW(1, 6), Chris@42: VTW(1, 7), Chris@42: VTW(1, 8), Chris@42: VTW(1, 9), Chris@42: VTW(1, 10), Chris@42: VTW(1, 11), Chris@42: VTW(1, 12), Chris@42: VTW(1, 13), Chris@42: VTW(1, 14), Chris@42: VTW(1, 15), Chris@42: VTW(1, 16), Chris@42: VTW(1, 17), Chris@42: VTW(1, 18), Chris@42: VTW(1, 19), Chris@42: VTW(1, 20), Chris@42: VTW(1, 21), Chris@42: VTW(1, 22), Chris@42: VTW(1, 23), Chris@42: VTW(1, 24), Chris@42: VTW(1, 25), Chris@42: VTW(1, 26), Chris@42: VTW(1, 27), Chris@42: VTW(1, 28), Chris@42: VTW(1, 29), Chris@42: VTW(1, 30), Chris@42: VTW(1, 31), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 32, XSIMD_STRING("hc2cbdftv_32"), twinstr, &GENUS, {119, 62, 130, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cbdftv_32) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cbdftv_32, &desc, HC2C_VIA_DFT); Chris@42: } Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 32 -dif -sign 1 -name hc2cbdftv_32 -include hc2cbv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 249 FP additions, 104 FP multiplications, Chris@42: * (or, 233 additions, 88 multiplications, 16 fused multiply/add), Chris@42: * 161 stack variables, 7 constants, and 64 memory accesses Chris@42: */ Chris@42: #include "hc2cbv.h" Chris@42: Chris@42: static void hc2cbdftv_32(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP195090322, +0.195090322016128267848284868477022240927691618); Chris@42: DVK(KP980785280, +0.980785280403230449126182236134239036973933731); Chris@42: DVK(KP555570233, +0.555570233019602224742830813948532874374937191); Chris@42: DVK(KP831469612, +0.831469612302545237078788377617905756738560812); Chris@42: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@42: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); Chris@42: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 62)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(128, rs)) { Chris@42: V T1W, T21, Tf, T2c, T1t, T2r, T3T, T4m, Ty, T2q, T3P, T4n, T1n, T2d, T1T; Chris@42: V T22, T1E, T24, T3I, T4p, TU, T2n, T1i, T2h, T1L, T25, T3L, T4q, T1f, T2o; Chris@42: V T1j, T2k; Chris@42: { Chris@42: V T2, T4, T1Z, T1p, T1r, T20, T9, T1U, Td, T1V, T3, T1q, T6, T8, T7; Chris@42: V Tc, Tb, Ta, T5, Te, T1o, T1s, T3R, T3S, Tj, T1N, Tw, T1Q, Tn, T1O; Chris@42: V Ts, T1R, Tg, Ti, Th, Tv, Tu, Tt, Tk, Tm, Tl, Tp, Tr, Tq, To; Chris@42: V Tx, T3N, T3O, T1l, T1m, T1P, T1S; Chris@42: T2 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: T3 = LD(&(Rm[WS(rs, 15)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T4 = VCONJ(T3); Chris@42: T1Z = VADD(T2, T4); Chris@42: T1p = LD(&(Rp[WS(rs, 8)]), ms, &(Rp[0])); Chris@42: T1q = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T1r = VCONJ(T1q); Chris@42: T20 = VADD(T1p, T1r); Chris@42: T6 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); Chris@42: T7 = LD(&(Rm[WS(rs, 11)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T8 = VCONJ(T7); Chris@42: T9 = VSUB(T6, T8); Chris@42: T1U = VADD(T6, T8); Chris@42: Tc = LD(&(Rp[WS(rs, 12)]), ms, &(Rp[0])); Chris@42: Ta = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tb = VCONJ(Ta); Chris@42: Td = VSUB(Tb, Tc); Chris@42: T1V = VADD(Tb, Tc); Chris@42: T1W = VSUB(T1U, T1V); Chris@42: T21 = VSUB(T1Z, T20); Chris@42: T5 = VSUB(T2, T4); Chris@42: Te = VMUL(LDK(KP707106781), VADD(T9, Td)); Chris@42: Tf = VSUB(T5, Te); Chris@42: T2c = VADD(T5, Te); Chris@42: T1o = VMUL(LDK(KP707106781), VSUB(T9, Td)); Chris@42: T1s = VSUB(T1p, T1r); Chris@42: T1t = VSUB(T1o, T1s); Chris@42: T2r = VADD(T1s, T1o); Chris@42: T3R = VADD(T1Z, T20); Chris@42: T3S = VADD(T1U, T1V); Chris@42: T3T = VSUB(T3R, T3S); Chris@42: T4m = VADD(T3R, T3S); Chris@42: Tg = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); Chris@42: Th = LD(&(Rm[WS(rs, 13)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Ti = VCONJ(Th); Chris@42: Tj = VSUB(Tg, Ti); Chris@42: T1N = VADD(Tg, Ti); Chris@42: Tv = LD(&(Rp[WS(rs, 14)]), ms, &(Rp[0])); Chris@42: Tt = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tu = VCONJ(Tt); Chris@42: Tw = VSUB(Tu, Tv); Chris@42: T1Q = VADD(Tu, Tv); Chris@42: Tk = LD(&(Rp[WS(rs, 10)]), ms, &(Rp[0])); Chris@42: Tl = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tm = VCONJ(Tl); Chris@42: Tn = VSUB(Tk, Tm); Chris@42: T1O = VADD(Tk, Tm); Chris@42: Tp = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); Chris@42: Tq = LD(&(Rm[WS(rs, 9)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tr = VCONJ(Tq); Chris@42: Ts = VSUB(Tp, Tr); Chris@42: T1R = VADD(Tp, Tr); Chris@42: To = VFMA(LDK(KP382683432), Tj, VMUL(LDK(KP923879532), Tn)); Chris@42: Tx = VFNMS(LDK(KP382683432), Tw, VMUL(LDK(KP923879532), Ts)); Chris@42: Ty = VSUB(To, Tx); Chris@42: T2q = VADD(To, Tx); Chris@42: T3N = VADD(T1N, T1O); Chris@42: T3O = VADD(T1Q, T1R); Chris@42: T3P = VSUB(T3N, T3O); Chris@42: T4n = VADD(T3N, T3O); Chris@42: T1l = VFNMS(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tj)); Chris@42: T1m = VFMA(LDK(KP923879532), Tw, VMUL(LDK(KP382683432), Ts)); Chris@42: T1n = VSUB(T1l, T1m); Chris@42: T2d = VADD(T1l, T1m); Chris@42: T1P = VSUB(T1N, T1O); Chris@42: T1S = VSUB(T1Q, T1R); Chris@42: T1T = VMUL(LDK(KP707106781), VSUB(T1P, T1S)); Chris@42: T22 = VMUL(LDK(KP707106781), VADD(T1P, T1S)); Chris@42: } Chris@42: { Chris@42: V TD, T1B, TR, T1y, TH, T1C, TM, T1z, TA, TC, TB, TO, TQ, TP, TG; Chris@42: V TF, TE, TJ, TL, TK, T1A, T1D, T3G, T3H, TN, T2f, TT, T2g, TI, TS; Chris@42: V TY, T1I, T1c, T1F, T12, T1J, T17, T1G, TV, TX, TW, T1b, T1a, T19, T11; Chris@42: V T10, TZ, T14, T16, T15, T1H, T1K, T3J, T3K, T18, T2i, T1e, T2j, T13, T1d; Chris@42: TA = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TB = LD(&(Rm[WS(rs, 10)]), -ms, &(Rm[0])); Chris@42: TC = VCONJ(TB); Chris@42: TD = VSUB(TA, TC); Chris@42: T1B = VADD(TA, TC); Chris@42: TO = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TP = LD(&(Rm[WS(rs, 14)]), -ms, &(Rm[0])); Chris@42: TQ = VCONJ(TP); Chris@42: TR = VSUB(TO, TQ); Chris@42: T1y = VADD(TO, TQ); Chris@42: TG = LD(&(Rp[WS(rs, 13)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TE = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); Chris@42: TF = VCONJ(TE); Chris@42: TH = VSUB(TF, TG); Chris@42: T1C = VADD(TF, TG); Chris@42: TJ = LD(&(Rp[WS(rs, 9)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TK = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); Chris@42: TL = VCONJ(TK); Chris@42: TM = VSUB(TJ, TL); Chris@42: T1z = VADD(TJ, TL); Chris@42: T1A = VSUB(T1y, T1z); Chris@42: T1D = VSUB(T1B, T1C); Chris@42: T1E = VFNMS(LDK(KP382683432), T1D, VMUL(LDK(KP923879532), T1A)); Chris@42: T24 = VFMA(LDK(KP382683432), T1A, VMUL(LDK(KP923879532), T1D)); Chris@42: T3G = VADD(T1y, T1z); Chris@42: T3H = VADD(T1B, T1C); Chris@42: T3I = VSUB(T3G, T3H); Chris@42: T4p = VADD(T3G, T3H); Chris@42: TI = VMUL(LDK(KP707106781), VSUB(TD, TH)); Chris@42: TN = VSUB(TI, TM); Chris@42: T2f = VADD(TM, TI); Chris@42: TS = VMUL(LDK(KP707106781), VADD(TD, TH)); Chris@42: TT = VSUB(TR, TS); Chris@42: T2g = VADD(TR, TS); Chris@42: TU = VFMA(LDK(KP831469612), TN, VMUL(LDK(KP555570233), TT)); Chris@42: T2n = VFNMS(LDK(KP195090322), T2f, VMUL(LDK(KP980785280), T2g)); Chris@42: T1i = VFNMS(LDK(KP555570233), TN, VMUL(LDK(KP831469612), TT)); Chris@42: T2h = VFMA(LDK(KP980785280), T2f, VMUL(LDK(KP195090322), T2g)); Chris@42: TV = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TW = LD(&(Rm[WS(rs, 12)]), -ms, &(Rm[0])); Chris@42: TX = VCONJ(TW); Chris@42: TY = VSUB(TV, TX); Chris@42: T1I = VADD(TV, TX); Chris@42: T1b = LD(&(Rp[WS(rs, 15)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T19 = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: T1a = VCONJ(T19); Chris@42: T1c = VSUB(T1a, T1b); Chris@42: T1F = VADD(T1a, T1b); Chris@42: T11 = LD(&(Rp[WS(rs, 11)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TZ = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); Chris@42: T10 = VCONJ(TZ); Chris@42: T12 = VSUB(T10, T11); Chris@42: T1J = VADD(T10, T11); Chris@42: T14 = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T15 = LD(&(Rm[WS(rs, 8)]), -ms, &(Rm[0])); Chris@42: T16 = VCONJ(T15); Chris@42: T17 = VSUB(T14, T16); Chris@42: T1G = VADD(T14, T16); Chris@42: T1H = VSUB(T1F, T1G); Chris@42: T1K = VSUB(T1I, T1J); Chris@42: T1L = VFMA(LDK(KP923879532), T1H, VMUL(LDK(KP382683432), T1K)); Chris@42: T25 = VFNMS(LDK(KP382683432), T1H, VMUL(LDK(KP923879532), T1K)); Chris@42: T3J = VADD(T1F, T1G); Chris@42: T3K = VADD(T1I, T1J); Chris@42: T3L = VSUB(T3J, T3K); Chris@42: T4q = VADD(T3J, T3K); Chris@42: T13 = VMUL(LDK(KP707106781), VSUB(TY, T12)); Chris@42: T18 = VSUB(T13, T17); Chris@42: T2i = VADD(T17, T13); Chris@42: T1d = VMUL(LDK(KP707106781), VADD(TY, T12)); Chris@42: T1e = VSUB(T1c, T1d); Chris@42: T2j = VADD(T1c, T1d); Chris@42: T1f = VFNMS(LDK(KP555570233), T1e, VMUL(LDK(KP831469612), T18)); Chris@42: T2o = VFMA(LDK(KP195090322), T2i, VMUL(LDK(KP980785280), T2j)); Chris@42: T1j = VFMA(LDK(KP555570233), T18, VMUL(LDK(KP831469612), T1e)); Chris@42: T2k = VFNMS(LDK(KP195090322), T2j, VMUL(LDK(KP980785280), T2i)); Chris@42: } Chris@42: { Chris@42: V T4L, T4G, T4s, T4y, T3W, T4g, T42, T4a, T3g, T4e, T3o, T3E, T1w, T46, T2M; Chris@42: V T40, T2u, T4w, T2C, T4k, T36, T3A, T3i, T3s, T28, T2O, T2w, T2G, T2Y, T4K; Chris@42: V T3y, T4C; Chris@42: { Chris@42: V T4E, T4F, T4D, T4o, T4r, T4l, T4x, T3Q, T48, T3V, T49, T3M, T3U, T3F, T4f; Chris@42: V T41, T47, T3c, T3n, T3f, T3m, T3a, T3b, T3d, T3e, T39, T4d, T3l, T3D, T1h; Chris@42: V T2K, T1v, T2L, Tz, T1g, T1k, T1u, T1, T45, T2J, T3Z, T2m, T2A, T2t, T2B; Chris@42: V T2e, T2l, T2p, T2s, T2b, T4v, T2z, T4j; Chris@42: T4E = VADD(T4m, T4n); Chris@42: T4F = VADD(T4p, T4q); Chris@42: T4L = VADD(T4E, T4F); Chris@42: T4D = LDW(&(W[TWVL * 30])); Chris@42: T4G = VZMUL(T4D, VSUB(T4E, T4F)); Chris@42: T4o = VSUB(T4m, T4n); Chris@42: T4r = VBYI(VSUB(T4p, T4q)); Chris@42: T4l = LDW(&(W[TWVL * 46])); Chris@42: T4s = VZMUL(T4l, VSUB(T4o, T4r)); Chris@42: T4x = LDW(&(W[TWVL * 14])); Chris@42: T4y = VZMUL(T4x, VADD(T4o, T4r)); Chris@42: T3M = VMUL(LDK(KP707106781), VSUB(T3I, T3L)); Chris@42: T3Q = VBYI(VSUB(T3M, T3P)); Chris@42: T48 = VBYI(VADD(T3P, T3M)); Chris@42: T3U = VMUL(LDK(KP707106781), VADD(T3I, T3L)); Chris@42: T3V = VSUB(T3T, T3U); Chris@42: T49 = VADD(T3T, T3U); Chris@42: T3F = LDW(&(W[TWVL * 22])); Chris@42: T3W = VZMUL(T3F, VADD(T3Q, T3V)); Chris@42: T4f = LDW(&(W[TWVL * 54])); Chris@42: T4g = VZMUL(T4f, VSUB(T49, T48)); Chris@42: T41 = LDW(&(W[TWVL * 38])); Chris@42: T42 = VZMUL(T41, VSUB(T3V, T3Q)); Chris@42: T47 = LDW(&(W[TWVL * 6])); Chris@42: T4a = VZMUL(T47, VADD(T48, T49)); Chris@42: T3a = VADD(T1t, T1n); Chris@42: T3b = VADD(TU, T1f); Chris@42: T3c = VBYI(VADD(T3a, T3b)); Chris@42: T3n = VBYI(VSUB(T3b, T3a)); Chris@42: T3d = VADD(Tf, Ty); Chris@42: T3e = VADD(T1i, T1j); Chris@42: T3f = VADD(T3d, T3e); Chris@42: T3m = VSUB(T3d, T3e); Chris@42: T39 = LDW(&(W[TWVL * 4])); Chris@42: T3g = VZMULI(T39, VADD(T3c, T3f)); Chris@42: T4d = LDW(&(W[TWVL * 56])); Chris@42: T4e = VZMULI(T4d, VSUB(T3f, T3c)); Chris@42: T3l = LDW(&(W[TWVL * 36])); Chris@42: T3o = VZMULI(T3l, VSUB(T3m, T3n)); Chris@42: T3D = LDW(&(W[TWVL * 24])); Chris@42: T3E = VZMULI(T3D, VADD(T3n, T3m)); Chris@42: Tz = VSUB(Tf, Ty); Chris@42: T1g = VSUB(TU, T1f); Chris@42: T1h = VSUB(Tz, T1g); Chris@42: T2K = VADD(Tz, T1g); Chris@42: T1k = VSUB(T1i, T1j); Chris@42: T1u = VSUB(T1n, T1t); Chris@42: T1v = VBYI(VSUB(T1k, T1u)); Chris@42: T2L = VBYI(VADD(T1u, T1k)); Chris@42: T1 = LDW(&(W[TWVL * 20])); Chris@42: T1w = VZMULI(T1, VADD(T1h, T1v)); Chris@42: T45 = LDW(&(W[TWVL * 8])); Chris@42: T46 = VZMULI(T45, VADD(T2K, T2L)); Chris@42: T2J = LDW(&(W[TWVL * 52])); Chris@42: T2M = VZMULI(T2J, VSUB(T2K, T2L)); Chris@42: T3Z = LDW(&(W[TWVL * 40])); Chris@42: T40 = VZMULI(T3Z, VSUB(T1h, T1v)); Chris@42: T2e = VSUB(T2c, T2d); Chris@42: T2l = VSUB(T2h, T2k); Chris@42: T2m = VSUB(T2e, T2l); Chris@42: T2A = VADD(T2e, T2l); Chris@42: T2p = VSUB(T2n, T2o); Chris@42: T2s = VSUB(T2q, T2r); Chris@42: T2t = VBYI(VSUB(T2p, T2s)); Chris@42: T2B = VBYI(VADD(T2s, T2p)); Chris@42: T2b = LDW(&(W[TWVL * 44])); Chris@42: T2u = VZMULI(T2b, VSUB(T2m, T2t)); Chris@42: T4v = LDW(&(W[TWVL * 16])); Chris@42: T4w = VZMULI(T4v, VADD(T2m, T2t)); Chris@42: T2z = LDW(&(W[TWVL * 12])); Chris@42: T2C = VZMULI(T2z, VADD(T2A, T2B)); Chris@42: T4j = LDW(&(W[TWVL * 48])); Chris@42: T4k = VZMULI(T4j, VSUB(T2A, T2B)); Chris@42: { Chris@42: V T32, T3q, T35, T3r, T30, T31, T33, T34, T2Z, T3z, T3h, T3p, T1Y, T2E, T27; Chris@42: V T2F, T1M, T1X, T23, T26, T1x, T2N, T2v, T2D, T2U, T3x, T2X, T3w, T2S, T2T; Chris@42: V T2V, T2W, T2R, T4J, T3v, T4B; Chris@42: T30 = VADD(T21, T22); Chris@42: T31 = VADD(T1E, T1L); Chris@42: T32 = VADD(T30, T31); Chris@42: T3q = VSUB(T30, T31); Chris@42: T33 = VADD(T1W, T1T); Chris@42: T34 = VADD(T24, T25); Chris@42: T35 = VBYI(VADD(T33, T34)); Chris@42: T3r = VBYI(VSUB(T34, T33)); Chris@42: T2Z = LDW(&(W[TWVL * 58])); Chris@42: T36 = VZMUL(T2Z, VSUB(T32, T35)); Chris@42: T3z = LDW(&(W[TWVL * 26])); Chris@42: T3A = VZMUL(T3z, VADD(T3q, T3r)); Chris@42: T3h = LDW(&(W[TWVL * 2])); Chris@42: T3i = VZMUL(T3h, VADD(T32, T35)); Chris@42: T3p = LDW(&(W[TWVL * 34])); Chris@42: T3s = VZMUL(T3p, VSUB(T3q, T3r)); Chris@42: T1M = VSUB(T1E, T1L); Chris@42: T1X = VSUB(T1T, T1W); Chris@42: T1Y = VBYI(VSUB(T1M, T1X)); Chris@42: T2E = VBYI(VADD(T1X, T1M)); Chris@42: T23 = VSUB(T21, T22); Chris@42: T26 = VSUB(T24, T25); Chris@42: T27 = VSUB(T23, T26); Chris@42: T2F = VADD(T23, T26); Chris@42: T1x = LDW(&(W[TWVL * 18])); Chris@42: T28 = VZMUL(T1x, VADD(T1Y, T27)); Chris@42: T2N = LDW(&(W[TWVL * 50])); Chris@42: T2O = VZMUL(T2N, VSUB(T2F, T2E)); Chris@42: T2v = LDW(&(W[TWVL * 42])); Chris@42: T2w = VZMUL(T2v, VSUB(T27, T1Y)); Chris@42: T2D = LDW(&(W[TWVL * 10])); Chris@42: T2G = VZMUL(T2D, VADD(T2E, T2F)); Chris@42: T2S = VADD(T2c, T2d); Chris@42: T2T = VADD(T2n, T2o); Chris@42: T2U = VADD(T2S, T2T); Chris@42: T3x = VSUB(T2S, T2T); Chris@42: T2V = VADD(T2r, T2q); Chris@42: T2W = VADD(T2h, T2k); Chris@42: T2X = VBYI(VADD(T2V, T2W)); Chris@42: T3w = VBYI(VSUB(T2W, T2V)); Chris@42: T2R = LDW(&(W[TWVL * 60])); Chris@42: T2Y = VZMULI(T2R, VSUB(T2U, T2X)); Chris@42: T4J = LDW(&(W[0])); Chris@42: T4K = VZMULI(T4J, VADD(T2X, T2U)); Chris@42: T3v = LDW(&(W[TWVL * 28])); Chris@42: T3y = VZMULI(T3v, VADD(T3w, T3x)); Chris@42: T4B = LDW(&(W[TWVL * 32])); Chris@42: T4C = VZMULI(T4B, VSUB(T3x, T3w)); Chris@42: } Chris@42: } Chris@42: { Chris@42: V T29, T4M, T2P, T4t, T4N, T2a, T4u, T2Q, T2x, T4H, T2H, T4z, T4I, T2y, T4A; Chris@42: V T2I, T37, T4h, T3B, T3X, T4i, T38, T3Y, T3C, T3j, T4b, T3t, T43, T4c, T3k; Chris@42: V T44, T3u; Chris@42: T29 = VADD(T1w, T28); Chris@42: ST(&(Rp[WS(rs, 5)]), T29, ms, &(Rp[WS(rs, 1)])); Chris@42: T4M = VADD(T4K, T4L); Chris@42: ST(&(Rp[0]), T4M, ms, &(Rp[0])); Chris@42: T2P = VADD(T2M, T2O); Chris@42: ST(&(Rp[WS(rs, 13)]), T2P, ms, &(Rp[WS(rs, 1)])); Chris@42: T4t = VADD(T4k, T4s); Chris@42: ST(&(Rp[WS(rs, 12)]), T4t, ms, &(Rp[0])); Chris@42: T4N = VCONJ(VSUB(T4L, T4K)); Chris@42: ST(&(Rm[0]), T4N, -ms, &(Rm[0])); Chris@42: T2a = VCONJ(VSUB(T28, T1w)); Chris@42: ST(&(Rm[WS(rs, 5)]), T2a, -ms, &(Rm[WS(rs, 1)])); Chris@42: T4u = VCONJ(VSUB(T4s, T4k)); Chris@42: ST(&(Rm[WS(rs, 12)]), T4u, -ms, &(Rm[0])); Chris@42: T2Q = VCONJ(VSUB(T2O, T2M)); Chris@42: ST(&(Rm[WS(rs, 13)]), T2Q, -ms, &(Rm[WS(rs, 1)])); Chris@42: T2x = VADD(T2u, T2w); Chris@42: ST(&(Rp[WS(rs, 11)]), T2x, ms, &(Rp[WS(rs, 1)])); Chris@42: T4H = VADD(T4C, T4G); Chris@42: ST(&(Rp[WS(rs, 8)]), T4H, ms, &(Rp[0])); Chris@42: T2H = VADD(T2C, T2G); Chris@42: ST(&(Rp[WS(rs, 3)]), T2H, ms, &(Rp[WS(rs, 1)])); Chris@42: T4z = VADD(T4w, T4y); Chris@42: ST(&(Rp[WS(rs, 4)]), T4z, ms, &(Rp[0])); Chris@42: T4I = VCONJ(VSUB(T4G, T4C)); Chris@42: ST(&(Rm[WS(rs, 8)]), T4I, -ms, &(Rm[0])); Chris@42: T2y = VCONJ(VSUB(T2w, T2u)); Chris@42: ST(&(Rm[WS(rs, 11)]), T2y, -ms, &(Rm[WS(rs, 1)])); Chris@42: T4A = VCONJ(VSUB(T4y, T4w)); Chris@42: ST(&(Rm[WS(rs, 4)]), T4A, -ms, &(Rm[0])); Chris@42: T2I = VCONJ(VSUB(T2G, T2C)); Chris@42: ST(&(Rm[WS(rs, 3)]), T2I, -ms, &(Rm[WS(rs, 1)])); Chris@42: T37 = VADD(T2Y, T36); Chris@42: ST(&(Rp[WS(rs, 15)]), T37, ms, &(Rp[WS(rs, 1)])); Chris@42: T4h = VADD(T4e, T4g); Chris@42: ST(&(Rp[WS(rs, 14)]), T4h, ms, &(Rp[0])); Chris@42: T3B = VADD(T3y, T3A); Chris@42: ST(&(Rp[WS(rs, 7)]), T3B, ms, &(Rp[WS(rs, 1)])); Chris@42: T3X = VADD(T3E, T3W); Chris@42: ST(&(Rp[WS(rs, 6)]), T3X, ms, &(Rp[0])); Chris@42: T4i = VCONJ(VSUB(T4g, T4e)); Chris@42: ST(&(Rm[WS(rs, 14)]), T4i, -ms, &(Rm[0])); Chris@42: T38 = VCONJ(VSUB(T36, T2Y)); Chris@42: ST(&(Rm[WS(rs, 15)]), T38, -ms, &(Rm[WS(rs, 1)])); Chris@42: T3Y = VCONJ(VSUB(T3W, T3E)); Chris@42: ST(&(Rm[WS(rs, 6)]), T3Y, -ms, &(Rm[0])); Chris@42: T3C = VCONJ(VSUB(T3A, T3y)); Chris@42: ST(&(Rm[WS(rs, 7)]), T3C, -ms, &(Rm[WS(rs, 1)])); Chris@42: T3j = VADD(T3g, T3i); Chris@42: ST(&(Rp[WS(rs, 1)]), T3j, ms, &(Rp[WS(rs, 1)])); Chris@42: T4b = VADD(T46, T4a); Chris@42: ST(&(Rp[WS(rs, 2)]), T4b, ms, &(Rp[0])); Chris@42: T3t = VADD(T3o, T3s); Chris@42: ST(&(Rp[WS(rs, 9)]), T3t, ms, &(Rp[WS(rs, 1)])); Chris@42: T43 = VADD(T40, T42); Chris@42: ST(&(Rp[WS(rs, 10)]), T43, ms, &(Rp[0])); Chris@42: T4c = VCONJ(VSUB(T4a, T46)); Chris@42: ST(&(Rm[WS(rs, 2)]), T4c, -ms, &(Rm[0])); Chris@42: T3k = VCONJ(VSUB(T3i, T3g)); Chris@42: ST(&(Rm[WS(rs, 1)]), T3k, -ms, &(Rm[WS(rs, 1)])); Chris@42: T44 = VCONJ(VSUB(T42, T40)); Chris@42: ST(&(Rm[WS(rs, 10)]), T44, -ms, &(Rm[0])); Chris@42: T3u = VCONJ(VSUB(T3s, T3o)); Chris@42: ST(&(Rm[WS(rs, 9)]), T3u, -ms, &(Rm[WS(rs, 1)])); Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(1, 1), Chris@42: VTW(1, 2), Chris@42: VTW(1, 3), Chris@42: VTW(1, 4), Chris@42: VTW(1, 5), Chris@42: VTW(1, 6), Chris@42: VTW(1, 7), Chris@42: VTW(1, 8), Chris@42: VTW(1, 9), Chris@42: VTW(1, 10), Chris@42: VTW(1, 11), Chris@42: VTW(1, 12), Chris@42: VTW(1, 13), Chris@42: VTW(1, 14), Chris@42: VTW(1, 15), Chris@42: VTW(1, 16), Chris@42: VTW(1, 17), Chris@42: VTW(1, 18), Chris@42: VTW(1, 19), Chris@42: VTW(1, 20), Chris@42: VTW(1, 21), Chris@42: VTW(1, 22), Chris@42: VTW(1, 23), Chris@42: VTW(1, 24), Chris@42: VTW(1, 25), Chris@42: VTW(1, 26), Chris@42: VTW(1, 27), Chris@42: VTW(1, 28), Chris@42: VTW(1, 29), Chris@42: VTW(1, 30), Chris@42: VTW(1, 31), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 32, XSIMD_STRING("hc2cbdftv_32"), twinstr, &GENUS, {233, 88, 16, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cbdftv_32) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cbdftv_32, &desc, HC2C_VIA_DFT); Chris@42: } Chris@42: #endif /* HAVE_FMA */