Mercurial > hg > sv-dependency-builds
diff src/fftw-3.3.3/dft/scalar/codelets/q1_4.c @ 10:37bf6b4a2645
Add FFTW3
| author | Chris Cannam |
|---|---|
| date | Wed, 20 Mar 2013 15:35:50 +0000 |
| parents | |
| children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/fftw-3.3.3/dft/scalar/codelets/q1_4.c Wed Mar 20 15:35:50 2013 +0000 @@ -0,0 +1,518 @@ +/* + * Copyright (c) 2003, 2007-11 Matteo Frigo + * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + * + */ + +/* This file was automatically generated --- DO NOT EDIT */ +/* Generated on Sun Nov 25 07:36:17 EST 2012 */ + +#include "codelet-dft.h" + +#ifdef HAVE_FMA + +/* Generated by: ../../../genfft/gen_twidsq.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include q.h */ + +/* + * This function contains 88 FP additions, 48 FP multiplications, + * (or, 64 additions, 24 multiplications, 24 fused multiply/add), + * 76 stack variables, 0 constants, and 64 memory accesses + */ +#include "q.h" + +static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms) +{ + { + INT m; + for (m = mb, W = W + (mb * 6); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(0, vs)) { + E T1X, T1S, T1L, T1Y, T1R; + { + E T3, Tf, Tv, Ti, Tw, Tx, T6, Tm, Tc, Ts, T1T, T1H, T29, T1W, T2a; + E T2b, T1K, T20, T1Q, T26, TN, TB, T13, TQ, T14, T15, TE, TU, TK, T10; + E T1l, T19, T1a, T1h, T1B, T1o, T1C, T1b, T1D, T1e, T1c; + { + E T1I, T1P, T1J, T1M; + { + E Tb, T4, T5, T8; + { + E T1, T2, T9, Ta, Tg, Th; + T1 = rio[0]; + T2 = rio[WS(rs, 2)]; + T9 = iio[0]; + Ta = iio[WS(rs, 2)]; + Tg = iio[WS(rs, 1)]; + T3 = T1 + T2; + Tf = T1 - T2; + Th = iio[WS(rs, 3)]; + Tv = T9 + Ta; + Tb = T9 - Ta; + T4 = rio[WS(rs, 1)]; + Ti = Tg - Th; + Tw = Tg + Th; + T5 = rio[WS(rs, 3)]; + } + Tx = Tv - Tw; + T8 = T4 - T5; + T6 = T4 + T5; + { + E T1N, T1O, T1F, T1G, T1U, T1V; + T1F = rio[WS(vs, 3)]; + T1G = rio[WS(vs, 3) + WS(rs, 2)]; + Tm = Tb - T8; + Tc = T8 + Tb; + Ts = T3 - T6; + T1T = T1F - T1G; + T1H = T1F + T1G; + T1N = iio[WS(vs, 3)]; + T1O = iio[WS(vs, 3) + WS(rs, 2)]; + T1U = iio[WS(vs, 3) + WS(rs, 1)]; + T1V = iio[WS(vs, 3) + WS(rs, 3)]; + T1I = rio[WS(vs, 3) + WS(rs, 1)]; + T1P = T1N - T1O; + T29 = T1N + T1O; + T1W = T1U - T1V; + T2a = T1U + T1V; + T1J = rio[WS(vs, 3) + WS(rs, 3)]; + } + } + T2b = T29 - T2a; + T1M = T1I - T1J; + T1K = T1I + T1J; + { + E TC, TJ, TD, TG; + { + E TH, TI, Tz, TA, TO, TP; + Tz = rio[WS(vs, 1)]; + TA = rio[WS(vs, 1) + WS(rs, 2)]; + T20 = T1P - T1M; + T1Q = T1M + T1P; + T26 = T1H - T1K; + TN = Tz - TA; + TB = Tz + TA; + TH = iio[WS(vs, 1)]; + TI = iio[WS(vs, 1) + WS(rs, 2)]; + TO = iio[WS(vs, 1) + WS(rs, 1)]; + TP = iio[WS(vs, 1) + WS(rs, 3)]; + TC = rio[WS(vs, 1) + WS(rs, 1)]; + TJ = TH - TI; + T13 = TH + TI; + TQ = TO - TP; + T14 = TO + TP; + TD = rio[WS(vs, 1) + WS(rs, 3)]; + } + T15 = T13 - T14; + TG = TC - TD; + TE = TC + TD; + { + E T1f, T1g, T17, T18, T1m, T1n; + T17 = rio[WS(vs, 2)]; + T18 = rio[WS(vs, 2) + WS(rs, 2)]; + TU = TJ - TG; + TK = TG + TJ; + T10 = TB - TE; + T1l = T17 - T18; + T19 = T17 + T18; + T1f = iio[WS(vs, 2)]; + T1g = iio[WS(vs, 2) + WS(rs, 2)]; + T1m = iio[WS(vs, 2) + WS(rs, 1)]; + T1n = iio[WS(vs, 2) + WS(rs, 3)]; + T1a = rio[WS(vs, 2) + WS(rs, 1)]; + T1h = T1f - T1g; + T1B = T1f + T1g; + T1o = T1m - T1n; + T1C = T1m + T1n; + T1b = rio[WS(vs, 2) + WS(rs, 3)]; + } + } + } + T1D = T1B - T1C; + T1e = T1a - T1b; + T1c = T1a + T1b; + { + E T1s, T1i, T1y, T28, T27, Tr, Tu; + rio[0] = T3 + T6; + iio[0] = Tv + Tw; + T1s = T1h - T1e; + T1i = T1e + T1h; + T1y = T19 - T1c; + rio[WS(rs, 1)] = TB + TE; + iio[WS(rs, 1)] = T13 + T14; + rio[WS(rs, 2)] = T19 + T1c; + iio[WS(rs, 2)] = T1B + T1C; + iio[WS(rs, 3)] = T29 + T2a; + rio[WS(rs, 3)] = T1H + T1K; + Tr = W[2]; + Tu = W[3]; + { + E T25, Ty, Tt, T2c; + T25 = W[2]; + T28 = W[3]; + Ty = Tr * Tx; + Tt = Tr * Ts; + T2c = T25 * T2b; + T27 = T25 * T26; + iio[WS(vs, 2)] = FNMS(Tu, Ts, Ty); + rio[WS(vs, 2)] = FMA(Tu, Tx, Tt); + iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T28, T26, T2c); + } + rio[WS(vs, 2) + WS(rs, 3)] = FMA(T28, T2b, T27); + { + E Tp, T1v, T23, T22, T1Z, TR, TM, TF; + { + E T1A, T1z, TZ, T12; + TZ = W[2]; + T12 = W[3]; + { + E T1x, T16, T11, T1E; + T1x = W[2]; + T1A = W[3]; + T16 = TZ * T15; + T11 = TZ * T10; + T1E = T1x * T1D; + T1z = T1x * T1y; + iio[WS(vs, 2) + WS(rs, 1)] = FNMS(T12, T10, T16); + rio[WS(vs, 2) + WS(rs, 1)] = FMA(T12, T15, T11); + iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1A, T1y, T1E); + } + rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1A, T1D, T1z); + { + E Tj, Te, T7, T1p, T1k, T1j; + Tp = Tf + Ti; + Tj = Tf - Ti; + Te = W[5]; + T7 = W[4]; + { + E T1d, T1q, Tk, Td; + T1p = T1l - T1o; + T1v = T1l + T1o; + T1k = W[5]; + Tk = Te * Tc; + Td = T7 * Tc; + T1d = W[4]; + T1q = T1k * T1i; + rio[WS(vs, 3)] = FMA(T7, Tj, Tk); + iio[WS(vs, 3)] = FNMS(Te, Tj, Td); + T1j = T1d * T1i; + rio[WS(vs, 3) + WS(rs, 2)] = FMA(T1d, T1p, T1q); + } + T23 = T1T + T1W; + T1X = T1T - T1W; + T22 = W[1]; + iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T1k, T1p, T1j); + T1Z = W[0]; + } + } + { + E TX, TW, TT, TY, TV, T24, T21; + TX = TN + TQ; + TR = TN - TQ; + T24 = T22 * T20; + TW = W[1]; + T21 = T1Z * T20; + TT = W[0]; + rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1Z, T23, T24); + TY = TW * TU; + iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T22, T23, T21); + TV = TT * TU; + rio[WS(vs, 1) + WS(rs, 1)] = FMA(TT, TX, TY); + TM = W[5]; + iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TW, TX, TV); + TF = W[4]; + } + { + E To, Tl, Tq, Tn, TS, TL; + TS = TM * TK; + To = W[1]; + TL = TF * TK; + Tl = W[0]; + rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TR, TS); + Tq = To * Tm; + iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TM, TR, TL); + Tn = Tl * Tm; + { + E T1u, T1r, T1w, T1t; + rio[WS(vs, 1)] = FMA(Tl, Tp, Tq); + T1u = W[1]; + iio[WS(vs, 1)] = FNMS(To, Tp, Tn); + T1r = W[0]; + T1w = T1u * T1s; + T1S = W[5]; + T1t = T1r * T1s; + T1L = W[4]; + rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1r, T1v, T1w); + T1Y = T1S * T1Q; + iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1u, T1v, T1t); + T1R = T1L * T1Q; + } + } + } + } + } + rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1L, T1X, T1Y); + iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1S, T1X, T1R); + } + } +} + +static const tw_instr twinstr[] = { + {TW_FULL, 0, 4}, + {TW_NEXT, 1, 0} +}; + +static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 }; + +void X(codelet_q1_4) (planner *p) { + X(kdft_difsq_register) (p, q1_4, &desc); +} +#else /* HAVE_FMA */ + +/* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include q.h */ + +/* + * This function contains 88 FP additions, 48 FP multiplications, + * (or, 64 additions, 24 multiplications, 24 fused multiply/add), + * 37 stack variables, 0 constants, and 64 memory accesses + */ +#include "q.h" + +static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms) +{ + { + INT m; + for (m = mb, W = W + (mb * 6); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(0, vs)) { + E T3, Te, Tb, Tq, T6, T8, Th, Tr, Tv, TG, TD, TS, Ty, TA, TJ; + E TT, TX, T18, T15, T1k, T10, T12, T1b, T1l, T1p, T1A, T1x, T1M, T1s, T1u; + E T1D, T1N; + { + E T1, T2, T9, Ta; + T1 = rio[0]; + T2 = rio[WS(rs, 2)]; + T3 = T1 + T2; + Te = T1 - T2; + T9 = iio[0]; + Ta = iio[WS(rs, 2)]; + Tb = T9 - Ta; + Tq = T9 + Ta; + } + { + E T4, T5, Tf, Tg; + T4 = rio[WS(rs, 1)]; + T5 = rio[WS(rs, 3)]; + T6 = T4 + T5; + T8 = T4 - T5; + Tf = iio[WS(rs, 1)]; + Tg = iio[WS(rs, 3)]; + Th = Tf - Tg; + Tr = Tf + Tg; + } + { + E Tt, Tu, TB, TC; + Tt = rio[WS(vs, 1)]; + Tu = rio[WS(vs, 1) + WS(rs, 2)]; + Tv = Tt + Tu; + TG = Tt - Tu; + TB = iio[WS(vs, 1)]; + TC = iio[WS(vs, 1) + WS(rs, 2)]; + TD = TB - TC; + TS = TB + TC; + } + { + E Tw, Tx, TH, TI; + Tw = rio[WS(vs, 1) + WS(rs, 1)]; + Tx = rio[WS(vs, 1) + WS(rs, 3)]; + Ty = Tw + Tx; + TA = Tw - Tx; + TH = iio[WS(vs, 1) + WS(rs, 1)]; + TI = iio[WS(vs, 1) + WS(rs, 3)]; + TJ = TH - TI; + TT = TH + TI; + } + { + E TV, TW, T13, T14; + TV = rio[WS(vs, 2)]; + TW = rio[WS(vs, 2) + WS(rs, 2)]; + TX = TV + TW; + T18 = TV - TW; + T13 = iio[WS(vs, 2)]; + T14 = iio[WS(vs, 2) + WS(rs, 2)]; + T15 = T13 - T14; + T1k = T13 + T14; + } + { + E TY, TZ, T19, T1a; + TY = rio[WS(vs, 2) + WS(rs, 1)]; + TZ = rio[WS(vs, 2) + WS(rs, 3)]; + T10 = TY + TZ; + T12 = TY - TZ; + T19 = iio[WS(vs, 2) + WS(rs, 1)]; + T1a = iio[WS(vs, 2) + WS(rs, 3)]; + T1b = T19 - T1a; + T1l = T19 + T1a; + } + { + E T1n, T1o, T1v, T1w; + T1n = rio[WS(vs, 3)]; + T1o = rio[WS(vs, 3) + WS(rs, 2)]; + T1p = T1n + T1o; + T1A = T1n - T1o; + T1v = iio[WS(vs, 3)]; + T1w = iio[WS(vs, 3) + WS(rs, 2)]; + T1x = T1v - T1w; + T1M = T1v + T1w; + } + { + E T1q, T1r, T1B, T1C; + T1q = rio[WS(vs, 3) + WS(rs, 1)]; + T1r = rio[WS(vs, 3) + WS(rs, 3)]; + T1s = T1q + T1r; + T1u = T1q - T1r; + T1B = iio[WS(vs, 3) + WS(rs, 1)]; + T1C = iio[WS(vs, 3) + WS(rs, 3)]; + T1D = T1B - T1C; + T1N = T1B + T1C; + } + rio[0] = T3 + T6; + iio[0] = Tq + Tr; + rio[WS(rs, 1)] = Tv + Ty; + iio[WS(rs, 1)] = TS + TT; + rio[WS(rs, 2)] = TX + T10; + iio[WS(rs, 2)] = T1k + T1l; + iio[WS(rs, 3)] = T1M + T1N; + rio[WS(rs, 3)] = T1p + T1s; + { + E Tc, Ti, T7, Td; + Tc = T8 + Tb; + Ti = Te - Th; + T7 = W[4]; + Td = W[5]; + iio[WS(vs, 3)] = FNMS(Td, Ti, T7 * Tc); + rio[WS(vs, 3)] = FMA(Td, Tc, T7 * Ti); + } + { + E T1K, T1O, T1J, T1L; + T1K = T1p - T1s; + T1O = T1M - T1N; + T1J = W[2]; + T1L = W[3]; + rio[WS(vs, 2) + WS(rs, 3)] = FMA(T1J, T1K, T1L * T1O); + iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T1L, T1K, T1J * T1O); + } + { + E Tk, Tm, Tj, Tl; + Tk = Tb - T8; + Tm = Te + Th; + Tj = W[0]; + Tl = W[1]; + iio[WS(vs, 1)] = FNMS(Tl, Tm, Tj * Tk); + rio[WS(vs, 1)] = FMA(Tl, Tk, Tj * Tm); + } + { + E To, Ts, Tn, Tp; + To = T3 - T6; + Ts = Tq - Tr; + Tn = W[2]; + Tp = W[3]; + rio[WS(vs, 2)] = FMA(Tn, To, Tp * Ts); + iio[WS(vs, 2)] = FNMS(Tp, To, Tn * Ts); + } + { + E T16, T1c, T11, T17; + T16 = T12 + T15; + T1c = T18 - T1b; + T11 = W[4]; + T17 = W[5]; + iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T17, T1c, T11 * T16); + rio[WS(vs, 3) + WS(rs, 2)] = FMA(T17, T16, T11 * T1c); + } + { + E T1G, T1I, T1F, T1H; + T1G = T1x - T1u; + T1I = T1A + T1D; + T1F = W[0]; + T1H = W[1]; + iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T1H, T1I, T1F * T1G); + rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1H, T1G, T1F * T1I); + } + { + E TQ, TU, TP, TR; + TQ = Tv - Ty; + TU = TS - TT; + TP = W[2]; + TR = W[3]; + rio[WS(vs, 2) + WS(rs, 1)] = FMA(TP, TQ, TR * TU); + iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TR, TQ, TP * TU); + } + { + E T1e, T1g, T1d, T1f; + T1e = T15 - T12; + T1g = T18 + T1b; + T1d = W[0]; + T1f = W[1]; + iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1f, T1g, T1d * T1e); + rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1f, T1e, T1d * T1g); + } + { + E T1i, T1m, T1h, T1j; + T1i = TX - T10; + T1m = T1k - T1l; + T1h = W[2]; + T1j = W[3]; + rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1h, T1i, T1j * T1m); + iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1j, T1i, T1h * T1m); + } + { + E T1y, T1E, T1t, T1z; + T1y = T1u + T1x; + T1E = T1A - T1D; + T1t = W[4]; + T1z = W[5]; + iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1z, T1E, T1t * T1y); + rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1z, T1y, T1t * T1E); + } + { + E TM, TO, TL, TN; + TM = TD - TA; + TO = TG + TJ; + TL = W[0]; + TN = W[1]; + iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TN, TO, TL * TM); + rio[WS(vs, 1) + WS(rs, 1)] = FMA(TN, TM, TL * TO); + } + { + E TE, TK, Tz, TF; + TE = TA + TD; + TK = TG - TJ; + Tz = W[4]; + TF = W[5]; + iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TF, TK, Tz * TE); + rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TE, Tz * TK); + } + } + } +} + +static const tw_instr twinstr[] = { + {TW_FULL, 0, 4}, + {TW_NEXT, 1, 0} +}; + +static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 }; + +void X(codelet_q1_4) (planner *p) { + X(kdft_difsq_register) (p, q1_4, &desc); +} +#endif /* HAVE_FMA */