cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:07:52 EDT 2018 */ cannam@167: cannam@167: #include "rdft/codelet-rdft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -dif -name hc2cb_10 -include rdft/scalar/hc2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 102 FP additions, 72 FP multiplications, cannam@167: * (or, 48 additions, 18 multiplications, 54 fused multiply/add), cannam@167: * 47 stack variables, 4 constants, and 40 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cb.h" cannam@167: cannam@167: static void hc2cb_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 18, MAKE_VOLATILE_STRIDE(40, rs)) { cannam@167: E TH, T1B, TB, T11, T1E, T1G, TK, TM, T1x, T1V, T3, T1g, Tl, T1I, T1J; cannam@167: E TO, TP, T1p, Ti, Tk, T1n, T1o, TF, TG; cannam@167: TF = Ip[0]; cannam@167: TG = Im[WS(rs, 4)]; cannam@167: TH = TF - TG; cannam@167: T1B = TF + TG; cannam@167: { cannam@167: E Tp, T1u, Tz, T1s, Ts, T1v, Tw, T1r; cannam@167: { cannam@167: E Tn, To, Tx, Ty; cannam@167: Tn = Ip[WS(rs, 4)]; cannam@167: To = Im[0]; cannam@167: Tp = Tn - To; cannam@167: T1u = Tn + To; cannam@167: Tx = Ip[WS(rs, 3)]; cannam@167: Ty = Im[WS(rs, 1)]; cannam@167: Tz = Tx - Ty; cannam@167: T1s = Tx + Ty; cannam@167: } cannam@167: { cannam@167: E Tq, Tr, Tu, Tv; cannam@167: Tq = Ip[WS(rs, 1)]; cannam@167: Tr = Im[WS(rs, 3)]; cannam@167: Ts = Tq - Tr; cannam@167: T1v = Tq + Tr; cannam@167: Tu = Ip[WS(rs, 2)]; cannam@167: Tv = Im[WS(rs, 2)]; cannam@167: Tw = Tu - Tv; cannam@167: T1r = Tu + Tv; cannam@167: } cannam@167: { cannam@167: E Tt, TA, T1C, T1D; cannam@167: Tt = Tp - Ts; cannam@167: TA = Tw - Tz; cannam@167: TB = FNMS(KP618033988, TA, Tt); cannam@167: T11 = FMA(KP618033988, Tt, TA); cannam@167: T1C = T1r - T1s; cannam@167: T1D = T1u - T1v; cannam@167: T1E = T1C + T1D; cannam@167: T1G = T1C - T1D; cannam@167: } cannam@167: { cannam@167: E TI, TJ, T1t, T1w; cannam@167: TI = Tw + Tz; cannam@167: TJ = Tp + Ts; cannam@167: TK = TI + TJ; cannam@167: TM = TI - TJ; cannam@167: T1t = T1r + T1s; cannam@167: T1w = T1u + T1v; cannam@167: T1x = FMA(KP618033988, T1w, T1t); cannam@167: T1V = FNMS(KP618033988, T1t, T1w); cannam@167: } cannam@167: } cannam@167: { cannam@167: E Td, T1k, Tg, T1l, Th, T1m, T6, T1h, T9, T1i, Ta, T1j, T1, T2; cannam@167: T1 = Rp[0]; cannam@167: T2 = Rm[WS(rs, 4)]; cannam@167: T3 = T1 + T2; cannam@167: T1g = T1 - T2; cannam@167: { cannam@167: E Tb, Tc, Te, Tf; cannam@167: Tb = Rp[WS(rs, 4)]; cannam@167: Tc = Rm[0]; cannam@167: Td = Tb + Tc; cannam@167: T1k = Tb - Tc; cannam@167: Te = Rm[WS(rs, 3)]; cannam@167: Tf = Rp[WS(rs, 1)]; cannam@167: Tg = Te + Tf; cannam@167: T1l = Te - Tf; cannam@167: } cannam@167: Th = Td + Tg; cannam@167: T1m = T1k + T1l; cannam@167: { cannam@167: E T4, T5, T7, T8; cannam@167: T4 = Rp[WS(rs, 2)]; cannam@167: T5 = Rm[WS(rs, 2)]; cannam@167: T6 = T4 + T5; cannam@167: T1h = T4 - T5; cannam@167: T7 = Rm[WS(rs, 1)]; cannam@167: T8 = Rp[WS(rs, 3)]; cannam@167: T9 = T7 + T8; cannam@167: T1i = T7 - T8; cannam@167: } cannam@167: Ta = T6 + T9; cannam@167: T1j = T1h + T1i; cannam@167: Tl = Ta - Th; cannam@167: T1I = T1h - T1i; cannam@167: T1J = T1k - T1l; cannam@167: TO = Td - Tg; cannam@167: TP = T6 - T9; cannam@167: T1p = T1j - T1m; cannam@167: Ti = Ta + Th; cannam@167: Tk = FNMS(KP250000000, Ti, T3); cannam@167: T1n = T1j + T1m; cannam@167: T1o = FNMS(KP250000000, T1n, T1g); cannam@167: } cannam@167: Rp[0] = T3 + Ti; cannam@167: Rm[0] = TH + TK; cannam@167: { cannam@167: E T2d, T29, T2b, T2c, T2e, T2a; cannam@167: T2d = T1B + T1E; cannam@167: T2a = T1g + T1n; cannam@167: T29 = W[8]; cannam@167: T2b = T29 * T2a; cannam@167: T2c = W[9]; cannam@167: T2e = T2c * T2a; cannam@167: Ip[WS(rs, 2)] = FNMS(T2c, T2d, T2b); cannam@167: Im[WS(rs, 2)] = FMA(T29, T2d, T2e); cannam@167: } cannam@167: { cannam@167: E TQ, T16, TC, TU, TN, T15, T12, T1a, Tm, TL, T10; cannam@167: TQ = FNMS(KP618033988, TP, TO); cannam@167: T16 = FMA(KP618033988, TO, TP); cannam@167: Tm = FNMS(KP559016994, Tl, Tk); cannam@167: TC = FMA(KP951056516, TB, Tm); cannam@167: TU = FNMS(KP951056516, TB, Tm); cannam@167: TL = FNMS(KP250000000, TK, TH); cannam@167: TN = FNMS(KP559016994, TM, TL); cannam@167: T15 = FMA(KP559016994, TM, TL); cannam@167: T10 = FMA(KP559016994, Tl, Tk); cannam@167: T12 = FMA(KP951056516, T11, T10); cannam@167: T1a = FNMS(KP951056516, T11, T10); cannam@167: { cannam@167: E TR, TE, TS, Tj, TD; cannam@167: TR = FNMS(KP951056516, TQ, TN); cannam@167: TE = W[3]; cannam@167: TS = TE * TC; cannam@167: Tj = W[2]; cannam@167: TD = Tj * TC; cannam@167: Rp[WS(rs, 1)] = FNMS(TE, TR, TD); cannam@167: Rm[WS(rs, 1)] = FMA(Tj, TR, TS); cannam@167: } cannam@167: { cannam@167: E T1d, T1c, T1e, T19, T1b; cannam@167: T1d = FMA(KP951056516, T16, T15); cannam@167: T1c = W[11]; cannam@167: T1e = T1c * T1a; cannam@167: T19 = W[10]; cannam@167: T1b = T19 * T1a; cannam@167: Rp[WS(rs, 3)] = FNMS(T1c, T1d, T1b); cannam@167: Rm[WS(rs, 3)] = FMA(T19, T1d, T1e); cannam@167: } cannam@167: { cannam@167: E TX, TW, TY, TT, TV; cannam@167: TX = FMA(KP951056516, TQ, TN); cannam@167: TW = W[15]; cannam@167: TY = TW * TU; cannam@167: TT = W[14]; cannam@167: TV = TT * TU; cannam@167: Rp[WS(rs, 4)] = FNMS(TW, TX, TV); cannam@167: Rm[WS(rs, 4)] = FMA(TT, TX, TY); cannam@167: } cannam@167: { cannam@167: E T17, T14, T18, TZ, T13; cannam@167: T17 = FNMS(KP951056516, T16, T15); cannam@167: T14 = W[7]; cannam@167: T18 = T14 * T12; cannam@167: TZ = W[6]; cannam@167: T13 = TZ * T12; cannam@167: Rp[WS(rs, 2)] = FNMS(T14, T17, T13); cannam@167: Rm[WS(rs, 2)] = FMA(TZ, T17, T18); cannam@167: } cannam@167: } cannam@167: { cannam@167: E T1K, T20, T1y, T1O, T1H, T1Z, T1W, T24, T1q, T1F, T1U; cannam@167: T1K = FMA(KP618033988, T1J, T1I); cannam@167: T20 = FNMS(KP618033988, T1I, T1J); cannam@167: T1q = FMA(KP559016994, T1p, T1o); cannam@167: T1y = FNMS(KP951056516, T1x, T1q); cannam@167: T1O = FMA(KP951056516, T1x, T1q); cannam@167: T1F = FNMS(KP250000000, T1E, T1B); cannam@167: T1H = FMA(KP559016994, T1G, T1F); cannam@167: T1Z = FNMS(KP559016994, T1G, T1F); cannam@167: T1U = FNMS(KP559016994, T1p, T1o); cannam@167: T1W = FNMS(KP951056516, T1V, T1U); cannam@167: T24 = FMA(KP951056516, T1V, T1U); cannam@167: { cannam@167: E T1L, T1A, T1M, T1f, T1z; cannam@167: T1L = FMA(KP951056516, T1K, T1H); cannam@167: T1A = W[1]; cannam@167: T1M = T1A * T1y; cannam@167: T1f = W[0]; cannam@167: T1z = T1f * T1y; cannam@167: Ip[0] = FNMS(T1A, T1L, T1z); cannam@167: Im[0] = FMA(T1f, T1L, T1M); cannam@167: } cannam@167: { cannam@167: E T27, T26, T28, T23, T25; cannam@167: T27 = FNMS(KP951056516, T20, T1Z); cannam@167: T26 = W[13]; cannam@167: T28 = T26 * T24; cannam@167: T23 = W[12]; cannam@167: T25 = T23 * T24; cannam@167: Ip[WS(rs, 3)] = FNMS(T26, T27, T25); cannam@167: Im[WS(rs, 3)] = FMA(T23, T27, T28); cannam@167: } cannam@167: { cannam@167: E T1R, T1Q, T1S, T1N, T1P; cannam@167: T1R = FNMS(KP951056516, T1K, T1H); cannam@167: T1Q = W[17]; cannam@167: T1S = T1Q * T1O; cannam@167: T1N = W[16]; cannam@167: T1P = T1N * T1O; cannam@167: Ip[WS(rs, 4)] = FNMS(T1Q, T1R, T1P); cannam@167: Im[WS(rs, 4)] = FMA(T1N, T1R, T1S); cannam@167: } cannam@167: { cannam@167: E T21, T1Y, T22, T1T, T1X; cannam@167: T21 = FMA(KP951056516, T20, T1Z); cannam@167: T1Y = W[5]; cannam@167: T22 = T1Y * T1W; cannam@167: T1T = W[4]; cannam@167: T1X = T1T * T1W; cannam@167: Ip[WS(rs, 1)] = FNMS(T1Y, T21, T1X); cannam@167: Im[WS(rs, 1)] = FMA(T1T, T21, T22); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 10}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 10, "hc2cb_10", twinstr, &GENUS, {48, 18, 54, 0} }; cannam@167: cannam@167: void X(codelet_hc2cb_10) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cb_10, &desc, HC2C_VIA_RDFT); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -dif -name hc2cb_10 -include rdft/scalar/hc2cb.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 102 FP additions, 60 FP multiplications, cannam@167: * (or, 72 additions, 30 multiplications, 30 fused multiply/add), cannam@167: * 39 stack variables, 4 constants, and 40 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/hc2cb.h" cannam@167: cannam@167: static void hc2cb_10(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@167: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 18, MAKE_VOLATILE_STRIDE(40, rs)) { cannam@167: E T3, T18, TJ, T1i, TE, TF, T1B, T1A, T1f, T1t, Ti, Tl, Tt, TA, T1w; cannam@167: E T1v, T1p, T1E, TM, TO; cannam@167: { cannam@167: E T1, T2, TH, TI; cannam@167: T1 = Rp[0]; cannam@167: T2 = Rm[WS(rs, 4)]; cannam@167: T3 = T1 + T2; cannam@167: T18 = T1 - T2; cannam@167: TH = Ip[0]; cannam@167: TI = Im[WS(rs, 4)]; cannam@167: TJ = TH - TI; cannam@167: T1i = TH + TI; cannam@167: } cannam@167: { cannam@167: E T6, T19, Tg, T1d, T9, T1a, Td, T1c; cannam@167: { cannam@167: E T4, T5, Te, Tf; cannam@167: T4 = Rp[WS(rs, 2)]; cannam@167: T5 = Rm[WS(rs, 2)]; cannam@167: T6 = T4 + T5; cannam@167: T19 = T4 - T5; cannam@167: Te = Rm[WS(rs, 3)]; cannam@167: Tf = Rp[WS(rs, 1)]; cannam@167: Tg = Te + Tf; cannam@167: T1d = Te - Tf; cannam@167: } cannam@167: { cannam@167: E T7, T8, Tb, Tc; cannam@167: T7 = Rm[WS(rs, 1)]; cannam@167: T8 = Rp[WS(rs, 3)]; cannam@167: T9 = T7 + T8; cannam@167: T1a = T7 - T8; cannam@167: Tb = Rp[WS(rs, 4)]; cannam@167: Tc = Rm[0]; cannam@167: Td = Tb + Tc; cannam@167: T1c = Tb - Tc; cannam@167: } cannam@167: TE = T6 - T9; cannam@167: TF = Td - Tg; cannam@167: T1B = T1c - T1d; cannam@167: T1A = T19 - T1a; cannam@167: { cannam@167: E T1b, T1e, Ta, Th; cannam@167: T1b = T19 + T1a; cannam@167: T1e = T1c + T1d; cannam@167: T1f = T1b + T1e; cannam@167: T1t = KP559016994 * (T1b - T1e); cannam@167: Ta = T6 + T9; cannam@167: Th = Td + Tg; cannam@167: Ti = Ta + Th; cannam@167: Tl = KP559016994 * (Ta - Th); cannam@167: } cannam@167: } cannam@167: { cannam@167: E Tp, T1j, Tz, T1n, Ts, T1k, Tw, T1m; cannam@167: { cannam@167: E Tn, To, Tx, Ty; cannam@167: Tn = Ip[WS(rs, 2)]; cannam@167: To = Im[WS(rs, 2)]; cannam@167: Tp = Tn - To; cannam@167: T1j = Tn + To; cannam@167: Tx = Ip[WS(rs, 1)]; cannam@167: Ty = Im[WS(rs, 3)]; cannam@167: Tz = Tx - Ty; cannam@167: T1n = Tx + Ty; cannam@167: } cannam@167: { cannam@167: E Tq, Tr, Tu, Tv; cannam@167: Tq = Ip[WS(rs, 3)]; cannam@167: Tr = Im[WS(rs, 1)]; cannam@167: Ts = Tq - Tr; cannam@167: T1k = Tq + Tr; cannam@167: Tu = Ip[WS(rs, 4)]; cannam@167: Tv = Im[0]; cannam@167: Tw = Tu - Tv; cannam@167: T1m = Tu + Tv; cannam@167: } cannam@167: Tt = Tp - Ts; cannam@167: TA = Tw - Tz; cannam@167: T1w = T1m + T1n; cannam@167: T1v = T1j + T1k; cannam@167: { cannam@167: E T1l, T1o, TK, TL; cannam@167: T1l = T1j - T1k; cannam@167: T1o = T1m - T1n; cannam@167: T1p = T1l + T1o; cannam@167: T1E = KP559016994 * (T1l - T1o); cannam@167: TK = Tp + Ts; cannam@167: TL = Tw + Tz; cannam@167: TM = TK + TL; cannam@167: TO = KP559016994 * (TK - TL); cannam@167: } cannam@167: } cannam@167: Rp[0] = T3 + Ti; cannam@167: Rm[0] = TJ + TM; cannam@167: { cannam@167: E T1g, T1q, T17, T1h; cannam@167: T1g = T18 + T1f; cannam@167: T1q = T1i + T1p; cannam@167: T17 = W[8]; cannam@167: T1h = W[9]; cannam@167: Ip[WS(rs, 2)] = FNMS(T1h, T1q, T17 * T1g); cannam@167: Im[WS(rs, 2)] = FMA(T1h, T1g, T17 * T1q); cannam@167: } cannam@167: { cannam@167: E TB, TG, T11, TX, TP, T10, Tm, TW, TN, Tk; cannam@167: TB = FNMS(KP951056516, TA, KP587785252 * Tt); cannam@167: TG = FNMS(KP951056516, TF, KP587785252 * TE); cannam@167: T11 = FMA(KP951056516, TE, KP587785252 * TF); cannam@167: TX = FMA(KP951056516, Tt, KP587785252 * TA); cannam@167: TN = FNMS(KP250000000, TM, TJ); cannam@167: TP = TN - TO; cannam@167: T10 = TO + TN; cannam@167: Tk = FNMS(KP250000000, Ti, T3); cannam@167: Tm = Tk - Tl; cannam@167: TW = Tl + Tk; cannam@167: { cannam@167: E TC, TQ, Tj, TD; cannam@167: TC = Tm - TB; cannam@167: TQ = TG + TP; cannam@167: Tj = W[2]; cannam@167: TD = W[3]; cannam@167: Rp[WS(rs, 1)] = FNMS(TD, TQ, Tj * TC); cannam@167: Rm[WS(rs, 1)] = FMA(TD, TC, Tj * TQ); cannam@167: } cannam@167: { cannam@167: E T14, T16, T13, T15; cannam@167: T14 = TW - TX; cannam@167: T16 = T11 + T10; cannam@167: T13 = W[10]; cannam@167: T15 = W[11]; cannam@167: Rp[WS(rs, 3)] = FNMS(T15, T16, T13 * T14); cannam@167: Rm[WS(rs, 3)] = FMA(T15, T14, T13 * T16); cannam@167: } cannam@167: { cannam@167: E TS, TU, TR, TT; cannam@167: TS = Tm + TB; cannam@167: TU = TP - TG; cannam@167: TR = W[14]; cannam@167: TT = W[15]; cannam@167: Rp[WS(rs, 4)] = FNMS(TT, TU, TR * TS); cannam@167: Rm[WS(rs, 4)] = FMA(TT, TS, TR * TU); cannam@167: } cannam@167: { cannam@167: E TY, T12, TV, TZ; cannam@167: TY = TW + TX; cannam@167: T12 = T10 - T11; cannam@167: TV = W[6]; cannam@167: TZ = W[7]; cannam@167: Rp[WS(rs, 2)] = FNMS(TZ, T12, TV * TY); cannam@167: Rm[WS(rs, 2)] = FMA(TZ, TY, TV * T12); cannam@167: } cannam@167: } cannam@167: { cannam@167: E T1x, T1C, T1Q, T1N, T1F, T1R, T1u, T1M, T1D, T1s; cannam@167: T1x = FNMS(KP951056516, T1w, KP587785252 * T1v); cannam@167: T1C = FNMS(KP951056516, T1B, KP587785252 * T1A); cannam@167: T1Q = FMA(KP951056516, T1A, KP587785252 * T1B); cannam@167: T1N = FMA(KP951056516, T1v, KP587785252 * T1w); cannam@167: T1D = FNMS(KP250000000, T1p, T1i); cannam@167: T1F = T1D - T1E; cannam@167: T1R = T1E + T1D; cannam@167: T1s = FNMS(KP250000000, T1f, T18); cannam@167: T1u = T1s - T1t; cannam@167: T1M = T1t + T1s; cannam@167: { cannam@167: E T1y, T1G, T1r, T1z; cannam@167: T1y = T1u - T1x; cannam@167: T1G = T1C + T1F; cannam@167: T1r = W[12]; cannam@167: T1z = W[13]; cannam@167: Ip[WS(rs, 3)] = FNMS(T1z, T1G, T1r * T1y); cannam@167: Im[WS(rs, 3)] = FMA(T1r, T1G, T1z * T1y); cannam@167: } cannam@167: { cannam@167: E T1U, T1W, T1T, T1V; cannam@167: T1U = T1M + T1N; cannam@167: T1W = T1R - T1Q; cannam@167: T1T = W[16]; cannam@167: T1V = W[17]; cannam@167: Ip[WS(rs, 4)] = FNMS(T1V, T1W, T1T * T1U); cannam@167: Im[WS(rs, 4)] = FMA(T1T, T1W, T1V * T1U); cannam@167: } cannam@167: { cannam@167: E T1I, T1K, T1H, T1J; cannam@167: T1I = T1u + T1x; cannam@167: T1K = T1F - T1C; cannam@167: T1H = W[4]; cannam@167: T1J = W[5]; cannam@167: Ip[WS(rs, 1)] = FNMS(T1J, T1K, T1H * T1I); cannam@167: Im[WS(rs, 1)] = FMA(T1H, T1K, T1J * T1I); cannam@167: } cannam@167: { cannam@167: E T1O, T1S, T1L, T1P; cannam@167: T1O = T1M - T1N; cannam@167: T1S = T1Q + T1R; cannam@167: T1L = W[0]; cannam@167: T1P = W[1]; cannam@167: Ip[0] = FNMS(T1P, T1S, T1L * T1O); cannam@167: Im[0] = FMA(T1L, T1S, T1P * T1O); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: {TW_FULL, 1, 10}, cannam@167: {TW_NEXT, 1, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 10, "hc2cb_10", twinstr, &GENUS, {72, 30, 30, 0} }; cannam@167: cannam@167: void X(codelet_hc2cb_10) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cb_10, &desc, HC2C_VIA_RDFT); cannam@167: } cannam@167: #endif