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 08:06:43 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_r2cf.native -fma -compact -variables 4 -pipeline-latency 4 -n 16 -name r2cfII_16 -dft-II -include rdft/scalar/r2cfII.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 66 FP additions, 48 FP multiplications, cannam@167: * (or, 18 additions, 0 multiplications, 48 fused multiply/add), cannam@167: * 32 stack variables, 7 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cfII.h" cannam@167: cannam@167: static void r2cfII_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DK(KP980785280, +0.980785280403230449126182236134239036973933731); cannam@167: DK(KP198912367, +0.198912367379658006911597622644676228597850501); cannam@167: DK(KP831469612, +0.831469612302545237078788377617905756738560812); cannam@167: DK(KP668178637, +0.668178637919298919997757686523080761552472251); cannam@167: DK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@167: DK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@167: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@167: E T5, TZ, TB, TT, Tr, TK, Tu, TJ, Ti, TH, Tl, TG, Tc, T10, TE; cannam@167: E TU; cannam@167: { cannam@167: E T1, TR, T4, TS, T2, T3; cannam@167: T1 = R0[0]; cannam@167: TR = R0[WS(rs, 4)]; cannam@167: T2 = R0[WS(rs, 2)]; cannam@167: T3 = R0[WS(rs, 6)]; cannam@167: T4 = T2 - T3; cannam@167: TS = T2 + T3; cannam@167: T5 = FNMS(KP707106781, T4, T1); cannam@167: TZ = FNMS(KP707106781, TS, TR); cannam@167: TB = FMA(KP707106781, T4, T1); cannam@167: TT = FMA(KP707106781, TS, TR); cannam@167: } cannam@167: { cannam@167: E Tn, Ts, Tq, Tt, To, Tp; cannam@167: Tn = R1[WS(rs, 7)]; cannam@167: Ts = R1[WS(rs, 3)]; cannam@167: To = R1[WS(rs, 1)]; cannam@167: Tp = R1[WS(rs, 5)]; cannam@167: Tq = To - Tp; cannam@167: Tt = To + Tp; cannam@167: Tr = FMA(KP707106781, Tq, Tn); cannam@167: TK = FMA(KP707106781, Tt, Ts); cannam@167: Tu = FNMS(KP707106781, Tt, Ts); cannam@167: TJ = FMS(KP707106781, Tq, Tn); cannam@167: } cannam@167: { cannam@167: E Te, Tj, Th, Tk, Tf, Tg; cannam@167: Te = R1[0]; cannam@167: Tj = R1[WS(rs, 4)]; cannam@167: Tf = R1[WS(rs, 2)]; cannam@167: Tg = R1[WS(rs, 6)]; cannam@167: Th = Tf - Tg; cannam@167: Tk = Tf + Tg; cannam@167: Ti = FNMS(KP707106781, Th, Te); cannam@167: TH = FMA(KP707106781, Tk, Tj); cannam@167: Tl = FNMS(KP707106781, Tk, Tj); cannam@167: TG = FMA(KP707106781, Th, Te); cannam@167: } cannam@167: { cannam@167: E T8, TC, Tb, TD; cannam@167: { cannam@167: E T6, T7, T9, Ta; cannam@167: T6 = R0[WS(rs, 5)]; cannam@167: T7 = R0[WS(rs, 1)]; cannam@167: T8 = FMA(KP414213562, T7, T6); cannam@167: TC = FNMS(KP414213562, T6, T7); cannam@167: T9 = R0[WS(rs, 3)]; cannam@167: Ta = R0[WS(rs, 7)]; cannam@167: Tb = FMA(KP414213562, Ta, T9); cannam@167: TD = FMS(KP414213562, T9, Ta); cannam@167: } cannam@167: Tc = T8 - Tb; cannam@167: T10 = TD - TC; cannam@167: TE = TC + TD; cannam@167: TU = T8 + Tb; cannam@167: } cannam@167: { cannam@167: E Td, T13, Tw, T14, Tm, Tv; cannam@167: Td = FMA(KP923879532, Tc, T5); cannam@167: T13 = FNMS(KP923879532, T10, TZ); cannam@167: Tm = FMA(KP668178637, Tl, Ti); cannam@167: Tv = FMA(KP668178637, Tu, Tr); cannam@167: Tw = Tm - Tv; cannam@167: T14 = Tm + Tv; cannam@167: Cr[WS(csr, 6)] = FNMS(KP831469612, Tw, Td); cannam@167: Ci[WS(csi, 5)] = FNMS(KP831469612, T14, T13); cannam@167: Cr[WS(csr, 1)] = FMA(KP831469612, Tw, Td); cannam@167: Ci[WS(csi, 2)] = -(FMA(KP831469612, T14, T13)); cannam@167: } cannam@167: { cannam@167: E Tx, T11, TA, T12, Ty, Tz; cannam@167: Tx = FNMS(KP923879532, Tc, T5); cannam@167: T11 = FMA(KP923879532, T10, TZ); cannam@167: Ty = FNMS(KP668178637, Tr, Tu); cannam@167: Tz = FNMS(KP668178637, Ti, Tl); cannam@167: TA = Ty - Tz; cannam@167: T12 = Tz + Ty; cannam@167: Cr[WS(csr, 5)] = FNMS(KP831469612, TA, Tx); cannam@167: Ci[WS(csi, 1)] = FMA(KP831469612, T12, T11); cannam@167: Cr[WS(csr, 2)] = FMA(KP831469612, TA, Tx); cannam@167: Ci[WS(csi, 6)] = FMS(KP831469612, T12, T11); cannam@167: } cannam@167: { cannam@167: E TF, TX, TM, TY, TI, TL; cannam@167: TF = FMA(KP923879532, TE, TB); cannam@167: TX = FNMS(KP923879532, TU, TT); cannam@167: TI = FNMS(KP198912367, TH, TG); cannam@167: TL = FMA(KP198912367, TK, TJ); cannam@167: TM = TI + TL; cannam@167: TY = TL - TI; cannam@167: Cr[WS(csr, 7)] = FNMS(KP980785280, TM, TF); cannam@167: Ci[WS(csi, 3)] = FMA(KP980785280, TY, TX); cannam@167: Cr[0] = FMA(KP980785280, TM, TF); cannam@167: Ci[WS(csi, 4)] = FMS(KP980785280, TY, TX); cannam@167: } cannam@167: { cannam@167: E TN, TV, TQ, TW, TO, TP; cannam@167: TN = FNMS(KP923879532, TE, TB); cannam@167: TV = FMA(KP923879532, TU, TT); cannam@167: TO = FMA(KP198912367, TG, TH); cannam@167: TP = FNMS(KP198912367, TJ, TK); cannam@167: TQ = TO - TP; cannam@167: TW = TO + TP; cannam@167: Cr[WS(csr, 4)] = FNMS(KP980785280, TQ, TN); cannam@167: Ci[WS(csi, 7)] = FNMS(KP980785280, TW, TV); cannam@167: Cr[WS(csr, 3)] = FMA(KP980785280, TQ, TN); cannam@167: Ci[0] = -(FMA(KP980785280, TW, TV)); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 16, "r2cfII_16", {18, 0, 48, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cfII_16) (planner *p) { cannam@167: X(kr2c_register) (p, r2cfII_16, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 16 -name r2cfII_16 -dft-II -include rdft/scalar/r2cfII.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 66 FP additions, 30 FP multiplications, cannam@167: * (or, 54 additions, 18 multiplications, 12 fused multiply/add), cannam@167: * 32 stack variables, 7 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/scalar/r2cfII.h" cannam@167: cannam@167: static void r2cfII_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DK(KP555570233, +0.555570233019602224742830813948532874374937191); cannam@167: DK(KP831469612, +0.831469612302545237078788377617905756738560812); cannam@167: DK(KP980785280, +0.980785280403230449126182236134239036973933731); cannam@167: DK(KP195090322, +0.195090322016128267848284868477022240927691618); cannam@167: DK(KP382683432, +0.382683432365089771728459984030398866761344562); cannam@167: DK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@167: DK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) { cannam@167: E T5, T11, TB, TV, Tr, TK, Tu, TJ, Ti, TH, Tl, TG, Tc, T10, TE; cannam@167: E TS; cannam@167: { cannam@167: E T1, TU, T4, TT, T2, T3; cannam@167: T1 = R0[0]; cannam@167: TU = R0[WS(rs, 4)]; cannam@167: T2 = R0[WS(rs, 2)]; cannam@167: T3 = R0[WS(rs, 6)]; cannam@167: T4 = KP707106781 * (T2 - T3); cannam@167: TT = KP707106781 * (T2 + T3); cannam@167: T5 = T1 + T4; cannam@167: T11 = TU - TT; cannam@167: TB = T1 - T4; cannam@167: TV = TT + TU; cannam@167: } cannam@167: { cannam@167: E Tq, Tt, Tp, Ts, Tn, To; cannam@167: Tq = R1[WS(rs, 7)]; cannam@167: Tt = R1[WS(rs, 3)]; cannam@167: Tn = R1[WS(rs, 1)]; cannam@167: To = R1[WS(rs, 5)]; cannam@167: Tp = KP707106781 * (Tn - To); cannam@167: Ts = KP707106781 * (Tn + To); cannam@167: Tr = Tp - Tq; cannam@167: TK = Tt - Ts; cannam@167: Tu = Ts + Tt; cannam@167: TJ = Tp + Tq; cannam@167: } cannam@167: { cannam@167: E Te, Tk, Th, Tj, Tf, Tg; cannam@167: Te = R1[0]; cannam@167: Tk = R1[WS(rs, 4)]; cannam@167: Tf = R1[WS(rs, 2)]; cannam@167: Tg = R1[WS(rs, 6)]; cannam@167: Th = KP707106781 * (Tf - Tg); cannam@167: Tj = KP707106781 * (Tf + Tg); cannam@167: Ti = Te + Th; cannam@167: TH = Tk - Tj; cannam@167: Tl = Tj + Tk; cannam@167: TG = Te - Th; cannam@167: } cannam@167: { cannam@167: E T8, TC, Tb, TD; cannam@167: { cannam@167: E T6, T7, T9, Ta; cannam@167: T6 = R0[WS(rs, 1)]; cannam@167: T7 = R0[WS(rs, 5)]; cannam@167: T8 = FNMS(KP382683432, T7, KP923879532 * T6); cannam@167: TC = FMA(KP382683432, T6, KP923879532 * T7); cannam@167: T9 = R0[WS(rs, 3)]; cannam@167: Ta = R0[WS(rs, 7)]; cannam@167: Tb = FNMS(KP923879532, Ta, KP382683432 * T9); cannam@167: TD = FMA(KP923879532, T9, KP382683432 * Ta); cannam@167: } cannam@167: Tc = T8 + Tb; cannam@167: T10 = Tb - T8; cannam@167: TE = TC - TD; cannam@167: TS = TC + TD; cannam@167: } cannam@167: { cannam@167: E Td, TW, Tw, TR, Tm, Tv; cannam@167: Td = T5 - Tc; cannam@167: TW = TS + TV; cannam@167: Tm = FMA(KP195090322, Ti, KP980785280 * Tl); cannam@167: Tv = FNMS(KP980785280, Tu, KP195090322 * Tr); cannam@167: Tw = Tm + Tv; cannam@167: TR = Tv - Tm; cannam@167: Cr[WS(csr, 4)] = Td - Tw; cannam@167: Ci[WS(csi, 7)] = TR + TW; cannam@167: Cr[WS(csr, 3)] = Td + Tw; cannam@167: Ci[0] = TR - TW; cannam@167: } cannam@167: { cannam@167: E Tx, TY, TA, TX, Ty, Tz; cannam@167: Tx = T5 + Tc; cannam@167: TY = TV - TS; cannam@167: Ty = FNMS(KP195090322, Tl, KP980785280 * Ti); cannam@167: Tz = FMA(KP980785280, Tr, KP195090322 * Tu); cannam@167: TA = Ty + Tz; cannam@167: TX = Tz - Ty; cannam@167: Cr[WS(csr, 7)] = Tx - TA; cannam@167: Ci[WS(csi, 3)] = TX + TY; cannam@167: Cr[0] = Tx + TA; cannam@167: Ci[WS(csi, 4)] = TX - TY; cannam@167: } cannam@167: { cannam@167: E TF, T12, TM, TZ, TI, TL; cannam@167: TF = TB + TE; cannam@167: T12 = T10 - T11; cannam@167: TI = FMA(KP831469612, TG, KP555570233 * TH); cannam@167: TL = FMA(KP831469612, TJ, KP555570233 * TK); cannam@167: TM = TI - TL; cannam@167: TZ = TI + TL; cannam@167: Cr[WS(csr, 6)] = TF - TM; cannam@167: Ci[WS(csi, 2)] = T12 - TZ; cannam@167: Cr[WS(csr, 1)] = TF + TM; cannam@167: Ci[WS(csi, 5)] = -(TZ + T12); cannam@167: } cannam@167: { cannam@167: E TN, T14, TQ, T13, TO, TP; cannam@167: TN = TB - TE; cannam@167: T14 = T10 + T11; cannam@167: TO = FNMS(KP555570233, TJ, KP831469612 * TK); cannam@167: TP = FNMS(KP555570233, TG, KP831469612 * TH); cannam@167: TQ = TO - TP; cannam@167: T13 = TP + TO; cannam@167: Cr[WS(csr, 5)] = TN - TQ; cannam@167: Ci[WS(csi, 1)] = T13 + T14; cannam@167: Cr[WS(csr, 2)] = TN + TQ; cannam@167: Ci[WS(csi, 6)] = T13 - T14; cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: cannam@167: static const kr2c_desc desc = { 16, "r2cfII_16", {54, 18, 12, 0}, &GENUS }; cannam@167: cannam@167: void X(codelet_r2cfII_16) (planner *p) { cannam@167: X(kr2c_register) (p, r2cfII_16, &desc); cannam@167: } cannam@167: cannam@167: #endif