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 08:08:12 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_hc2cdft_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 16 -dif -sign 1 -name hc2cbdftv_16 -include rdft/simd/hc2cbv.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 103 FP additions, 80 FP multiplications, cannam@167: * (or, 53 additions, 30 multiplications, 50 fused multiply/add), cannam@167: * 79 stack variables, 3 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/simd/hc2cbv.h" cannam@167: cannam@167: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@167: DVK(KP414213562, +0.414213562373095048801688724209698078569671875); cannam@167: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 30)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(64, rs)) { cannam@167: V T8, Tv, TE, T1t, TP, T1w, T10, T1p, Tn, Tw, T13, T1q, TL, T1x, TS; cannam@167: V T1u; cannam@167: { cannam@167: V T4, TA, Tu, TC, T7, TN, Tr, TB, T2, T3, Ts, Tt, T5, T6, Tp; cannam@167: V Tq, TD, TO, TY, TZ, Tb, TF, Tl, TJ, Te, TG, Ti, TI, T9, Ta; cannam@167: V Tj, Tk, Tc, Td, Tg, Th, Tf, Tm, T11, T12, TH, TK, TQ, TR; cannam@167: T2 = LD(&(Rp[0]), ms, &(Rp[0])); cannam@167: T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: T4 = VFMACONJ(T3, T2); cannam@167: TA = VFNMSCONJ(T3, T2); cannam@167: Ts = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); cannam@167: Tt = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: Tu = VFMACONJ(Tt, Ts); cannam@167: TC = VFMSCONJ(Tt, Ts); cannam@167: T5 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); cannam@167: T6 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: T7 = VFMACONJ(T6, T5); cannam@167: TN = VFNMSCONJ(T6, T5); cannam@167: Tp = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); cannam@167: Tq = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: Tr = VFMACONJ(Tq, Tp); cannam@167: TB = VFNMSCONJ(Tq, Tp); cannam@167: T8 = VSUB(T4, T7); cannam@167: Tv = VSUB(Tr, Tu); cannam@167: TD = VADD(TB, TC); cannam@167: TE = VFMA(LDK(KP707106781), TD, TA); cannam@167: T1t = VFNMS(LDK(KP707106781), TD, TA); cannam@167: TO = VSUB(TB, TC); cannam@167: TP = VFMA(LDK(KP707106781), TO, TN); cannam@167: T1w = VFNMS(LDK(KP707106781), TO, TN); cannam@167: TY = VADD(T4, T7); cannam@167: TZ = VADD(Tr, Tu); cannam@167: T10 = VADD(TY, TZ); cannam@167: T1p = VSUB(TY, TZ); cannam@167: T9 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Ta = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); cannam@167: Tb = VFMACONJ(Ta, T9); cannam@167: TF = VFNMSCONJ(Ta, T9); cannam@167: Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Tk = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); cannam@167: Tl = VFMACONJ(Tk, Tj); cannam@167: TJ = VFNMSCONJ(Tk, Tj); cannam@167: Tc = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Td = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); cannam@167: Te = VFMACONJ(Td, Tc); cannam@167: TG = VFNMSCONJ(Td, Tc); cannam@167: Tg = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Th = LD(&(Rm[0]), -ms, &(Rm[0])); cannam@167: Ti = VFMACONJ(Th, Tg); cannam@167: TI = VFMSCONJ(Th, Tg); cannam@167: Tf = VSUB(Tb, Te); cannam@167: Tm = VSUB(Ti, Tl); cannam@167: Tn = VADD(Tf, Tm); cannam@167: Tw = VSUB(Tf, Tm); cannam@167: T11 = VADD(Tb, Te); cannam@167: T12 = VADD(Ti, Tl); cannam@167: T13 = VADD(T11, T12); cannam@167: T1q = VSUB(T11, T12); cannam@167: TH = VFNMS(LDK(KP414213562), TG, TF); cannam@167: TK = VFMA(LDK(KP414213562), TJ, TI); cannam@167: TL = VADD(TH, TK); cannam@167: T1x = VSUB(TH, TK); cannam@167: TQ = VFMA(LDK(KP414213562), TF, TG); cannam@167: TR = VFNMS(LDK(KP414213562), TI, TJ); cannam@167: TS = VADD(TQ, TR); cannam@167: T1u = VSUB(TQ, TR); cannam@167: } cannam@167: { cannam@167: V T1j, T1R, T1c, T1J, T1g, T1l, T1N, T1T, T1Q, T1a, T1b, T19, T1I, T1e, T1f; cannam@167: V T1d, T1k, T1L, T1M, T1K, T1S, T1h, T1U, T1V, T1i, T1m, T1O, T1P, T1n, T14; cannam@167: V T1r, Ty, T1D, TU, T16, T1z, T1F, TX, T1o, To, Tx, T1, T1C, TM, TT; cannam@167: V Tz, T15, T1v, T1y, T1s, T1E, TV, T1G, T1H, TW, T17, T1A, T1B, T18; cannam@167: T1j = VADD(T10, T13); cannam@167: T1Q = LDW(&(W[TWVL * 22])); cannam@167: T1R = VZMUL(T1Q, VFNMSI(T1q, T1p)); cannam@167: T1a = VFMA(LDK(KP707106781), Tn, T8); cannam@167: T1b = VFMA(LDK(KP707106781), Tw, Tv); cannam@167: T19 = LDW(&(W[TWVL * 26])); cannam@167: T1c = VZMUL(T19, VFNMSI(T1b, T1a)); cannam@167: T1I = LDW(&(W[TWVL * 2])); cannam@167: T1J = VZMUL(T1I, VFMAI(T1b, T1a)); cannam@167: T1e = VFMA(LDK(KP923879532), TL, TE); cannam@167: T1f = VFMA(LDK(KP923879532), TS, TP); cannam@167: T1d = LDW(&(W[TWVL * 28])); cannam@167: T1g = VZMULI(T1d, VFNMSI(T1f, T1e)); cannam@167: T1k = LDW(&(W[0])); cannam@167: T1l = VZMULI(T1k, VFMAI(T1f, T1e)); cannam@167: T1L = VFMA(LDK(KP923879532), T1u, T1t); cannam@167: T1M = VFNMS(LDK(KP923879532), T1x, T1w); cannam@167: T1K = LDW(&(W[TWVL * 4])); cannam@167: T1N = VZMULI(T1K, VFNMSI(T1M, T1L)); cannam@167: T1S = LDW(&(W[TWVL * 24])); cannam@167: T1T = VZMULI(T1S, VFMAI(T1M, T1L)); cannam@167: T1h = VCONJ(VSUB(T1c, T1g)); cannam@167: ST(&(Rm[WS(rs, 7)]), T1h, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1U = VCONJ(VSUB(T1R, T1T)); cannam@167: ST(&(Rm[WS(rs, 6)]), T1U, -ms, &(Rm[0])); cannam@167: T1V = VADD(T1R, T1T); cannam@167: ST(&(Rp[WS(rs, 6)]), T1V, ms, &(Rp[0])); cannam@167: T1i = VADD(T1c, T1g); cannam@167: ST(&(Rp[WS(rs, 7)]), T1i, ms, &(Rp[WS(rs, 1)])); cannam@167: T1m = VCONJ(VSUB(T1j, T1l)); cannam@167: ST(&(Rm[0]), T1m, -ms, &(Rm[0])); cannam@167: T1O = VCONJ(VSUB(T1J, T1N)); cannam@167: ST(&(Rm[WS(rs, 1)]), T1O, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1P = VADD(T1J, T1N); cannam@167: ST(&(Rp[WS(rs, 1)]), T1P, ms, &(Rp[WS(rs, 1)])); cannam@167: T1n = VADD(T1j, T1l); cannam@167: ST(&(Rp[0]), T1n, ms, &(Rp[0])); cannam@167: TX = LDW(&(W[TWVL * 14])); cannam@167: T14 = VZMUL(TX, VSUB(T10, T13)); cannam@167: T1o = LDW(&(W[TWVL * 6])); cannam@167: T1r = VZMUL(T1o, VFMAI(T1q, T1p)); cannam@167: To = VFNMS(LDK(KP707106781), Tn, T8); cannam@167: Tx = VFNMS(LDK(KP707106781), Tw, Tv); cannam@167: T1 = LDW(&(W[TWVL * 10])); cannam@167: Ty = VZMUL(T1, VFNMSI(Tx, To)); cannam@167: T1C = LDW(&(W[TWVL * 18])); cannam@167: T1D = VZMUL(T1C, VFMAI(Tx, To)); cannam@167: TM = VFNMS(LDK(KP923879532), TL, TE); cannam@167: TT = VFNMS(LDK(KP923879532), TS, TP); cannam@167: Tz = LDW(&(W[TWVL * 12])); cannam@167: TU = VZMULI(Tz, VFNMSI(TT, TM)); cannam@167: T15 = LDW(&(W[TWVL * 16])); cannam@167: T16 = VZMULI(T15, VFMAI(TT, TM)); cannam@167: T1v = VFNMS(LDK(KP923879532), T1u, T1t); cannam@167: T1y = VFMA(LDK(KP923879532), T1x, T1w); cannam@167: T1s = LDW(&(W[TWVL * 8])); cannam@167: T1z = VZMULI(T1s, VFMAI(T1y, T1v)); cannam@167: T1E = LDW(&(W[TWVL * 20])); cannam@167: T1F = VZMULI(T1E, VFNMSI(T1y, T1v)); cannam@167: TV = VCONJ(VSUB(Ty, TU)); cannam@167: ST(&(Rm[WS(rs, 3)]), TV, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1G = VCONJ(VSUB(T1D, T1F)); cannam@167: ST(&(Rm[WS(rs, 5)]), T1G, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1H = VADD(T1D, T1F); cannam@167: ST(&(Rp[WS(rs, 5)]), T1H, ms, &(Rp[WS(rs, 1)])); cannam@167: TW = VADD(Ty, TU); cannam@167: ST(&(Rp[WS(rs, 3)]), TW, ms, &(Rp[WS(rs, 1)])); cannam@167: T17 = VCONJ(VSUB(T14, T16)); cannam@167: ST(&(Rm[WS(rs, 4)]), T17, -ms, &(Rm[0])); cannam@167: T1A = VCONJ(VSUB(T1r, T1z)); cannam@167: ST(&(Rm[WS(rs, 2)]), T1A, -ms, &(Rm[0])); cannam@167: T1B = VADD(T1r, T1z); cannam@167: ST(&(Rp[WS(rs, 2)]), T1B, ms, &(Rp[0])); cannam@167: T18 = VADD(T14, T16); cannam@167: ST(&(Rp[WS(rs, 4)]), T18, ms, &(Rp[0])); cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: VTW(1, 1), cannam@167: VTW(1, 2), cannam@167: VTW(1, 3), cannam@167: VTW(1, 4), cannam@167: VTW(1, 5), cannam@167: VTW(1, 6), cannam@167: VTW(1, 7), cannam@167: VTW(1, 8), cannam@167: VTW(1, 9), cannam@167: VTW(1, 10), cannam@167: VTW(1, 11), cannam@167: VTW(1, 12), cannam@167: VTW(1, 13), cannam@167: VTW(1, 14), cannam@167: VTW(1, 15), cannam@167: {TW_NEXT, VL, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {53, 30, 50, 0} }; cannam@167: cannam@167: void XSIMD(codelet_hc2cbdftv_16) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cbdftv_16, &desc, HC2C_VIA_DFT); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 16 -dif -sign 1 -name hc2cbdftv_16 -include rdft/simd/hc2cbv.h */ cannam@167: cannam@167: /* cannam@167: * This function contains 103 FP additions, 42 FP multiplications, cannam@167: * (or, 99 additions, 38 multiplications, 4 fused multiply/add), cannam@167: * 83 stack variables, 3 constants, and 32 memory accesses cannam@167: */ cannam@167: #include "rdft/simd/hc2cbv.h" cannam@167: cannam@167: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); cannam@167: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); cannam@167: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT m; cannam@167: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 30)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(64, rs)) { cannam@167: V Tf, T16, TZ, T1C, TI, T1a, TV, T1D, T1F, T1G, Ty, T19, TC, T17, TS; cannam@167: V T10; cannam@167: { cannam@167: V T2, TD, T4, TF, Tc, Tb, Td, T6, T8, T9, T3, TE, Ta, T7, T5; cannam@167: V Te, TX, TY, TG, TH, TT, TU, Tj, TM, Tw, TQ, Tn, TN, Ts, TP; cannam@167: V Tg, Ti, Th, Tt, Tv, Tu, Tk, Tm, Tl, Tr, Tq, Tp, To, Tx, TA; cannam@167: V TB, TO, TR; cannam@167: T2 = LD(&(Rp[0]), ms, &(Rp[0])); cannam@167: TD = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); cannam@167: T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: T4 = VCONJ(T3); cannam@167: TE = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: TF = VCONJ(TE); cannam@167: Tc = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); cannam@167: Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: Tb = VCONJ(Ta); cannam@167: Td = VSUB(Tb, Tc); cannam@167: T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); cannam@167: T7 = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); cannam@167: T8 = VCONJ(T7); cannam@167: T9 = VSUB(T6, T8); cannam@167: T5 = VSUB(T2, T4); cannam@167: Te = VMUL(LDK(KP707106781), VADD(T9, Td)); cannam@167: Tf = VADD(T5, Te); cannam@167: T16 = VSUB(T5, Te); cannam@167: TX = VADD(T2, T4); cannam@167: TY = VADD(TD, TF); cannam@167: TZ = VSUB(TX, TY); cannam@167: T1C = VADD(TX, TY); cannam@167: TG = VSUB(TD, TF); cannam@167: TH = VMUL(LDK(KP707106781), VSUB(T9, Td)); cannam@167: TI = VADD(TG, TH); cannam@167: T1a = VSUB(TH, TG); cannam@167: TT = VADD(T6, T8); cannam@167: TU = VADD(Tb, Tc); cannam@167: TV = VSUB(TT, TU); cannam@167: T1D = VADD(TT, TU); cannam@167: Tg = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Th = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); cannam@167: Ti = VCONJ(Th); cannam@167: Tj = VSUB(Tg, Ti); cannam@167: TM = VADD(Tg, Ti); cannam@167: Tt = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Tu = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); cannam@167: Tv = VCONJ(Tu); cannam@167: Tw = VSUB(Tt, Tv); cannam@167: TQ = VADD(Tt, Tv); cannam@167: Tk = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Tl = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); cannam@167: Tm = VCONJ(Tl); cannam@167: Tn = VSUB(Tk, Tm); cannam@167: TN = VADD(Tk, Tm); cannam@167: Tr = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); cannam@167: Tp = LD(&(Rm[0]), -ms, &(Rm[0])); cannam@167: Tq = VCONJ(Tp); cannam@167: Ts = VSUB(Tq, Tr); cannam@167: TP = VADD(Tq, Tr); cannam@167: T1F = VADD(TM, TN); cannam@167: T1G = VADD(TP, TQ); cannam@167: To = VFNMS(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tj)); cannam@167: Tx = VFMA(LDK(KP923879532), Ts, VMUL(LDK(KP382683432), Tw)); cannam@167: Ty = VADD(To, Tx); cannam@167: T19 = VSUB(To, Tx); cannam@167: TA = VFMA(LDK(KP382683432), Tj, VMUL(LDK(KP923879532), Tn)); cannam@167: TB = VFNMS(LDK(KP382683432), Ts, VMUL(LDK(KP923879532), Tw)); cannam@167: TC = VADD(TA, TB); cannam@167: T17 = VSUB(TA, TB); cannam@167: TO = VSUB(TM, TN); cannam@167: TR = VSUB(TP, TQ); cannam@167: TS = VMUL(LDK(KP707106781), VSUB(TO, TR)); cannam@167: T10 = VMUL(LDK(KP707106781), VADD(TO, TR)); cannam@167: } cannam@167: { cannam@167: V T21, T1W, T1u, T20, T1I, T1O, TK, T1S, T12, T1e, T1k, T1A, T1o, T1w, T1c; cannam@167: V T1M, T1U, T1V, T1T, T1s, T1t, T1r, T1Z, T1E, T1H, T1B, T1N, Tz, TJ, T1; cannam@167: V T1R, TW, T11, TL, T1d, T1i, T1j, T1h, T1z, T1m, T1n, T1l, T1v, T18, T1b; cannam@167: V T15, T1L, T13, T1g, T1X, T23, T14, T1f, T1Y, T22, T1p, T1y, T1J, T1Q, T1q; cannam@167: V T1x, T1K, T1P; cannam@167: T1U = VADD(T1C, T1D); cannam@167: T1V = VADD(T1F, T1G); cannam@167: T21 = VADD(T1U, T1V); cannam@167: T1T = LDW(&(W[TWVL * 14])); cannam@167: T1W = VZMUL(T1T, VSUB(T1U, T1V)); cannam@167: T1s = VADD(Tf, Ty); cannam@167: T1t = VBYI(VADD(TI, TC)); cannam@167: T1r = LDW(&(W[TWVL * 28])); cannam@167: T1u = VZMULI(T1r, VSUB(T1s, T1t)); cannam@167: T1Z = LDW(&(W[0])); cannam@167: T20 = VZMULI(T1Z, VADD(T1s, T1t)); cannam@167: T1E = VSUB(T1C, T1D); cannam@167: T1H = VBYI(VSUB(T1F, T1G)); cannam@167: T1B = LDW(&(W[TWVL * 22])); cannam@167: T1I = VZMUL(T1B, VSUB(T1E, T1H)); cannam@167: T1N = LDW(&(W[TWVL * 6])); cannam@167: T1O = VZMUL(T1N, VADD(T1E, T1H)); cannam@167: Tz = VSUB(Tf, Ty); cannam@167: TJ = VBYI(VSUB(TC, TI)); cannam@167: T1 = LDW(&(W[TWVL * 12])); cannam@167: TK = VZMULI(T1, VADD(Tz, TJ)); cannam@167: T1R = LDW(&(W[TWVL * 16])); cannam@167: T1S = VZMULI(T1R, VSUB(Tz, TJ)); cannam@167: TW = VBYI(VSUB(TS, TV)); cannam@167: T11 = VSUB(TZ, T10); cannam@167: TL = LDW(&(W[TWVL * 10])); cannam@167: T12 = VZMUL(TL, VADD(TW, T11)); cannam@167: T1d = LDW(&(W[TWVL * 18])); cannam@167: T1e = VZMUL(T1d, VSUB(T11, TW)); cannam@167: T1i = VBYI(VADD(T1a, T19)); cannam@167: T1j = VADD(T16, T17); cannam@167: T1h = LDW(&(W[TWVL * 4])); cannam@167: T1k = VZMULI(T1h, VADD(T1i, T1j)); cannam@167: T1z = LDW(&(W[TWVL * 24])); cannam@167: T1A = VZMULI(T1z, VSUB(T1j, T1i)); cannam@167: T1m = VBYI(VADD(TV, TS)); cannam@167: T1n = VADD(TZ, T10); cannam@167: T1l = LDW(&(W[TWVL * 2])); cannam@167: T1o = VZMUL(T1l, VADD(T1m, T1n)); cannam@167: T1v = LDW(&(W[TWVL * 26])); cannam@167: T1w = VZMUL(T1v, VSUB(T1n, T1m)); cannam@167: T18 = VSUB(T16, T17); cannam@167: T1b = VBYI(VSUB(T19, T1a)); cannam@167: T15 = LDW(&(W[TWVL * 20])); cannam@167: T1c = VZMULI(T15, VSUB(T18, T1b)); cannam@167: T1L = LDW(&(W[TWVL * 8])); cannam@167: T1M = VZMULI(T1L, VADD(T1b, T18)); cannam@167: T13 = VADD(TK, T12); cannam@167: ST(&(Rp[WS(rs, 3)]), T13, ms, &(Rp[WS(rs, 1)])); cannam@167: T1g = VCONJ(VSUB(T1e, T1c)); cannam@167: ST(&(Rm[WS(rs, 5)]), T1g, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1X = VADD(T1S, T1W); cannam@167: ST(&(Rp[WS(rs, 4)]), T1X, ms, &(Rp[0])); cannam@167: T23 = VCONJ(VSUB(T21, T20)); cannam@167: ST(&(Rm[0]), T23, -ms, &(Rm[0])); cannam@167: T14 = VCONJ(VSUB(T12, TK)); cannam@167: ST(&(Rm[WS(rs, 3)]), T14, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1f = VADD(T1c, T1e); cannam@167: ST(&(Rp[WS(rs, 5)]), T1f, ms, &(Rp[WS(rs, 1)])); cannam@167: T1Y = VCONJ(VSUB(T1W, T1S)); cannam@167: ST(&(Rm[WS(rs, 4)]), T1Y, -ms, &(Rm[0])); cannam@167: T22 = VADD(T20, T21); cannam@167: ST(&(Rp[0]), T22, ms, &(Rp[0])); cannam@167: T1p = VADD(T1k, T1o); cannam@167: ST(&(Rp[WS(rs, 1)]), T1p, ms, &(Rp[WS(rs, 1)])); cannam@167: T1y = VCONJ(VSUB(T1w, T1u)); cannam@167: ST(&(Rm[WS(rs, 7)]), T1y, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1J = VADD(T1A, T1I); cannam@167: ST(&(Rp[WS(rs, 6)]), T1J, ms, &(Rp[0])); cannam@167: T1Q = VCONJ(VSUB(T1O, T1M)); cannam@167: ST(&(Rm[WS(rs, 2)]), T1Q, -ms, &(Rm[0])); cannam@167: T1q = VCONJ(VSUB(T1o, T1k)); cannam@167: ST(&(Rm[WS(rs, 1)]), T1q, -ms, &(Rm[WS(rs, 1)])); cannam@167: T1x = VADD(T1u, T1w); cannam@167: ST(&(Rp[WS(rs, 7)]), T1x, ms, &(Rp[WS(rs, 1)])); cannam@167: T1K = VCONJ(VSUB(T1I, T1A)); cannam@167: ST(&(Rm[WS(rs, 6)]), T1K, -ms, &(Rm[0])); cannam@167: T1P = VADD(T1M, T1O); cannam@167: ST(&(Rp[WS(rs, 2)]), T1P, ms, &(Rp[0])); cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: VTW(1, 1), cannam@167: VTW(1, 2), cannam@167: VTW(1, 3), cannam@167: VTW(1, 4), cannam@167: VTW(1, 5), cannam@167: VTW(1, 6), cannam@167: VTW(1, 7), cannam@167: VTW(1, 8), cannam@167: VTW(1, 9), cannam@167: VTW(1, 10), cannam@167: VTW(1, 11), cannam@167: VTW(1, 12), cannam@167: VTW(1, 13), cannam@167: VTW(1, 14), cannam@167: VTW(1, 15), cannam@167: {TW_NEXT, VL, 0} cannam@167: }; cannam@167: cannam@167: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {99, 38, 4, 0} }; cannam@167: cannam@167: void XSIMD(codelet_hc2cbdftv_16) (planner *p) { cannam@167: X(khc2c_register) (p, hc2cbdftv_16, &desc, HC2C_VIA_DFT); cannam@167: } cannam@167: #endif