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 16:52:44 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 16 -dif -sign 1 -name hc2cbdftv_16 -include hc2cbv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 103 FP additions, 80 FP multiplications, Chris@42: * (or, 53 additions, 30 multiplications, 50 fused multiply/add), Chris@42: * 123 stack variables, 3 constants, and 32 memory accesses Chris@42: */ Chris@42: #include "hc2cbv.h" Chris@42: Chris@42: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { 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) * 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)) { Chris@42: V T1D, T1F, TV, TW, T17, T18, T1B, T1A, T1H, T1G; Chris@42: { Chris@42: V T8, Tv, Tb, TF, Tl, TJ, TP, T1w, TE, T1t, T10, T1p, TG, Te, Tg; Chris@42: V Th, T2, T3, Ts, Tt, T5, T6, Tp, Tq, T9, TA, T4, TC, Tu, TN; Chris@42: V T7, TB, Tr, Ta, Tj, Tk, Tc, Td, TY, TD, TO, TZ, T1Q, T19, T1I; Chris@42: V T1d, Tf, T11, TH, TQ, Ti, TI, T1k, T1K, T1S, T1r, T14, T16, TU, Ty; Chris@42: V T1z, TX, T1o, T1, TK, TR, Tm, T12, T1C, Tz, T15; Chris@42: T2 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Ts = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); Chris@42: Tt = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T5 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); Chris@42: T6 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tp = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); Chris@42: Tq = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T9 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: TA = VFNMSCONJ(T3, T2); Chris@42: T4 = VFMACONJ(T3, T2); Chris@42: TC = VFMSCONJ(Tt, Ts); Chris@42: Tu = VFMACONJ(Tt, Ts); Chris@42: TN = VFNMSCONJ(T6, T5); Chris@42: T7 = VFMACONJ(T6, T5); Chris@42: TB = VFNMSCONJ(Tq, Tp); Chris@42: Tr = VFMACONJ(Tq, Tp); Chris@42: Ta = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); Chris@42: Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tk = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); Chris@42: Tc = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Td = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); Chris@42: T8 = VSUB(T4, T7); Chris@42: TY = VADD(T4, T7); Chris@42: TD = VADD(TB, TC); Chris@42: TO = VSUB(TB, TC); Chris@42: Tv = VSUB(Tr, Tu); Chris@42: TZ = VADD(Tr, Tu); Chris@42: Tb = VFMACONJ(Ta, T9); Chris@42: TF = VFNMSCONJ(Ta, T9); Chris@42: Tl = VFMACONJ(Tk, Tj); Chris@42: TJ = VFNMSCONJ(Tk, Tj); Chris@42: TP = VFMA(LDK(KP707106781), TO, TN); Chris@42: T1w = VFNMS(LDK(KP707106781), TO, TN); Chris@42: TE = VFMA(LDK(KP707106781), TD, TA); Chris@42: T1t = VFNMS(LDK(KP707106781), TD, TA); Chris@42: T10 = VADD(TY, TZ); Chris@42: T1p = VSUB(TY, TZ); Chris@42: TG = VFNMSCONJ(Td, Tc); Chris@42: Te = VFMACONJ(Td, Tc); Chris@42: Tg = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Th = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: T1Q = LDW(&(W[TWVL * 22])); Chris@42: T19 = LDW(&(W[TWVL * 26])); Chris@42: T1I = LDW(&(W[TWVL * 2])); Chris@42: T1d = LDW(&(W[TWVL * 28])); Chris@42: Tf = VSUB(Tb, Te); Chris@42: T11 = VADD(Tb, Te); Chris@42: TH = VFNMS(LDK(KP414213562), TG, TF); Chris@42: TQ = VFMA(LDK(KP414213562), TF, TG); Chris@42: Ti = VFMACONJ(Th, Tg); Chris@42: TI = VFMSCONJ(Th, Tg); Chris@42: T1k = LDW(&(W[0])); Chris@42: T1K = LDW(&(W[TWVL * 4])); Chris@42: T1S = LDW(&(W[TWVL * 24])); Chris@42: TX = LDW(&(W[TWVL * 14])); Chris@42: T1o = LDW(&(W[TWVL * 6])); Chris@42: T1 = LDW(&(W[TWVL * 10])); Chris@42: TK = VFMA(LDK(KP414213562), TJ, TI); Chris@42: TR = VFNMS(LDK(KP414213562), TI, TJ); Chris@42: Tm = VSUB(Ti, Tl); Chris@42: T12 = VADD(Ti, Tl); Chris@42: T1C = LDW(&(W[TWVL * 18])); Chris@42: Tz = LDW(&(W[TWVL * 12])); Chris@42: T15 = LDW(&(W[TWVL * 16])); Chris@42: { Chris@42: V T1v, T1y, T1N, T1g, T1J, T1c, T1U, T1V, T1m, T1n, T1s, TS, T1u, TL, T1x; Chris@42: V T13, T1q, Tn, Tw, T1L, T1f, TT, T1M, T1e, TM, T1R, T1j, T1b, Tx, T1a; Chris@42: V To, T1T, T1l, T1E, T1O, T1P, T1h, T1i; Chris@42: T1s = LDW(&(W[TWVL * 8])); Chris@42: TS = VADD(TQ, TR); Chris@42: T1u = VSUB(TQ, TR); Chris@42: TL = VADD(TH, TK); Chris@42: T1x = VSUB(TH, TK); Chris@42: T13 = VADD(T11, T12); Chris@42: T1q = VSUB(T11, T12); Chris@42: Tn = VADD(Tf, Tm); Chris@42: Tw = VSUB(Tf, Tm); Chris@42: T1L = VFMA(LDK(KP923879532), T1u, T1t); Chris@42: T1v = VFNMS(LDK(KP923879532), T1u, T1t); Chris@42: T1f = VFMA(LDK(KP923879532), TS, TP); Chris@42: TT = VFNMS(LDK(KP923879532), TS, TP); Chris@42: T1M = VFNMS(LDK(KP923879532), T1x, T1w); Chris@42: T1y = VFMA(LDK(KP923879532), T1x, T1w); Chris@42: T1e = VFMA(LDK(KP923879532), TL, TE); Chris@42: TM = VFNMS(LDK(KP923879532), TL, TE); Chris@42: T1r = VZMUL(T1o, VFMAI(T1q, T1p)); Chris@42: T1R = VZMUL(T1Q, VFNMSI(T1q, T1p)); Chris@42: T14 = VZMUL(TX, VSUB(T10, T13)); Chris@42: T1j = VADD(T10, T13); Chris@42: T1b = VFMA(LDK(KP707106781), Tw, Tv); Chris@42: Tx = VFNMS(LDK(KP707106781), Tw, Tv); Chris@42: T1a = VFMA(LDK(KP707106781), Tn, T8); Chris@42: To = VFNMS(LDK(KP707106781), Tn, T8); Chris@42: T1T = VZMULI(T1S, VFMAI(T1M, T1L)); Chris@42: T1N = VZMULI(T1K, VFNMSI(T1M, T1L)); Chris@42: T16 = VZMULI(T15, VFMAI(TT, TM)); Chris@42: TU = VZMULI(Tz, VFNMSI(TT, TM)); Chris@42: T1l = VZMULI(T1k, VFMAI(T1f, T1e)); Chris@42: T1g = VZMULI(T1d, VFNMSI(T1f, T1e)); Chris@42: T1D = VZMUL(T1C, VFMAI(Tx, To)); Chris@42: Ty = VZMUL(T1, VFNMSI(Tx, To)); Chris@42: T1J = VZMUL(T1I, VFMAI(T1b, T1a)); Chris@42: T1c = VZMUL(T19, VFNMSI(T1b, T1a)); Chris@42: T1U = VCONJ(VSUB(T1R, T1T)); Chris@42: T1V = VADD(T1R, T1T); Chris@42: T1m = VCONJ(VSUB(T1j, T1l)); Chris@42: T1n = VADD(T1j, T1l); Chris@42: T1z = VZMULI(T1s, VFMAI(T1y, T1v)); Chris@42: T1E = LDW(&(W[TWVL * 20])); Chris@42: T1O = VCONJ(VSUB(T1J, T1N)); Chris@42: T1P = VADD(T1J, T1N); Chris@42: T1h = VCONJ(VSUB(T1c, T1g)); Chris@42: T1i = VADD(T1c, T1g); Chris@42: ST(&(Rp[WS(rs, 6)]), T1V, ms, &(Rp[0])); Chris@42: ST(&(Rm[WS(rs, 6)]), T1U, -ms, &(Rm[0])); Chris@42: ST(&(Rp[0]), T1n, ms, &(Rp[0])); Chris@42: ST(&(Rm[0]), T1m, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 1)]), T1P, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 1)]), T1O, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rp[WS(rs, 7)]), T1i, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 7)]), T1h, -ms, &(Rm[WS(rs, 1)])); Chris@42: T1F = VZMULI(T1E, VFNMSI(T1y, T1v)); Chris@42: } Chris@42: TV = VCONJ(VSUB(Ty, TU)); Chris@42: TW = VADD(Ty, TU); Chris@42: T17 = VCONJ(VSUB(T14, T16)); Chris@42: T18 = VADD(T14, T16); Chris@42: T1B = VADD(T1r, T1z); Chris@42: T1A = VCONJ(VSUB(T1r, T1z)); Chris@42: } Chris@42: T1H = VADD(T1D, T1F); Chris@42: T1G = VCONJ(VSUB(T1D, T1F)); Chris@42: ST(&(Rm[WS(rs, 3)]), TV, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rp[WS(rs, 3)]), TW, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 4)]), T17, -ms, &(Rm[0])); Chris@42: ST(&(Rm[WS(rs, 2)]), T1A, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 2)]), T1B, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 4)]), T18, ms, &(Rp[0])); Chris@42: ST(&(Rp[WS(rs, 5)]), T1H, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[WS(rs, 5)]), T1G, -ms, &(Rm[WS(rs, 1)])); 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: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {53, 30, 50, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cbdftv_16) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cbdftv_16, &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 16 -dif -sign 1 -name hc2cbdftv_16 -include hc2cbv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 103 FP additions, 42 FP multiplications, Chris@42: * (or, 99 additions, 38 multiplications, 4 fused multiply/add), Chris@42: * 83 stack variables, 3 constants, and 32 memory accesses Chris@42: */ Chris@42: #include "hc2cbv.h" Chris@42: Chris@42: static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP382683432, +0.382683432365089771728459984030398866761344562); Chris@42: DVK(KP923879532, +0.923879532511286756128183189396788286822416626); Chris@42: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@42: { Chris@42: INT m; Chris@42: 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)) { Chris@42: V Tf, T16, TZ, T1C, TI, T1a, TV, T1D, T1F, T1G, Ty, T19, TC, T17, TS; Chris@42: V T10; Chris@42: { Chris@42: V T2, TD, T4, TF, Tc, Tb, Td, T6, T8, T9, T3, TE, Ta, T7, T5; Chris@42: V Te, TX, TY, TG, TH, TT, TU, Tj, TM, Tw, TQ, Tn, TN, Ts, TP; Chris@42: V Tg, Ti, Th, Tt, Tv, Tu, Tk, Tm, Tl, Tr, Tq, Tp, To, Tx, TA; Chris@42: V TB, TO, TR; Chris@42: T2 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: TD = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0])); Chris@42: T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T4 = VCONJ(T3); Chris@42: TE = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: TF = VCONJ(TE); Chris@42: Tc = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0])); Chris@42: Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: Tb = VCONJ(Ta); Chris@42: Td = VSUB(Tb, Tc); Chris@42: T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0])); Chris@42: T7 = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T8 = VCONJ(T7); Chris@42: T9 = VSUB(T6, T8); Chris@42: T5 = VSUB(T2, T4); Chris@42: Te = VMUL(LDK(KP707106781), VADD(T9, Td)); Chris@42: Tf = VADD(T5, Te); Chris@42: T16 = VSUB(T5, Te); Chris@42: TX = VADD(T2, T4); Chris@42: TY = VADD(TD, TF); Chris@42: TZ = VSUB(TX, TY); Chris@42: T1C = VADD(TX, TY); Chris@42: TG = VSUB(TD, TF); Chris@42: TH = VMUL(LDK(KP707106781), VSUB(T9, Td)); Chris@42: TI = VADD(TG, TH); Chris@42: T1a = VSUB(TH, TG); Chris@42: TT = VADD(T6, T8); Chris@42: TU = VADD(Tb, Tc); Chris@42: TV = VSUB(TT, TU); Chris@42: T1D = VADD(TT, TU); Chris@42: Tg = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Th = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0])); Chris@42: Ti = VCONJ(Th); Chris@42: Tj = VSUB(Tg, Ti); Chris@42: TM = VADD(Tg, Ti); Chris@42: Tt = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tu = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0])); Chris@42: Tv = VCONJ(Tu); Chris@42: Tw = VSUB(Tt, Tv); Chris@42: TQ = VADD(Tt, Tv); Chris@42: Tk = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tl = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0])); Chris@42: Tm = VCONJ(Tl); Chris@42: Tn = VSUB(Tk, Tm); Chris@42: TN = VADD(Tk, Tm); Chris@42: Tr = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)])); Chris@42: Tp = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: Tq = VCONJ(Tp); Chris@42: Ts = VSUB(Tq, Tr); Chris@42: TP = VADD(Tq, Tr); Chris@42: T1F = VADD(TM, TN); Chris@42: T1G = VADD(TP, TQ); Chris@42: To = VFNMS(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tj)); Chris@42: Tx = VFMA(LDK(KP923879532), Ts, VMUL(LDK(KP382683432), Tw)); Chris@42: Ty = VADD(To, Tx); Chris@42: T19 = VSUB(To, Tx); Chris@42: TA = VFMA(LDK(KP382683432), Tj, VMUL(LDK(KP923879532), Tn)); Chris@42: TB = VFNMS(LDK(KP382683432), Ts, VMUL(LDK(KP923879532), Tw)); Chris@42: TC = VADD(TA, TB); Chris@42: T17 = VSUB(TA, TB); Chris@42: TO = VSUB(TM, TN); Chris@42: TR = VSUB(TP, TQ); Chris@42: TS = VMUL(LDK(KP707106781), VSUB(TO, TR)); Chris@42: T10 = VMUL(LDK(KP707106781), VADD(TO, TR)); Chris@42: } Chris@42: { Chris@42: V T21, T1W, T1u, T20, T1I, T1O, TK, T1S, T12, T1e, T1k, T1A, T1o, T1w, T1c; Chris@42: V T1M, T1U, T1V, T1T, T1s, T1t, T1r, T1Z, T1E, T1H, T1B, T1N, Tz, TJ, T1; Chris@42: V T1R, TW, T11, TL, T1d, T1i, T1j, T1h, T1z, T1m, T1n, T1l, T1v, T18, T1b; Chris@42: V T15, T1L, T13, T1g, T1X, T23, T14, T1f, T1Y, T22, T1p, T1y, T1J, T1Q, T1q; Chris@42: V T1x, T1K, T1P; Chris@42: T1U = VADD(T1C, T1D); Chris@42: T1V = VADD(T1F, T1G); Chris@42: T21 = VADD(T1U, T1V); Chris@42: T1T = LDW(&(W[TWVL * 14])); Chris@42: T1W = VZMUL(T1T, VSUB(T1U, T1V)); Chris@42: T1s = VADD(Tf, Ty); Chris@42: T1t = VBYI(VADD(TI, TC)); Chris@42: T1r = LDW(&(W[TWVL * 28])); Chris@42: T1u = VZMULI(T1r, VSUB(T1s, T1t)); Chris@42: T1Z = LDW(&(W[0])); Chris@42: T20 = VZMULI(T1Z, VADD(T1s, T1t)); Chris@42: T1E = VSUB(T1C, T1D); Chris@42: T1H = VBYI(VSUB(T1F, T1G)); Chris@42: T1B = LDW(&(W[TWVL * 22])); Chris@42: T1I = VZMUL(T1B, VSUB(T1E, T1H)); Chris@42: T1N = LDW(&(W[TWVL * 6])); Chris@42: T1O = VZMUL(T1N, VADD(T1E, T1H)); Chris@42: Tz = VSUB(Tf, Ty); Chris@42: TJ = VBYI(VSUB(TC, TI)); Chris@42: T1 = LDW(&(W[TWVL * 12])); Chris@42: TK = VZMULI(T1, VADD(Tz, TJ)); Chris@42: T1R = LDW(&(W[TWVL * 16])); Chris@42: T1S = VZMULI(T1R, VSUB(Tz, TJ)); Chris@42: TW = VBYI(VSUB(TS, TV)); Chris@42: T11 = VSUB(TZ, T10); Chris@42: TL = LDW(&(W[TWVL * 10])); Chris@42: T12 = VZMUL(TL, VADD(TW, T11)); Chris@42: T1d = LDW(&(W[TWVL * 18])); Chris@42: T1e = VZMUL(T1d, VSUB(T11, TW)); Chris@42: T1i = VBYI(VADD(T1a, T19)); Chris@42: T1j = VADD(T16, T17); Chris@42: T1h = LDW(&(W[TWVL * 4])); Chris@42: T1k = VZMULI(T1h, VADD(T1i, T1j)); Chris@42: T1z = LDW(&(W[TWVL * 24])); Chris@42: T1A = VZMULI(T1z, VSUB(T1j, T1i)); Chris@42: T1m = VBYI(VADD(TV, TS)); Chris@42: T1n = VADD(TZ, T10); Chris@42: T1l = LDW(&(W[TWVL * 2])); Chris@42: T1o = VZMUL(T1l, VADD(T1m, T1n)); Chris@42: T1v = LDW(&(W[TWVL * 26])); Chris@42: T1w = VZMUL(T1v, VSUB(T1n, T1m)); Chris@42: T18 = VSUB(T16, T17); Chris@42: T1b = VBYI(VSUB(T19, T1a)); Chris@42: T15 = LDW(&(W[TWVL * 20])); Chris@42: T1c = VZMULI(T15, VSUB(T18, T1b)); Chris@42: T1L = LDW(&(W[TWVL * 8])); Chris@42: T1M = VZMULI(T1L, VADD(T1b, T18)); Chris@42: T13 = VADD(TK, T12); Chris@42: ST(&(Rp[WS(rs, 3)]), T13, ms, &(Rp[WS(rs, 1)])); Chris@42: T1g = VCONJ(VSUB(T1e, T1c)); Chris@42: ST(&(Rm[WS(rs, 5)]), T1g, -ms, &(Rm[WS(rs, 1)])); Chris@42: T1X = VADD(T1S, T1W); Chris@42: ST(&(Rp[WS(rs, 4)]), T1X, ms, &(Rp[0])); Chris@42: T23 = VCONJ(VSUB(T21, T20)); Chris@42: ST(&(Rm[0]), T23, -ms, &(Rm[0])); Chris@42: T14 = VCONJ(VSUB(T12, TK)); Chris@42: ST(&(Rm[WS(rs, 3)]), T14, -ms, &(Rm[WS(rs, 1)])); Chris@42: T1f = VADD(T1c, T1e); Chris@42: ST(&(Rp[WS(rs, 5)]), T1f, ms, &(Rp[WS(rs, 1)])); Chris@42: T1Y = VCONJ(VSUB(T1W, T1S)); Chris@42: ST(&(Rm[WS(rs, 4)]), T1Y, -ms, &(Rm[0])); Chris@42: T22 = VADD(T20, T21); Chris@42: ST(&(Rp[0]), T22, ms, &(Rp[0])); Chris@42: T1p = VADD(T1k, T1o); Chris@42: ST(&(Rp[WS(rs, 1)]), T1p, ms, &(Rp[WS(rs, 1)])); Chris@42: T1y = VCONJ(VSUB(T1w, T1u)); Chris@42: ST(&(Rm[WS(rs, 7)]), T1y, -ms, &(Rm[WS(rs, 1)])); Chris@42: T1J = VADD(T1A, T1I); Chris@42: ST(&(Rp[WS(rs, 6)]), T1J, ms, &(Rp[0])); Chris@42: T1Q = VCONJ(VSUB(T1O, T1M)); Chris@42: ST(&(Rm[WS(rs, 2)]), T1Q, -ms, &(Rm[0])); Chris@42: T1q = VCONJ(VSUB(T1o, T1k)); Chris@42: ST(&(Rm[WS(rs, 1)]), T1q, -ms, &(Rm[WS(rs, 1)])); Chris@42: T1x = VADD(T1u, T1w); Chris@42: ST(&(Rp[WS(rs, 7)]), T1x, ms, &(Rp[WS(rs, 1)])); Chris@42: T1K = VCONJ(VSUB(T1I, T1A)); Chris@42: ST(&(Rm[WS(rs, 6)]), T1K, -ms, &(Rm[0])); Chris@42: T1P = VADD(T1M, T1O); Chris@42: ST(&(Rp[WS(rs, 2)]), T1P, ms, &(Rp[0])); 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: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 16, XSIMD_STRING("hc2cbdftv_16"), twinstr, &GENUS, {99, 38, 4, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cbdftv_16) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cbdftv_16, &desc, HC2C_VIA_DFT); Chris@42: } Chris@42: #endif /* HAVE_FMA */