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 08:06:09 EDT 2018 */ cannam@167: cannam@167: #include "dft/codelet-dft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */ cannam@167: cannam@167: /* cannam@167: * This function contains 57 FP additions, 52 FP multiplications, cannam@167: * (or, 39 additions, 34 multiplications, 18 fused multiply/add), cannam@167: * 41 stack variables, 4 constants, and 20 memory accesses cannam@167: */ cannam@167: #include "dft/simd/t3b.h" cannam@167: cannam@167: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@167: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: { cannam@167: INT m; cannam@167: R *x; cannam@167: x = ii; cannam@167: for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(10, rs)) { cannam@167: V T2, T3, T4, Ta, T5, T6, Tt, Td, Th; cannam@167: T2 = LDW(&(W[0])); cannam@167: T3 = LDW(&(W[TWVL * 2])); cannam@167: T4 = VZMUL(T2, T3); cannam@167: Ta = VZMULJ(T2, T3); cannam@167: T5 = LDW(&(W[TWVL * 4])); cannam@167: T6 = VZMULJ(T4, T5); cannam@167: Tt = VZMULJ(T3, T5); cannam@167: Td = VZMULJ(Ta, T5); cannam@167: Th = VZMULJ(T2, T5); cannam@167: { cannam@167: V T9, TJ, Ts, Ty, Tz, TN, TO, TP, Tg, Tm, Tn, TK, TL, TM, T1; cannam@167: V T8, T7; cannam@167: T1 = LD(&(x[0]), ms, &(x[0])); cannam@167: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@167: T8 = VZMUL(T6, T7); cannam@167: T9 = VSUB(T1, T8); cannam@167: TJ = VADD(T1, T8); cannam@167: { cannam@167: V Tp, Tx, Tr, Tv; cannam@167: { cannam@167: V To, Tw, Tq, Tu; cannam@167: To = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@167: Tp = VZMUL(T4, To); cannam@167: Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@167: Tx = VZMUL(T2, Tw); cannam@167: Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@167: Tr = VZMUL(T5, Tq); cannam@167: Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@167: Tv = VZMUL(Tt, Tu); cannam@167: } cannam@167: Ts = VSUB(Tp, Tr); cannam@167: Ty = VSUB(Tv, Tx); cannam@167: Tz = VADD(Ts, Ty); cannam@167: TN = VADD(Tp, Tr); cannam@167: TO = VADD(Tv, Tx); cannam@167: TP = VADD(TN, TO); cannam@167: } cannam@167: { cannam@167: V Tc, Tl, Tf, Tj; cannam@167: { cannam@167: V Tb, Tk, Te, Ti; cannam@167: Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@167: Tc = VZMUL(Ta, Tb); cannam@167: Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@167: Tl = VZMUL(T3, Tk); cannam@167: Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@167: Tf = VZMUL(Td, Te); cannam@167: Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@167: Tj = VZMUL(Th, Ti); cannam@167: } cannam@167: Tg = VSUB(Tc, Tf); cannam@167: Tm = VSUB(Tj, Tl); cannam@167: Tn = VADD(Tg, Tm); cannam@167: TK = VADD(Tc, Tf); cannam@167: TL = VADD(Tj, Tl); cannam@167: TM = VADD(TK, TL); cannam@167: } cannam@167: { cannam@167: V TC, TA, TB, TG, TI, TE, TF, TH, TD; cannam@167: TC = VSUB(Tn, Tz); cannam@167: TA = VADD(Tn, Tz); cannam@167: TB = VFNMS(LDK(KP250000000), TA, T9); cannam@167: TE = VSUB(Tg, Tm); cannam@167: TF = VSUB(Ts, Ty); cannam@167: TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE)); cannam@167: TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF)); cannam@167: ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)])); cannam@167: TH = VFNMS(LDK(KP559016994), TC, TB); cannam@167: ST(&(x[WS(rs, 3)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)])); cannam@167: ST(&(x[WS(rs, 7)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)])); cannam@167: TD = VFMA(LDK(KP559016994), TC, TB); cannam@167: ST(&(x[WS(rs, 1)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)])); cannam@167: ST(&(x[WS(rs, 9)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)])); cannam@167: } cannam@167: { cannam@167: V TS, TQ, TR, TW, TY, TU, TV, TX, TT; cannam@167: TS = VSUB(TM, TP); cannam@167: TQ = VADD(TM, TP); cannam@167: TR = VFNMS(LDK(KP250000000), TQ, TJ); cannam@167: TU = VSUB(TN, TO); cannam@167: TV = VSUB(TK, TL); cannam@167: TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU)); cannam@167: TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV)); cannam@167: ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0])); cannam@167: TX = VFMA(LDK(KP559016994), TS, TR); cannam@167: ST(&(x[WS(rs, 4)]), VFNMSI(TY, TX), ms, &(x[0])); cannam@167: ST(&(x[WS(rs, 6)]), VFMAI(TY, TX), ms, &(x[0])); cannam@167: TT = VFNMS(LDK(KP559016994), TS, TR); cannam@167: ST(&(x[WS(rs, 2)]), VFNMSI(TW, TT), ms, &(x[0])); cannam@167: ST(&(x[WS(rs, 8)]), VFMAI(TW, TT), ms, &(x[0])); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: VTW(0, 1), cannam@167: VTW(0, 3), cannam@167: VTW(0, 9), cannam@167: {TW_NEXT, VL, 0} cannam@167: }; cannam@167: cannam@167: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {39, 34, 18, 0}, 0, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_t3bv_10) (planner *p) { cannam@167: X(kdft_dit_register) (p, t3bv_10, &desc); cannam@167: } cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */ cannam@167: cannam@167: /* cannam@167: * This function contains 57 FP additions, 42 FP multiplications, cannam@167: * (or, 51 additions, 36 multiplications, 6 fused multiply/add), cannam@167: * 41 stack variables, 4 constants, and 20 memory accesses cannam@167: */ cannam@167: #include "dft/simd/t3b.h" cannam@167: cannam@167: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@167: { cannam@167: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@167: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@167: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@167: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@167: { cannam@167: INT m; cannam@167: R *x; cannam@167: x = ii; cannam@167: for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(10, rs)) { cannam@167: V T1, T2, T3, Ti, T6, T7, TA, Tb, To; cannam@167: T1 = LDW(&(W[0])); cannam@167: T2 = LDW(&(W[TWVL * 2])); cannam@167: T3 = VZMULJ(T1, T2); cannam@167: Ti = VZMUL(T1, T2); cannam@167: T6 = LDW(&(W[TWVL * 4])); cannam@167: T7 = VZMULJ(T3, T6); cannam@167: TA = VZMULJ(Ti, T6); cannam@167: Tb = VZMULJ(T1, T6); cannam@167: To = VZMULJ(T2, T6); cannam@167: { cannam@167: V TD, TQ, Tn, Tt, Tx, TM, TN, TS, Ta, Tg, Tw, TJ, TK, TR, Tz; cannam@167: V TC, TB; cannam@167: Tz = LD(&(x[0]), ms, &(x[0])); cannam@167: TB = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@167: TC = VZMUL(TA, TB); cannam@167: TD = VSUB(Tz, TC); cannam@167: TQ = VADD(Tz, TC); cannam@167: { cannam@167: V Tk, Ts, Tm, Tq; cannam@167: { cannam@167: V Tj, Tr, Tl, Tp; cannam@167: Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@167: Tk = VZMUL(Ti, Tj); cannam@167: Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@167: Ts = VZMUL(T1, Tr); cannam@167: Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); cannam@167: Tm = VZMUL(T6, Tl); cannam@167: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); cannam@167: Tq = VZMUL(To, Tp); cannam@167: } cannam@167: Tn = VSUB(Tk, Tm); cannam@167: Tt = VSUB(Tq, Ts); cannam@167: Tx = VADD(Tn, Tt); cannam@167: TM = VADD(Tk, Tm); cannam@167: TN = VADD(Tq, Ts); cannam@167: TS = VADD(TM, TN); cannam@167: } cannam@167: { cannam@167: V T5, Tf, T9, Td; cannam@167: { cannam@167: V T4, Te, T8, Tc; cannam@167: T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@167: T5 = VZMUL(T3, T4); cannam@167: Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@167: Tf = VZMUL(T2, Te); cannam@167: T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); cannam@167: T9 = VZMUL(T7, T8); cannam@167: Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0])); cannam@167: Td = VZMUL(Tb, Tc); cannam@167: } cannam@167: Ta = VSUB(T5, T9); cannam@167: Tg = VSUB(Td, Tf); cannam@167: Tw = VADD(Ta, Tg); cannam@167: TJ = VADD(T5, T9); cannam@167: TK = VADD(Td, Tf); cannam@167: TR = VADD(TJ, TK); cannam@167: } cannam@167: { cannam@167: V Ty, TE, TF, Tv, TI, Th, Tu, TH, TG; cannam@167: Ty = VMUL(LDK(KP559016994), VSUB(Tw, Tx)); cannam@167: TE = VADD(Tw, Tx); cannam@167: TF = VFNMS(LDK(KP250000000), TE, TD); cannam@167: Th = VSUB(Ta, Tg); cannam@167: Tu = VSUB(Tn, Tt); cannam@167: Tv = VBYI(VFMA(LDK(KP951056516), Th, VMUL(LDK(KP587785252), Tu))); cannam@167: TI = VBYI(VFNMS(LDK(KP951056516), Tu, VMUL(LDK(KP587785252), Th))); cannam@167: ST(&(x[WS(rs, 5)]), VADD(TD, TE), ms, &(x[WS(rs, 1)])); cannam@167: TH = VSUB(TF, Ty); cannam@167: ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)])); cannam@167: ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)])); cannam@167: TG = VADD(Ty, TF); cannam@167: ST(&(x[WS(rs, 1)]), VADD(Tv, TG), ms, &(x[WS(rs, 1)])); cannam@167: ST(&(x[WS(rs, 9)]), VSUB(TG, Tv), ms, &(x[WS(rs, 1)])); cannam@167: } cannam@167: { cannam@167: V TV, TT, TU, TP, TY, TL, TO, TX, TW; cannam@167: TV = VMUL(LDK(KP559016994), VSUB(TR, TS)); cannam@167: TT = VADD(TR, TS); cannam@167: TU = VFNMS(LDK(KP250000000), TT, TQ); cannam@167: TL = VSUB(TJ, TK); cannam@167: TO = VSUB(TM, TN); cannam@167: TP = VBYI(VFNMS(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL))); cannam@167: TY = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TO))); cannam@167: ST(&(x[0]), VADD(TQ, TT), ms, &(x[0])); cannam@167: TX = VADD(TV, TU); cannam@167: ST(&(x[WS(rs, 4)]), VSUB(TX, TY), ms, &(x[0])); cannam@167: ST(&(x[WS(rs, 6)]), VADD(TY, TX), ms, &(x[0])); cannam@167: TW = VSUB(TU, TV); cannam@167: ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0])); cannam@167: ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0])); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const tw_instr twinstr[] = { cannam@167: VTW(0, 1), cannam@167: VTW(0, 3), cannam@167: VTW(0, 9), cannam@167: {TW_NEXT, VL, 0} cannam@167: }; cannam@167: cannam@167: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {51, 36, 6, 0}, 0, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_t3bv_10) (planner *p) { cannam@167: X(kdft_dit_register) (p, t3bv_10, &desc); cannam@167: } cannam@167: #endif