Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:06:09 EDT 2018 */ Chris@82: Chris@82: #include "dft/codelet-dft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* 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 */ Chris@82: Chris@82: /* Chris@82: * This function contains 57 FP additions, 52 FP multiplications, Chris@82: * (or, 39 additions, 34 multiplications, 18 fused multiply/add), Chris@82: * 41 stack variables, 4 constants, and 20 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t3b.h" Chris@82: Chris@82: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ii; Chris@82: 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)) { Chris@82: V T2, T3, T4, Ta, T5, T6, Tt, Td, Th; Chris@82: T2 = LDW(&(W[0])); Chris@82: T3 = LDW(&(W[TWVL * 2])); Chris@82: T4 = VZMUL(T2, T3); Chris@82: Ta = VZMULJ(T2, T3); Chris@82: T5 = LDW(&(W[TWVL * 4])); Chris@82: T6 = VZMULJ(T4, T5); Chris@82: Tt = VZMULJ(T3, T5); Chris@82: Td = VZMULJ(Ta, T5); Chris@82: Th = VZMULJ(T2, T5); Chris@82: { Chris@82: V T9, TJ, Ts, Ty, Tz, TN, TO, TP, Tg, Tm, Tn, TK, TL, TM, T1; Chris@82: V T8, T7; Chris@82: T1 = LD(&(x[0]), ms, &(x[0])); Chris@82: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: T8 = VZMUL(T6, T7); Chris@82: T9 = VSUB(T1, T8); Chris@82: TJ = VADD(T1, T8); Chris@82: { Chris@82: V Tp, Tx, Tr, Tv; Chris@82: { Chris@82: V To, Tw, Tq, Tu; Chris@82: To = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tp = VZMUL(T4, To); Chris@82: Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: Tx = VZMUL(T2, Tw); Chris@82: Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@82: Tr = VZMUL(T5, Tq); Chris@82: Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@82: Tv = VZMUL(Tt, Tu); Chris@82: } Chris@82: Ts = VSUB(Tp, Tr); Chris@82: Ty = VSUB(Tv, Tx); Chris@82: Tz = VADD(Ts, Ty); Chris@82: TN = VADD(Tp, Tr); Chris@82: TO = VADD(Tv, Tx); Chris@82: TP = VADD(TN, TO); Chris@82: } Chris@82: { Chris@82: V Tc, Tl, Tf, Tj; Chris@82: { Chris@82: V Tb, Tk, Te, Ti; Chris@82: Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: Tc = VZMUL(Ta, Tb); Chris@82: Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: Tl = VZMUL(T3, Tk); Chris@82: Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@82: Tf = VZMUL(Td, Te); Chris@82: Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@82: Tj = VZMUL(Th, Ti); Chris@82: } Chris@82: Tg = VSUB(Tc, Tf); Chris@82: Tm = VSUB(Tj, Tl); Chris@82: Tn = VADD(Tg, Tm); Chris@82: TK = VADD(Tc, Tf); Chris@82: TL = VADD(Tj, Tl); Chris@82: TM = VADD(TK, TL); Chris@82: } Chris@82: { Chris@82: V TC, TA, TB, TG, TI, TE, TF, TH, TD; Chris@82: TC = VSUB(Tn, Tz); Chris@82: TA = VADD(Tn, Tz); Chris@82: TB = VFNMS(LDK(KP250000000), TA, T9); Chris@82: TE = VSUB(Tg, Tm); Chris@82: TF = VSUB(Ts, Ty); Chris@82: TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE)); Chris@82: TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF)); Chris@82: ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)])); Chris@82: TH = VFNMS(LDK(KP559016994), TC, TB); Chris@82: ST(&(x[WS(rs, 3)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 7)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)])); Chris@82: TD = VFMA(LDK(KP559016994), TC, TB); Chris@82: ST(&(x[WS(rs, 1)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 9)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)])); Chris@82: } Chris@82: { Chris@82: V TS, TQ, TR, TW, TY, TU, TV, TX, TT; Chris@82: TS = VSUB(TM, TP); Chris@82: TQ = VADD(TM, TP); Chris@82: TR = VFNMS(LDK(KP250000000), TQ, TJ); Chris@82: TU = VSUB(TN, TO); Chris@82: TV = VSUB(TK, TL); Chris@82: TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU)); Chris@82: TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV)); Chris@82: ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0])); Chris@82: TX = VFMA(LDK(KP559016994), TS, TR); Chris@82: ST(&(x[WS(rs, 4)]), VFNMSI(TY, TX), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 6)]), VFMAI(TY, TX), ms, &(x[0])); Chris@82: TT = VFNMS(LDK(KP559016994), TS, TR); Chris@82: ST(&(x[WS(rs, 2)]), VFNMSI(TW, TT), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 8)]), VFMAI(TW, TT), ms, &(x[0])); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const tw_instr twinstr[] = { Chris@82: VTW(0, 1), Chris@82: VTW(0, 3), Chris@82: VTW(0, 9), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {39, 34, 18, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t3bv_10) (planner *p) { Chris@82: X(kdft_dit_register) (p, t3bv_10, &desc); Chris@82: } Chris@82: #else Chris@82: Chris@82: /* 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 */ Chris@82: Chris@82: /* Chris@82: * This function contains 57 FP additions, 42 FP multiplications, Chris@82: * (or, 51 additions, 36 multiplications, 6 fused multiply/add), Chris@82: * 41 stack variables, 4 constants, and 20 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t3b.h" Chris@82: Chris@82: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ii; Chris@82: 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)) { Chris@82: V T1, T2, T3, Ti, T6, T7, TA, Tb, To; Chris@82: T1 = LDW(&(W[0])); Chris@82: T2 = LDW(&(W[TWVL * 2])); Chris@82: T3 = VZMULJ(T1, T2); Chris@82: Ti = VZMUL(T1, T2); Chris@82: T6 = LDW(&(W[TWVL * 4])); Chris@82: T7 = VZMULJ(T3, T6); Chris@82: TA = VZMULJ(Ti, T6); Chris@82: Tb = VZMULJ(T1, T6); Chris@82: To = VZMULJ(T2, T6); Chris@82: { Chris@82: V TD, TQ, Tn, Tt, Tx, TM, TN, TS, Ta, Tg, Tw, TJ, TK, TR, Tz; Chris@82: V TC, TB; Chris@82: Tz = LD(&(x[0]), ms, &(x[0])); Chris@82: TB = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: TC = VZMUL(TA, TB); Chris@82: TD = VSUB(Tz, TC); Chris@82: TQ = VADD(Tz, TC); Chris@82: { Chris@82: V Tk, Ts, Tm, Tq; Chris@82: { Chris@82: V Tj, Tr, Tl, Tp; Chris@82: Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tk = VZMUL(Ti, Tj); Chris@82: Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: Ts = VZMUL(T1, Tr); Chris@82: Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@82: Tm = VZMUL(T6, Tl); Chris@82: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@82: Tq = VZMUL(To, Tp); Chris@82: } Chris@82: Tn = VSUB(Tk, Tm); Chris@82: Tt = VSUB(Tq, Ts); Chris@82: Tx = VADD(Tn, Tt); Chris@82: TM = VADD(Tk, Tm); Chris@82: TN = VADD(Tq, Ts); Chris@82: TS = VADD(TM, TN); Chris@82: } Chris@82: { Chris@82: V T5, Tf, T9, Td; Chris@82: { Chris@82: V T4, Te, T8, Tc; Chris@82: T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: T5 = VZMUL(T3, T4); Chris@82: Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: Tf = VZMUL(T2, Te); Chris@82: T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@82: T9 = VZMUL(T7, T8); Chris@82: Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@82: Td = VZMUL(Tb, Tc); Chris@82: } Chris@82: Ta = VSUB(T5, T9); Chris@82: Tg = VSUB(Td, Tf); Chris@82: Tw = VADD(Ta, Tg); Chris@82: TJ = VADD(T5, T9); Chris@82: TK = VADD(Td, Tf); Chris@82: TR = VADD(TJ, TK); Chris@82: } Chris@82: { Chris@82: V Ty, TE, TF, Tv, TI, Th, Tu, TH, TG; Chris@82: Ty = VMUL(LDK(KP559016994), VSUB(Tw, Tx)); Chris@82: TE = VADD(Tw, Tx); Chris@82: TF = VFNMS(LDK(KP250000000), TE, TD); Chris@82: Th = VSUB(Ta, Tg); Chris@82: Tu = VSUB(Tn, Tt); Chris@82: Tv = VBYI(VFMA(LDK(KP951056516), Th, VMUL(LDK(KP587785252), Tu))); Chris@82: TI = VBYI(VFNMS(LDK(KP951056516), Tu, VMUL(LDK(KP587785252), Th))); Chris@82: ST(&(x[WS(rs, 5)]), VADD(TD, TE), ms, &(x[WS(rs, 1)])); Chris@82: TH = VSUB(TF, Ty); Chris@82: ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)])); Chris@82: TG = VADD(Ty, TF); Chris@82: ST(&(x[WS(rs, 1)]), VADD(Tv, TG), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 9)]), VSUB(TG, Tv), ms, &(x[WS(rs, 1)])); Chris@82: } Chris@82: { Chris@82: V TV, TT, TU, TP, TY, TL, TO, TX, TW; Chris@82: TV = VMUL(LDK(KP559016994), VSUB(TR, TS)); Chris@82: TT = VADD(TR, TS); Chris@82: TU = VFNMS(LDK(KP250000000), TT, TQ); Chris@82: TL = VSUB(TJ, TK); Chris@82: TO = VSUB(TM, TN); Chris@82: TP = VBYI(VFNMS(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL))); Chris@82: TY = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TO))); Chris@82: ST(&(x[0]), VADD(TQ, TT), ms, &(x[0])); Chris@82: TX = VADD(TV, TU); Chris@82: ST(&(x[WS(rs, 4)]), VSUB(TX, TY), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 6)]), VADD(TY, TX), ms, &(x[0])); Chris@82: TW = VSUB(TU, TV); Chris@82: ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0])); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: VLEAVE(); Chris@82: } Chris@82: Chris@82: static const tw_instr twinstr[] = { Chris@82: VTW(0, 1), Chris@82: VTW(0, 3), Chris@82: VTW(0, 9), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {51, 36, 6, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t3bv_10) (planner *p) { Chris@82: X(kdft_dit_register) (p, t3bv_10, &desc); Chris@82: } Chris@82: #endif