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 16:45:01 EDT 2016 */ Chris@42: Chris@42: #include "codelet-dft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include t3b.h -sign 1 */ Chris@42: Chris@42: /* Chris@42: * This function contains 57 FP additions, 52 FP multiplications, Chris@42: * (or, 39 additions, 34 multiplications, 18 fused multiply/add), Chris@42: * 57 stack variables, 4 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "t3b.h" Chris@42: Chris@42: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@42: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@42: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@42: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@42: { Chris@42: INT m; Chris@42: R *x; Chris@42: x = ii; Chris@42: 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@42: V T1, T7, Th, Tx, Tr, Td, Tp, T6, Tv, Tc, Te, Ti, Tl, T2, T3; Chris@42: V T5; Chris@42: T2 = LDW(&(W[0])); Chris@42: T3 = LDW(&(W[TWVL * 2])); Chris@42: T5 = LDW(&(W[TWVL * 4])); Chris@42: T1 = LD(&(x[0]), ms, &(x[0])); Chris@42: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@42: { Chris@42: V To, Tw, Tq, Tu, Ta, T4, Tt, Tk, Tb; Chris@42: To = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@42: Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@42: Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@42: Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@42: Ta = VZMULJ(T2, T3); Chris@42: T4 = VZMUL(T2, T3); Chris@42: Th = VZMULJ(T2, T5); Chris@42: Tt = VZMULJ(T3, T5); Chris@42: Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@42: Tx = VZMUL(T2, Tw); Chris@42: Tr = VZMUL(T5, Tq); Chris@42: Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@42: Td = VZMULJ(Ta, T5); Chris@42: Tp = VZMUL(T4, To); Chris@42: T6 = VZMULJ(T4, T5); Chris@42: Tv = VZMUL(Tt, Tu); Chris@42: Tc = VZMUL(Ta, Tb); Chris@42: Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@42: Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@42: Tl = VZMUL(T3, Tk); Chris@42: } Chris@42: { Chris@42: V TN, Ts, T8, Ty, TO, Tf, Tj; Chris@42: TN = VADD(Tp, Tr); Chris@42: Ts = VSUB(Tp, Tr); Chris@42: T8 = VZMUL(T6, T7); Chris@42: Ty = VSUB(Tv, Tx); Chris@42: TO = VADD(Tv, Tx); Chris@42: Tf = VZMUL(Td, Te); Chris@42: Tj = VZMUL(Th, Ti); Chris@42: { Chris@42: V T9, TJ, TP, TU, Tz, TF, Tg, TK, Tm, TL; Chris@42: T9 = VSUB(T1, T8); Chris@42: TJ = VADD(T1, T8); Chris@42: TP = VADD(TN, TO); Chris@42: TU = VSUB(TN, TO); Chris@42: Tz = VADD(Ts, Ty); Chris@42: TF = VSUB(Ts, Ty); Chris@42: Tg = VSUB(Tc, Tf); Chris@42: TK = VADD(Tc, Tf); Chris@42: Tm = VSUB(Tj, Tl); Chris@42: TL = VADD(Tj, Tl); Chris@42: { Chris@42: V TM, TV, Tn, TE; Chris@42: TM = VADD(TK, TL); Chris@42: TV = VSUB(TK, TL); Chris@42: Tn = VADD(Tg, Tm); Chris@42: TE = VSUB(Tg, Tm); Chris@42: { Chris@42: V TW, TY, TS, TQ, TG, TI, TC, TA, TR, TB; Chris@42: TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU)); Chris@42: TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV)); Chris@42: TS = VSUB(TM, TP); Chris@42: TQ = VADD(TM, TP); Chris@42: TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE)); Chris@42: TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF)); Chris@42: TC = VSUB(Tn, Tz); Chris@42: TA = VADD(Tn, Tz); Chris@42: ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0])); Chris@42: TR = VFNMS(LDK(KP250000000), TQ, TJ); Chris@42: ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)])); Chris@42: TB = VFNMS(LDK(KP250000000), TA, T9); Chris@42: { Chris@42: V TX, TT, TH, TD; Chris@42: TX = VFMA(LDK(KP559016994), TS, TR); Chris@42: TT = VFNMS(LDK(KP559016994), TS, TR); Chris@42: TH = VFNMS(LDK(KP559016994), TC, TB); Chris@42: TD = VFMA(LDK(KP559016994), TC, TB); Chris@42: ST(&(x[WS(rs, 8)]), VFMAI(TW, TT), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 2)]), VFNMSI(TW, TT), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 6)]), VFMAI(TY, TX), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 4)]), VFNMSI(TY, TX), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 9)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)])); Chris@42: ST(&(x[WS(rs, 1)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)])); Chris@42: ST(&(x[WS(rs, 7)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)])); Chris@42: ST(&(x[WS(rs, 3)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)])); Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(0, 1), Chris@42: VTW(0, 3), Chris@42: VTW(0, 9), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {39, 34, 18, 0}, 0, 0, 0 }; Chris@42: Chris@42: void XSIMD(codelet_t3bv_10) (planner *p) { Chris@42: X(kdft_dit_register) (p, t3bv_10, &desc); Chris@42: } Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* 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 t3b.h -sign 1 */ Chris@42: Chris@42: /* Chris@42: * This function contains 57 FP additions, 42 FP multiplications, Chris@42: * (or, 51 additions, 36 multiplications, 6 fused multiply/add), Chris@42: * 41 stack variables, 4 constants, and 20 memory accesses Chris@42: */ Chris@42: #include "t3b.h" Chris@42: Chris@42: static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@42: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@42: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@42: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@42: { Chris@42: INT m; Chris@42: R *x; Chris@42: x = ii; Chris@42: 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@42: V T1, T2, T3, Ti, T6, T7, TA, Tb, To; Chris@42: T1 = LDW(&(W[0])); Chris@42: T2 = LDW(&(W[TWVL * 2])); Chris@42: T3 = VZMULJ(T1, T2); Chris@42: Ti = VZMUL(T1, T2); Chris@42: T6 = LDW(&(W[TWVL * 4])); Chris@42: T7 = VZMULJ(T3, T6); Chris@42: TA = VZMULJ(Ti, T6); Chris@42: Tb = VZMULJ(T1, T6); Chris@42: To = VZMULJ(T2, T6); Chris@42: { Chris@42: V TD, TQ, Tn, Tt, Tx, TM, TN, TS, Ta, Tg, Tw, TJ, TK, TR, Tz; Chris@42: V TC, TB; Chris@42: Tz = LD(&(x[0]), ms, &(x[0])); Chris@42: TB = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@42: TC = VZMUL(TA, TB); Chris@42: TD = VSUB(Tz, TC); Chris@42: TQ = VADD(Tz, TC); Chris@42: { Chris@42: V Tk, Ts, Tm, Tq; Chris@42: { Chris@42: V Tj, Tr, Tl, Tp; Chris@42: Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@42: Tk = VZMUL(Ti, Tj); Chris@42: Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@42: Ts = VZMUL(T1, Tr); Chris@42: Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Chris@42: Tm = VZMUL(T6, Tl); Chris@42: Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@42: Tq = VZMUL(To, Tp); Chris@42: } Chris@42: Tn = VSUB(Tk, Tm); Chris@42: Tt = VSUB(Tq, Ts); Chris@42: Tx = VADD(Tn, Tt); Chris@42: TM = VADD(Tk, Tm); Chris@42: TN = VADD(Tq, Ts); Chris@42: TS = VADD(TM, TN); Chris@42: } Chris@42: { Chris@42: V T5, Tf, T9, Td; Chris@42: { Chris@42: V T4, Te, T8, Tc; Chris@42: T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@42: T5 = VZMUL(T3, T4); Chris@42: Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@42: Tf = VZMUL(T2, Te); Chris@42: T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@42: T9 = VZMUL(T7, T8); Chris@42: Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Chris@42: Td = VZMUL(Tb, Tc); Chris@42: } Chris@42: Ta = VSUB(T5, T9); Chris@42: Tg = VSUB(Td, Tf); Chris@42: Tw = VADD(Ta, Tg); Chris@42: TJ = VADD(T5, T9); Chris@42: TK = VADD(Td, Tf); Chris@42: TR = VADD(TJ, TK); Chris@42: } Chris@42: { Chris@42: V Ty, TE, TF, Tv, TI, Th, Tu, TH, TG; Chris@42: Ty = VMUL(LDK(KP559016994), VSUB(Tw, Tx)); Chris@42: TE = VADD(Tw, Tx); Chris@42: TF = VFNMS(LDK(KP250000000), TE, TD); Chris@42: Th = VSUB(Ta, Tg); Chris@42: Tu = VSUB(Tn, Tt); Chris@42: Tv = VBYI(VFMA(LDK(KP951056516), Th, VMUL(LDK(KP587785252), Tu))); Chris@42: TI = VBYI(VFNMS(LDK(KP951056516), Tu, VMUL(LDK(KP587785252), Th))); Chris@42: ST(&(x[WS(rs, 5)]), VADD(TD, TE), ms, &(x[WS(rs, 1)])); Chris@42: TH = VSUB(TF, Ty); Chris@42: ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)])); Chris@42: ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)])); Chris@42: TG = VADD(Ty, TF); Chris@42: ST(&(x[WS(rs, 1)]), VADD(Tv, TG), ms, &(x[WS(rs, 1)])); Chris@42: ST(&(x[WS(rs, 9)]), VSUB(TG, Tv), ms, &(x[WS(rs, 1)])); Chris@42: } Chris@42: { Chris@42: V TV, TT, TU, TP, TY, TL, TO, TX, TW; Chris@42: TV = VMUL(LDK(KP559016994), VSUB(TR, TS)); Chris@42: TT = VADD(TR, TS); Chris@42: TU = VFNMS(LDK(KP250000000), TT, TQ); Chris@42: TL = VSUB(TJ, TK); Chris@42: TO = VSUB(TM, TN); Chris@42: TP = VBYI(VFNMS(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL))); Chris@42: TY = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TO))); Chris@42: ST(&(x[0]), VADD(TQ, TT), ms, &(x[0])); Chris@42: TX = VADD(TV, TU); Chris@42: ST(&(x[WS(rs, 4)]), VSUB(TX, TY), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 6)]), VADD(TY, TX), ms, &(x[0])); Chris@42: TW = VSUB(TU, TV); Chris@42: ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0])); Chris@42: ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0])); Chris@42: } Chris@42: } Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(0, 1), Chris@42: VTW(0, 3), Chris@42: VTW(0, 9), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {51, 36, 6, 0}, 0, 0, 0 }; Chris@42: Chris@42: void XSIMD(codelet_t3bv_10) (planner *p) { Chris@42: X(kdft_dit_register) (p, t3bv_10, &desc); Chris@42: } Chris@42: #endif /* HAVE_FMA */