Chris@19: /* Chris@19: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@19: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@19: * Chris@19: * This program is free software; you can redistribute it and/or modify Chris@19: * it under the terms of the GNU General Public License as published by Chris@19: * the Free Software Foundation; either version 2 of the License, or Chris@19: * (at your option) any later version. Chris@19: * Chris@19: * This program is distributed in the hope that it will be useful, Chris@19: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@19: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@19: * GNU General Public License for more details. Chris@19: * Chris@19: * You should have received a copy of the GNU General Public License Chris@19: * along with this program; if not, write to the Free Software Chris@19: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@19: * Chris@19: */ Chris@19: Chris@19: /* This file was automatically generated --- DO NOT EDIT */ Chris@19: /* Generated on Tue Mar 4 13:47:14 EST 2014 */ Chris@19: Chris@19: #include "codelet-dft.h" Chris@19: Chris@19: #ifdef HAVE_FMA Chris@19: Chris@19: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1fv_8 -include t1f.h */ Chris@19: Chris@19: /* Chris@19: * This function contains 33 FP additions, 24 FP multiplications, Chris@19: * (or, 23 additions, 14 multiplications, 10 fused multiply/add), Chris@19: * 36 stack variables, 1 constants, and 16 memory accesses Chris@19: */ Chris@19: #include "t1f.h" Chris@19: Chris@19: static void t1fv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@19: { Chris@19: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@19: { Chris@19: INT m; Chris@19: R *x; Chris@19: x = ri; Chris@19: for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) { Chris@19: V T1, T2, Th, Tj, T5, T7, Ta, Tc; Chris@19: T1 = LD(&(x[0]), ms, &(x[0])); Chris@19: T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@19: Th = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@19: Tj = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@19: T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@19: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@19: Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@19: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@19: { Chris@19: V T3, Ti, Tk, T6, T8, Tb, Td; Chris@19: T3 = BYTWJ(&(W[TWVL * 6]), T2); Chris@19: Ti = BYTWJ(&(W[TWVL * 2]), Th); Chris@19: Tk = BYTWJ(&(W[TWVL * 10]), Tj); Chris@19: T6 = BYTWJ(&(W[0]), T5); Chris@19: T8 = BYTWJ(&(W[TWVL * 8]), T7); Chris@19: Tb = BYTWJ(&(W[TWVL * 12]), Ta); Chris@19: Td = BYTWJ(&(W[TWVL * 4]), Tc); Chris@19: { Chris@19: V Tq, T4, Tr, Tl, Tt, T9, Tu, Te, Tw, Ts; Chris@19: Tq = VADD(T1, T3); Chris@19: T4 = VSUB(T1, T3); Chris@19: Tr = VADD(Ti, Tk); Chris@19: Tl = VSUB(Ti, Tk); Chris@19: Tt = VADD(T6, T8); Chris@19: T9 = VSUB(T6, T8); Chris@19: Tu = VADD(Tb, Td); Chris@19: Te = VSUB(Tb, Td); Chris@19: Tw = VSUB(Tq, Tr); Chris@19: Ts = VADD(Tq, Tr); Chris@19: { Chris@19: V Tx, Tv, Tm, Tf; Chris@19: Tx = VSUB(Tu, Tt); Chris@19: Tv = VADD(Tt, Tu); Chris@19: Tm = VSUB(Te, T9); Chris@19: Tf = VADD(T9, Te); Chris@19: { Chris@19: V Tp, Tn, To, Tg; Chris@19: ST(&(x[WS(rs, 2)]), VFMAI(Tx, Tw), ms, &(x[0])); Chris@19: ST(&(x[WS(rs, 6)]), VFNMSI(Tx, Tw), ms, &(x[0])); Chris@19: ST(&(x[0]), VADD(Ts, Tv), ms, &(x[0])); Chris@19: ST(&(x[WS(rs, 4)]), VSUB(Ts, Tv), ms, &(x[0])); Chris@19: Tp = VFMA(LDK(KP707106781), Tm, Tl); Chris@19: Tn = VFNMS(LDK(KP707106781), Tm, Tl); Chris@19: To = VFNMS(LDK(KP707106781), Tf, T4); Chris@19: Tg = VFMA(LDK(KP707106781), Tf, T4); Chris@19: ST(&(x[WS(rs, 5)]), VFNMSI(Tp, To), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 3)]), VFMAI(Tp, To), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 7)]), VFMAI(Tn, Tg), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 1)]), VFNMSI(Tn, Tg), ms, &(x[WS(rs, 1)])); Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: VLEAVE(); Chris@19: } Chris@19: Chris@19: static const tw_instr twinstr[] = { Chris@19: VTW(0, 1), Chris@19: VTW(0, 2), Chris@19: VTW(0, 3), Chris@19: VTW(0, 4), Chris@19: VTW(0, 5), Chris@19: VTW(0, 6), Chris@19: VTW(0, 7), Chris@19: {TW_NEXT, VL, 0} Chris@19: }; Chris@19: Chris@19: static const ct_desc desc = { 8, XSIMD_STRING("t1fv_8"), twinstr, &GENUS, {23, 14, 10, 0}, 0, 0, 0 }; Chris@19: Chris@19: void XSIMD(codelet_t1fv_8) (planner *p) { Chris@19: X(kdft_dit_register) (p, t1fv_8, &desc); Chris@19: } Chris@19: #else /* HAVE_FMA */ Chris@19: Chris@19: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1fv_8 -include t1f.h */ Chris@19: Chris@19: /* Chris@19: * This function contains 33 FP additions, 16 FP multiplications, Chris@19: * (or, 33 additions, 16 multiplications, 0 fused multiply/add), Chris@19: * 24 stack variables, 1 constants, and 16 memory accesses Chris@19: */ Chris@19: #include "t1f.h" Chris@19: Chris@19: static void t1fv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@19: { Chris@19: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); Chris@19: { Chris@19: INT m; Chris@19: R *x; Chris@19: x = ri; Chris@19: for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) { Chris@19: V T4, Tq, Tm, Tr, T9, Tt, Te, Tu, T1, T3, T2; Chris@19: T1 = LD(&(x[0]), ms, &(x[0])); Chris@19: T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@19: T3 = BYTWJ(&(W[TWVL * 6]), T2); Chris@19: T4 = VSUB(T1, T3); Chris@19: Tq = VADD(T1, T3); Chris@19: { Chris@19: V Tj, Tl, Ti, Tk; Chris@19: Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@19: Tj = BYTWJ(&(W[TWVL * 2]), Ti); Chris@19: Tk = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Chris@19: Tl = BYTWJ(&(W[TWVL * 10]), Tk); Chris@19: Tm = VSUB(Tj, Tl); Chris@19: Tr = VADD(Tj, Tl); Chris@19: } Chris@19: { Chris@19: V T6, T8, T5, T7; Chris@19: T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@19: T6 = BYTWJ(&(W[0]), T5); Chris@19: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@19: T8 = BYTWJ(&(W[TWVL * 8]), T7); Chris@19: T9 = VSUB(T6, T8); Chris@19: Tt = VADD(T6, T8); Chris@19: } Chris@19: { Chris@19: V Tb, Td, Ta, Tc; Chris@19: Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Chris@19: Tb = BYTWJ(&(W[TWVL * 12]), Ta); Chris@19: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@19: Td = BYTWJ(&(W[TWVL * 4]), Tc); Chris@19: Te = VSUB(Tb, Td); Chris@19: Tu = VADD(Tb, Td); Chris@19: } Chris@19: { Chris@19: V Ts, Tv, Tw, Tx; Chris@19: Ts = VADD(Tq, Tr); Chris@19: Tv = VADD(Tt, Tu); Chris@19: ST(&(x[WS(rs, 4)]), VSUB(Ts, Tv), ms, &(x[0])); Chris@19: ST(&(x[0]), VADD(Ts, Tv), ms, &(x[0])); Chris@19: Tw = VSUB(Tq, Tr); Chris@19: Tx = VBYI(VSUB(Tu, Tt)); Chris@19: ST(&(x[WS(rs, 6)]), VSUB(Tw, Tx), ms, &(x[0])); Chris@19: ST(&(x[WS(rs, 2)]), VADD(Tw, Tx), ms, &(x[0])); Chris@19: { Chris@19: V Tg, To, Tn, Tp, Tf, Th; Chris@19: Tf = VMUL(LDK(KP707106781), VADD(T9, Te)); Chris@19: Tg = VADD(T4, Tf); Chris@19: To = VSUB(T4, Tf); Chris@19: Th = VMUL(LDK(KP707106781), VSUB(Te, T9)); Chris@19: Tn = VBYI(VSUB(Th, Tm)); Chris@19: Tp = VBYI(VADD(Tm, Th)); Chris@19: ST(&(x[WS(rs, 7)]), VSUB(Tg, Tn), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 1)]), VADD(Tg, Tn), ms, &(x[WS(rs, 1)])); Chris@19: ST(&(x[WS(rs, 5)]), VSUB(To, Tp), ms, &(x[WS(rs, 1)])); Chris@19: } Chris@19: } Chris@19: } Chris@19: } Chris@19: VLEAVE(); Chris@19: } Chris@19: Chris@19: static const tw_instr twinstr[] = { Chris@19: VTW(0, 1), Chris@19: VTW(0, 2), Chris@19: VTW(0, 3), Chris@19: VTW(0, 4), Chris@19: VTW(0, 5), Chris@19: VTW(0, 6), Chris@19: VTW(0, 7), Chris@19: {TW_NEXT, VL, 0} Chris@19: }; Chris@19: Chris@19: static const ct_desc desc = { 8, XSIMD_STRING("t1fv_8"), twinstr, &GENUS, {33, 16, 0, 0}, 0, 0, 0 }; Chris@19: Chris@19: void XSIMD(codelet_t1fv_8) (planner *p) { Chris@19: X(kdft_dit_register) (p, t1fv_8, &desc); Chris@19: } Chris@19: #endif /* HAVE_FMA */