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 08:05:26 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 -n 6 -name t1fv_6 -include dft/simd/t1f.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 23 FP additions, 18 FP multiplications, Chris@82: * (or, 17 additions, 12 multiplications, 6 fused multiply/add), Chris@82: * 19 stack variables, 2 constants, and 12 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t1f.h" Chris@82: Chris@82: static void t1fv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ri; Chris@82: for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) { Chris@82: V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2; Chris@82: T1 = LD(&(x[0]), ms, &(x[0])); Chris@82: T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: T3 = BYTWJ(&(W[TWVL * 4]), T2); Chris@82: T4 = VSUB(T1, T3); Chris@82: Ti = VADD(T1, T3); Chris@82: { Chris@82: V Tb, Td, Ta, Tc; Chris@82: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tb = BYTWJ(&(W[TWVL * 6]), Ta); Chris@82: Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: Td = BYTWJ(&(W[0]), Tc); Chris@82: Te = VSUB(Tb, Td); Chris@82: Tk = VADD(Tb, Td); Chris@82: } Chris@82: { Chris@82: V T6, T8, T5, T7; Chris@82: T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: T6 = BYTWJ(&(W[TWVL * 2]), T5); Chris@82: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: T8 = BYTWJ(&(W[TWVL * 8]), T7); Chris@82: T9 = VSUB(T6, T8); Chris@82: Tj = VADD(T6, T8); Chris@82: } Chris@82: { Chris@82: V Th, Tf, Tg, Tn, Tl, Tm; Chris@82: Th = VMUL(LDK(KP866025403), VSUB(Te, T9)); Chris@82: Tf = VADD(T9, Te); Chris@82: Tg = VFNMS(LDK(KP500000000), Tf, T4); Chris@82: ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)])); Chris@82: Tn = VMUL(LDK(KP866025403), VSUB(Tk, Tj)); Chris@82: Tl = VADD(Tj, Tk); Chris@82: Tm = VFNMS(LDK(KP500000000), Tl, Ti); Chris@82: ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0])); 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, 2), Chris@82: VTW(0, 3), Chris@82: VTW(0, 4), Chris@82: VTW(0, 5), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 6, XSIMD_STRING("t1fv_6"), twinstr, &GENUS, {17, 12, 6, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t1fv_6) (planner *p) { Chris@82: X(kdft_dit_register) (p, t1fv_6, &desc); Chris@82: } Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1fv_6 -include dft/simd/t1f.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 23 FP additions, 14 FP multiplications, Chris@82: * (or, 21 additions, 12 multiplications, 2 fused multiply/add), Chris@82: * 19 stack variables, 2 constants, and 12 memory accesses Chris@82: */ Chris@82: #include "dft/simd/t1f.h" Chris@82: Chris@82: static void t1fv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@82: { Chris@82: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@82: { Chris@82: INT m; Chris@82: R *x; Chris@82: x = ri; Chris@82: for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) { Chris@82: V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2; Chris@82: T1 = LD(&(x[0]), ms, &(x[0])); Chris@82: T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@82: T3 = BYTWJ(&(W[TWVL * 4]), T2); Chris@82: T4 = VSUB(T1, T3); Chris@82: Ti = VADD(T1, T3); Chris@82: { Chris@82: V Tb, Td, Ta, Tc; Chris@82: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@82: Tb = BYTWJ(&(W[TWVL * 6]), Ta); Chris@82: Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@82: Td = BYTWJ(&(W[0]), Tc); Chris@82: Te = VSUB(Tb, Td); Chris@82: Tk = VADD(Tb, Td); Chris@82: } Chris@82: { Chris@82: V T6, T8, T5, T7; Chris@82: T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@82: T6 = BYTWJ(&(W[TWVL * 2]), T5); Chris@82: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Chris@82: T8 = BYTWJ(&(W[TWVL * 8]), T7); Chris@82: T9 = VSUB(T6, T8); Chris@82: Tj = VADD(T6, T8); Chris@82: } Chris@82: { Chris@82: V Th, Tf, Tg, Tn, Tl, Tm; Chris@82: Th = VBYI(VMUL(LDK(KP866025403), VSUB(Te, T9))); Chris@82: Tf = VADD(T9, Te); Chris@82: Tg = VFNMS(LDK(KP500000000), Tf, T4); Chris@82: ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 1)]), VADD(Tg, Th), ms, &(x[WS(rs, 1)])); Chris@82: ST(&(x[WS(rs, 5)]), VSUB(Tg, Th), ms, &(x[WS(rs, 1)])); Chris@82: Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tk, Tj))); Chris@82: Tl = VADD(Tj, Tk); Chris@82: Tm = VFNMS(LDK(KP500000000), Tl, Ti); Chris@82: ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 4)]), VADD(Tm, Tn), ms, &(x[0])); Chris@82: ST(&(x[WS(rs, 2)]), VSUB(Tm, Tn), ms, &(x[0])); 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, 2), Chris@82: VTW(0, 3), Chris@82: VTW(0, 4), Chris@82: VTW(0, 5), Chris@82: {TW_NEXT, VL, 0} Chris@82: }; Chris@82: Chris@82: static const ct_desc desc = { 6, XSIMD_STRING("t1fv_6"), twinstr, &GENUS, {21, 12, 2, 0}, 0, 0, 0 }; Chris@82: Chris@82: void XSIMD(codelet_t1fv_6) (planner *p) { Chris@82: X(kdft_dit_register) (p, t1fv_6, &desc); Chris@82: } Chris@82: #endif