cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:44:45 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* 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 4 -name t3bv_4 -include t3b.h -sign 1 */ cannam@127: cannam@127: /* cannam@127: * This function contains 12 FP additions, 10 FP multiplications, cannam@127: * (or, 10 additions, 8 multiplications, 2 fused multiply/add), cannam@127: * 16 stack variables, 0 constants, and 8 memory accesses cannam@127: */ cannam@127: #include "t3b.h" cannam@127: cannam@127: static void t3bv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ii; cannam@127: for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(4, rs)) { cannam@127: V T2, T3, T1, Ta, T5, T8; cannam@127: T2 = LDW(&(W[0])); cannam@127: T3 = LDW(&(W[TWVL * 2])); cannam@127: T1 = LD(&(x[0]), ms, &(x[0])); cannam@127: Ta = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: { cannam@127: V T4, Tb, T9, T6; cannam@127: T4 = VZMULJ(T2, T3); cannam@127: Tb = VZMUL(T3, Ta); cannam@127: T9 = VZMUL(T2, T8); cannam@127: T6 = VZMUL(T4, T5); cannam@127: { cannam@127: V Tc, Te, T7, Td; cannam@127: Tc = VSUB(T9, Tb); cannam@127: Te = VADD(T9, Tb); cannam@127: T7 = VSUB(T1, T6); cannam@127: Td = VADD(T1, T6); cannam@127: ST(&(x[0]), VADD(Td, Te), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 2)]), VSUB(Td, Te), ms, &(x[0])); cannam@127: ST(&(x[WS(rs, 1)]), VFMAI(Tc, T7), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 3)]), VFNMSI(Tc, T7), ms, &(x[WS(rs, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 3), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 4, XSIMD_STRING("t3bv_4"), twinstr, &GENUS, {10, 8, 2, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t3bv_4) (planner *p) { cannam@127: X(kdft_dit_register) (p, t3bv_4, &desc); cannam@127: } cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 4 -name t3bv_4 -include t3b.h -sign 1 */ cannam@127: cannam@127: /* cannam@127: * This function contains 12 FP additions, 8 FP multiplications, cannam@127: * (or, 12 additions, 8 multiplications, 0 fused multiply/add), cannam@127: * 16 stack variables, 0 constants, and 8 memory accesses cannam@127: */ cannam@127: #include "t3b.h" cannam@127: cannam@127: static void t3bv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@127: { cannam@127: { cannam@127: INT m; cannam@127: R *x; cannam@127: x = ii; cannam@127: for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(4, rs)) { cannam@127: V T2, T3, T4; cannam@127: T2 = LDW(&(W[0])); cannam@127: T3 = LDW(&(W[TWVL * 2])); cannam@127: T4 = VZMULJ(T2, T3); cannam@127: { cannam@127: V T1, Tb, T6, T9, Ta, T5, T8; cannam@127: T1 = LD(&(x[0]), ms, &(x[0])); cannam@127: Ta = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@127: Tb = VZMUL(T3, Ta); cannam@127: T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@127: T6 = VZMUL(T4, T5); cannam@127: T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@127: T9 = VZMUL(T2, T8); cannam@127: { cannam@127: V T7, Tc, Td, Te; cannam@127: T7 = VSUB(T1, T6); cannam@127: Tc = VBYI(VSUB(T9, Tb)); cannam@127: ST(&(x[WS(rs, 3)]), VSUB(T7, Tc), ms, &(x[WS(rs, 1)])); cannam@127: ST(&(x[WS(rs, 1)]), VADD(T7, Tc), ms, &(x[WS(rs, 1)])); cannam@127: Td = VADD(T1, T6); cannam@127: Te = VADD(T9, Tb); cannam@127: ST(&(x[WS(rs, 2)]), VSUB(Td, Te), ms, &(x[0])); cannam@127: ST(&(x[0]), VADD(Td, Te), ms, &(x[0])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const tw_instr twinstr[] = { cannam@127: VTW(0, 1), cannam@127: VTW(0, 3), cannam@127: {TW_NEXT, VL, 0} cannam@127: }; cannam@127: cannam@127: static const ct_desc desc = { 4, XSIMD_STRING("t3bv_4"), twinstr, &GENUS, {12, 8, 0, 0}, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_t3bv_4) (planner *p) { cannam@127: X(kdft_dit_register) (p, t3bv_4, &desc); cannam@127: } cannam@127: #endif /* HAVE_FMA */