cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:38:41 EST 2012 */ cannam@95: cannam@95: #include "codelet-dft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 5 -name t2fv_5 -include t2f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 20 FP additions, 19 FP multiplications, cannam@95: * (or, 11 additions, 10 multiplications, 9 fused multiply/add), cannam@95: * 26 stack variables, 4 constants, and 10 memory accesses cannam@95: */ cannam@95: #include "t2f.h" cannam@95: cannam@95: static void t2fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) { cannam@95: V T1, T2, T9, T4, T7; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T9 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: T4 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: { cannam@95: V T3, Ta, T5, T8; cannam@95: T3 = BYTWJ(&(W[0]), T2); cannam@95: Ta = BYTWJ(&(W[TWVL * 4]), T9); cannam@95: T5 = BYTWJ(&(W[TWVL * 6]), T4); cannam@95: T8 = BYTWJ(&(W[TWVL * 2]), T7); cannam@95: { cannam@95: V T6, Tg, Tb, Th; cannam@95: T6 = VADD(T3, T5); cannam@95: Tg = VSUB(T3, T5); cannam@95: Tb = VADD(T8, Ta); cannam@95: Th = VSUB(T8, Ta); cannam@95: { cannam@95: V Te, Tc, Tk, Ti, Td, Tj, Tf; cannam@95: Te = VSUB(T6, Tb); cannam@95: Tc = VADD(T6, Tb); cannam@95: Tk = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tg, Th)); cannam@95: Ti = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Th, Tg)); cannam@95: Td = VFNMS(LDK(KP250000000), Tc, T1); cannam@95: ST(&(x[0]), VADD(T1, Tc), ms, &(x[0])); cannam@95: Tj = VFNMS(LDK(KP559016994), Te, Td); cannam@95: Tf = VFMA(LDK(KP559016994), Te, Td); cannam@95: ST(&(x[WS(rs, 2)]), VFMAI(Tk, Tj), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 3)]), VFNMSI(Tk, Tj), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VFMAI(Ti, Tf), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 1)]), VFNMSI(Ti, Tf), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 5, XSIMD_STRING("t2fv_5"), twinstr, &GENUS, {11, 10, 9, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t2fv_5) (planner *p) { cannam@95: X(kdft_dit_register) (p, t2fv_5, &desc); cannam@95: } cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -name t2fv_5 -include t2f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 20 FP additions, 14 FP multiplications, cannam@95: * (or, 17 additions, 11 multiplications, 3 fused multiply/add), cannam@95: * 20 stack variables, 4 constants, and 10 memory accesses cannam@95: */ cannam@95: #include "t2f.h" cannam@95: cannam@95: static void t2fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ri; cannam@95: for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) { cannam@95: V Tc, Tg, Th, T5, Ta, Td; cannam@95: Tc = LD(&(x[0]), ms, &(x[0])); cannam@95: { cannam@95: V T2, T9, T4, T7; cannam@95: { cannam@95: V T1, T8, T3, T6; cannam@95: T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T2 = BYTWJ(&(W[0]), T1); cannam@95: T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: T9 = BYTWJ(&(W[TWVL * 4]), T8); cannam@95: T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: T4 = BYTWJ(&(W[TWVL * 6]), T3); cannam@95: T6 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: T7 = BYTWJ(&(W[TWVL * 2]), T6); cannam@95: } cannam@95: Tg = VSUB(T2, T4); cannam@95: Th = VSUB(T7, T9); cannam@95: T5 = VADD(T2, T4); cannam@95: Ta = VADD(T7, T9); cannam@95: Td = VADD(T5, Ta); cannam@95: } cannam@95: ST(&(x[0]), VADD(Tc, Td), ms, &(x[0])); cannam@95: { cannam@95: V Ti, Tj, Tf, Tk, Tb, Te; cannam@95: Ti = VBYI(VFMA(LDK(KP951056516), Tg, VMUL(LDK(KP587785252), Th))); cannam@95: Tj = VBYI(VFNMS(LDK(KP587785252), Tg, VMUL(LDK(KP951056516), Th))); cannam@95: Tb = VMUL(LDK(KP559016994), VSUB(T5, Ta)); cannam@95: Te = VFNMS(LDK(KP250000000), Td, Tc); cannam@95: Tf = VADD(Tb, Te); cannam@95: Tk = VSUB(Te, Tb); cannam@95: ST(&(x[WS(rs, 1)]), VSUB(Tf, Ti), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 3)]), VSUB(Tk, Tj), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VADD(Ti, Tf), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 2)]), VADD(Tj, Tk), ms, &(x[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 5, XSIMD_STRING("t2fv_5"), twinstr, &GENUS, {17, 11, 3, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t2fv_5) (planner *p) { cannam@95: X(kdft_dit_register) (p, t2fv_5, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */