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 07:38:54 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3fv_5 -include t3f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 22 FP additions, 23 FP multiplications, cannam@95: * (or, 13 additions, 14 multiplications, 9 fused multiply/add), cannam@95: * 30 stack variables, 4 constants, and 10 memory accesses cannam@95: */ cannam@95: #include "t3f.h" cannam@95: cannam@95: static void t3fv_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) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(5, rs)) { cannam@95: V T2, T5, T1, T3, Td, T7, Tb; cannam@95: T2 = LDW(&(W[0])); cannam@95: T5 = LDW(&(W[TWVL * 2])); cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T3 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: T7 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: { cannam@95: V Ta, T6, T4, Te, Tc, T8; cannam@95: Ta = VZMULJ(T2, T5); cannam@95: T6 = VZMUL(T2, T5); cannam@95: T4 = VZMULJ(T2, T3); cannam@95: Te = VZMULJ(T5, Td); cannam@95: Tc = VZMULJ(Ta, Tb); cannam@95: T8 = VZMULJ(T6, T7); cannam@95: { cannam@95: V Tf, Tl, T9, Tk; cannam@95: Tf = VADD(Tc, Te); cannam@95: Tl = VSUB(Tc, Te); cannam@95: T9 = VADD(T4, T8); cannam@95: Tk = VSUB(T4, T8); cannam@95: { cannam@95: V Ti, Tg, To, Tm, Th, Tn, Tj; cannam@95: Ti = VSUB(T9, Tf); cannam@95: Tg = VADD(T9, Tf); cannam@95: To = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tk, Tl)); cannam@95: Tm = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tl, Tk)); cannam@95: Th = VFNMS(LDK(KP250000000), Tg, T1); cannam@95: ST(&(x[0]), VADD(T1, Tg), ms, &(x[0])); cannam@95: Tn = VFNMS(LDK(KP559016994), Ti, Th); cannam@95: Tj = VFMA(LDK(KP559016994), Ti, Th); cannam@95: ST(&(x[WS(rs, 2)]), VFMAI(To, Tn), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 3)]), VFNMSI(To, Tn), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VFMAI(Tm, Tj), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 1)]), VFNMSI(Tm, Tj), 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, 3), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 5, XSIMD_STRING("t3fv_5"), twinstr, &GENUS, {13, 14, 9, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t3fv_5) (planner *p) { cannam@95: X(kdft_dit_register) (p, t3fv_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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3fv_5 -include t3f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 22 FP additions, 18 FP multiplications, cannam@95: * (or, 19 additions, 15 multiplications, 3 fused multiply/add), cannam@95: * 24 stack variables, 4 constants, and 10 memory accesses cannam@95: */ cannam@95: #include "t3f.h" cannam@95: cannam@95: static void t3fv_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) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(5, rs)) { cannam@95: V T1, T4, T5, T9; cannam@95: T1 = LDW(&(W[0])); cannam@95: T4 = LDW(&(W[TWVL * 2])); cannam@95: T5 = VZMUL(T1, T4); cannam@95: T9 = VZMULJ(T1, T4); cannam@95: { cannam@95: V Tg, Tk, Tl, T8, Te, Th; cannam@95: Tg = LD(&(x[0]), ms, &(x[0])); cannam@95: { cannam@95: V T3, Td, T7, Tb; cannam@95: { cannam@95: V T2, Tc, T6, Ta; cannam@95: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T3 = VZMULJ(T1, T2); cannam@95: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: Td = VZMULJ(T4, Tc); cannam@95: T6 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: T7 = VZMULJ(T5, T6); cannam@95: Ta = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: Tb = VZMULJ(T9, Ta); cannam@95: } cannam@95: Tk = VSUB(T3, T7); cannam@95: Tl = VSUB(Tb, Td); cannam@95: T8 = VADD(T3, T7); cannam@95: Te = VADD(Tb, Td); cannam@95: Th = VADD(T8, Te); cannam@95: } cannam@95: ST(&(x[0]), VADD(Tg, Th), ms, &(x[0])); cannam@95: { cannam@95: V Tm, Tn, Tj, To, Tf, Ti; cannam@95: Tm = VBYI(VFMA(LDK(KP951056516), Tk, VMUL(LDK(KP587785252), Tl))); cannam@95: Tn = VBYI(VFNMS(LDK(KP587785252), Tk, VMUL(LDK(KP951056516), Tl))); cannam@95: Tf = VMUL(LDK(KP559016994), VSUB(T8, Te)); cannam@95: Ti = VFNMS(LDK(KP250000000), Th, Tg); cannam@95: Tj = VADD(Tf, Ti); cannam@95: To = VSUB(Ti, Tf); cannam@95: ST(&(x[WS(rs, 1)]), VSUB(Tj, Tm), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 3)]), VSUB(To, Tn), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 4)]), VADD(Tm, Tj), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 2)]), VADD(Tn, To), ms, &(x[0])); 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, 3), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 5, XSIMD_STRING("t3fv_5"), twinstr, &GENUS, {19, 15, 3, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t3fv_5) (planner *p) { cannam@95: X(kdft_dit_register) (p, t3fv_5, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */