Chris@42: /* Chris@42: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@42: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@42: * Chris@42: * This program is free software; you can redistribute it and/or modify Chris@42: * it under the terms of the GNU General Public License as published by Chris@42: * the Free Software Foundation; either version 2 of the License, or Chris@42: * (at your option) any later version. Chris@42: * Chris@42: * This program is distributed in the hope that it will be useful, Chris@42: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@42: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@42: * GNU General Public License for more details. Chris@42: * Chris@42: * You should have received a copy of the GNU General Public License Chris@42: * along with this program; if not, write to the Free Software Chris@42: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@42: * Chris@42: */ Chris@42: Chris@42: /* This file was automatically generated --- DO NOT EDIT */ Chris@42: /* Generated on Sat Jul 30 16:52:40 EDT 2016 */ Chris@42: Chris@42: #include "codelet-rdft.h" Chris@42: Chris@42: #ifdef HAVE_FMA Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 4 -dit -name hc2cfdftv_4 -include hc2cfv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 15 FP additions, 16 FP multiplications, Chris@42: * (or, 9 additions, 10 multiplications, 6 fused multiply/add), Chris@42: * 21 stack variables, 1 constants, and 8 memory accesses Chris@42: */ Chris@42: #include "hc2cfv.h" Chris@42: Chris@42: static void hc2cfdftv_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 6)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(16, rs)) { Chris@42: V T1, T2, Tb, T5, T6, T4, T9, T3, Tc, T7, Ta, Tg, T8, Td, Th; Chris@42: V Tf, Te, Ti, Tj; Chris@42: T1 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: T2 = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: Tb = LDW(&(W[0])); Chris@42: T5 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T6 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T4 = LDW(&(W[TWVL * 2])); Chris@42: T9 = LDW(&(W[TWVL * 4])); Chris@42: T3 = VFMACONJ(T2, T1); Chris@42: Tc = VZMULIJ(Tb, VFNMSCONJ(T2, T1)); Chris@42: T7 = VZMULJ(T4, VFMACONJ(T6, T5)); Chris@42: Ta = VZMULIJ(T9, VFNMSCONJ(T6, T5)); Chris@42: Tg = VADD(T3, T7); Chris@42: T8 = VSUB(T3, T7); Chris@42: Td = VSUB(Ta, Tc); Chris@42: Th = VADD(Tc, Ta); Chris@42: Tf = VCONJ(VMUL(LDK(KP500000000), VFMAI(Td, T8))); Chris@42: Te = VMUL(LDK(KP500000000), VFNMSI(Td, T8)); Chris@42: Ti = VMUL(LDK(KP500000000), VSUB(Tg, Th)); Chris@42: Tj = VCONJ(VMUL(LDK(KP500000000), VADD(Th, Tg))); Chris@42: ST(&(Rm[0]), Tf, -ms, &(Rm[0])); Chris@42: ST(&(Rp[WS(rs, 1)]), Te, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rp[0]), Ti, ms, &(Rp[0])); Chris@42: ST(&(Rm[WS(rs, 1)]), Tj, -ms, &(Rm[WS(rs, 1)])); Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(1, 1), Chris@42: VTW(1, 2), Chris@42: VTW(1, 3), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 4, XSIMD_STRING("hc2cfdftv_4"), twinstr, &GENUS, {9, 10, 6, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cfdftv_4) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cfdftv_4, &desc, HC2C_VIA_DFT); Chris@42: } Chris@42: #else /* HAVE_FMA */ Chris@42: Chris@42: /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 4 -dit -name hc2cfdftv_4 -include hc2cfv.h */ Chris@42: Chris@42: /* Chris@42: * This function contains 15 FP additions, 10 FP multiplications, Chris@42: * (or, 15 additions, 10 multiplications, 0 fused multiply/add), Chris@42: * 23 stack variables, 1 constants, and 8 memory accesses Chris@42: */ Chris@42: #include "hc2cfv.h" Chris@42: Chris@42: static void hc2cfdftv_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms) Chris@42: { Chris@42: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@42: { Chris@42: INT m; Chris@42: for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 6)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(16, rs)) { Chris@42: V T4, Tc, T9, Te, T1, T3, T2, Tb, T6, T8, T7, T5, Td, Tg, Th; Chris@42: V Ta, Tf, Tk, Tl, Ti, Tj; Chris@42: T1 = LD(&(Rp[0]), ms, &(Rp[0])); Chris@42: T2 = LD(&(Rm[0]), -ms, &(Rm[0])); Chris@42: T3 = VCONJ(T2); Chris@42: T4 = VADD(T1, T3); Chris@42: Tb = LDW(&(W[0])); Chris@42: Tc = VZMULIJ(Tb, VSUB(T3, T1)); Chris@42: T6 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)])); Chris@42: T7 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)])); Chris@42: T8 = VCONJ(T7); Chris@42: T5 = LDW(&(W[TWVL * 2])); Chris@42: T9 = VZMULJ(T5, VADD(T6, T8)); Chris@42: Td = LDW(&(W[TWVL * 4])); Chris@42: Te = VZMULIJ(Td, VSUB(T8, T6)); Chris@42: Ta = VSUB(T4, T9); Chris@42: Tf = VBYI(VSUB(Tc, Te)); Chris@42: Tg = VMUL(LDK(KP500000000), VSUB(Ta, Tf)); Chris@42: Th = VCONJ(VMUL(LDK(KP500000000), VADD(Ta, Tf))); Chris@42: ST(&(Rp[WS(rs, 1)]), Tg, ms, &(Rp[WS(rs, 1)])); Chris@42: ST(&(Rm[0]), Th, -ms, &(Rm[0])); Chris@42: Ti = VADD(T4, T9); Chris@42: Tj = VADD(Tc, Te); Chris@42: Tk = VCONJ(VMUL(LDK(KP500000000), VSUB(Ti, Tj))); Chris@42: Tl = VMUL(LDK(KP500000000), VADD(Ti, Tj)); Chris@42: ST(&(Rm[WS(rs, 1)]), Tk, -ms, &(Rm[WS(rs, 1)])); Chris@42: ST(&(Rp[0]), Tl, ms, &(Rp[0])); Chris@42: } Chris@42: } Chris@42: VLEAVE(); Chris@42: } Chris@42: Chris@42: static const tw_instr twinstr[] = { Chris@42: VTW(1, 1), Chris@42: VTW(1, 2), Chris@42: VTW(1, 3), Chris@42: {TW_NEXT, VL, 0} Chris@42: }; Chris@42: Chris@42: static const hc2c_desc desc = { 4, XSIMD_STRING("hc2cfdftv_4"), twinstr, &GENUS, {15, 10, 0, 0} }; Chris@42: Chris@42: void XSIMD(codelet_hc2cfdftv_4) (planner *p) { Chris@42: X(khc2c_register) (p, hc2cfdftv_4, &desc, HC2C_VIA_DFT); Chris@42: } Chris@42: #endif /* HAVE_FMA */