cannam@167: /*
cannam@167:  * Copyright (c) 2003, 2007-14 Matteo Frigo
cannam@167:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
cannam@167:  *
cannam@167:  * This program is free software; you can redistribute it and/or modify
cannam@167:  * it under the terms of the GNU General Public License as published by
cannam@167:  * the Free Software Foundation; either version 2 of the License, or
cannam@167:  * (at your option) any later version.
cannam@167:  *
cannam@167:  * This program is distributed in the hope that it will be useful,
cannam@167:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@167:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
cannam@167:  * GNU General Public License for more details.
cannam@167:  *
cannam@167:  * You should have received a copy of the GNU General Public License
cannam@167:  * along with this program; if not, write to the Free Software
cannam@167:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
cannam@167:  *
cannam@167:  */
cannam@167: 
cannam@167: 
cannam@167: /*
cannam@167:  * This header file must include every file or define every
cannam@167:  * type or macro which is required to compile a codelet.
cannam@167:  */
cannam@167: 
cannam@167: #ifndef __DFT_CODELET_H__
cannam@167: #define __DFT_CODELET_H__
cannam@167: 
cannam@167: #include "kernel/ifftw.h"
cannam@167: 
cannam@167: /**************************************************************
cannam@167:  * types of codelets
cannam@167:  **************************************************************/
cannam@167: 
cannam@167: /* DFT codelets */
cannam@167: typedef struct kdft_desc_s kdft_desc;
cannam@167: 
cannam@167: typedef struct {
cannam@167:      int (*okp)(
cannam@167: 	  const kdft_desc *desc,
cannam@167: 	  const R *ri, const R *ii, const R *ro, const R *io,
cannam@167: 	  INT is, INT os, INT vl, INT ivs, INT ovs,
cannam@167: 	  const planner *plnr);
cannam@167:      INT vl;
cannam@167: } kdft_genus;
cannam@167: 
cannam@167: struct kdft_desc_s {
cannam@167:      INT sz;    /* size of transform computed */
cannam@167:      const char *nam;
cannam@167:      opcnt ops;
cannam@167:      const kdft_genus *genus;
cannam@167:      INT is;
cannam@167:      INT os;
cannam@167:      INT ivs;
cannam@167:      INT ovs;
cannam@167: };
cannam@167: 
cannam@167: typedef void (*kdft) (const R *ri, const R *ii, R *ro, R *io,
cannam@167:                       stride is, stride os, INT vl, INT ivs, INT ovs);
cannam@167: void X(kdft_register)(planner *p, kdft codelet, const kdft_desc *desc);
cannam@167: 
cannam@167: 
cannam@167: typedef struct ct_desc_s ct_desc;
cannam@167: 
cannam@167: typedef struct {
cannam@167:      int (*okp)(
cannam@167: 	  const struct ct_desc_s *desc,
cannam@167: 	  const R *rio, const R *iio, 
cannam@167: 	  INT rs, INT vs, INT m, INT mb, INT me, INT ms,
cannam@167: 	  const planner *plnr);
cannam@167:      INT vl;
cannam@167: } ct_genus;
cannam@167: 
cannam@167: struct ct_desc_s {
cannam@167:      INT radix;
cannam@167:      const char *nam;
cannam@167:      const tw_instr *tw;
cannam@167:      const ct_genus *genus;
cannam@167:      opcnt ops;
cannam@167:      INT rs;
cannam@167:      INT vs;
cannam@167:      INT ms;
cannam@167: };
cannam@167: 
cannam@167: typedef void (*kdftw) (R *rioarray, R *iioarray, const R *W,
cannam@167: 		       stride ios, INT mb, INT me, INT ms);
cannam@167: void X(kdft_dit_register)(planner *p, kdftw codelet, const ct_desc *desc);
cannam@167: void X(kdft_dif_register)(planner *p, kdftw codelet, const ct_desc *desc);
cannam@167: 
cannam@167: 
cannam@167: typedef void (*kdftwsq) (R *rioarray, R *iioarray,
cannam@167: 			 const R *W, stride is, stride vs,
cannam@167: 			 INT mb, INT me, INT ms);
cannam@167: void X(kdft_difsq_register)(planner *p, kdftwsq codelet, const ct_desc *desc);
cannam@167: 
cannam@167: 
cannam@167: extern const solvtab X(solvtab_dft_standard);
cannam@167: extern const solvtab X(solvtab_dft_sse2);
cannam@167: extern const solvtab X(solvtab_dft_avx);
cannam@167: extern const solvtab X(solvtab_dft_avx_128_fma);
cannam@167: extern const solvtab X(solvtab_dft_avx2);
cannam@167: extern const solvtab X(solvtab_dft_avx2_128);
cannam@167: extern const solvtab X(solvtab_dft_avx512);
cannam@167: extern const solvtab X(solvtab_dft_kcvi);
cannam@167: extern const solvtab X(solvtab_dft_altivec);
cannam@167: extern const solvtab X(solvtab_dft_vsx);
cannam@167: extern const solvtab X(solvtab_dft_neon);
cannam@167: extern const solvtab X(solvtab_dft_generic_simd128);
cannam@167: extern const solvtab X(solvtab_dft_generic_simd256);
cannam@167: 
cannam@167: #endif				/* __DFT_CODELET_H__ */