Chris@10: /*
Chris@10:  * Copyright (c) 2003, 2007-11 Matteo Frigo
Chris@10:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
Chris@10:  *
Chris@10:  * This program is free software; you can redistribute it and/or modify
Chris@10:  * it under the terms of the GNU General Public License as published by
Chris@10:  * the Free Software Foundation; either version 2 of the License, or
Chris@10:  * (at your option) any later version.
Chris@10:  *
Chris@10:  * This program is distributed in the hope that it will be useful,
Chris@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@10:  * GNU General Public License for more details.
Chris@10:  *
Chris@10:  * You should have received a copy of the GNU General Public License
Chris@10:  * along with this program; if not, write to the Free Software
Chris@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@10:  *
Chris@10:  */
Chris@10: 
Chris@10: /* detection of alignment.  This is complicated because a machine may
Chris@10:    support multiple SIMD extensions (e.g. SSE2 and AVX) but only one
Chris@10:    set of alignment contraints.  So this alignment stuff cannot be
Chris@10:    defined in the SIMD header files.  Rather than defining a separate
Chris@10:    set of "machine" header files, we just do this ugly ifdef here. */
Chris@10: #if defined(HAVE_SSE2) || defined(HAVE_AVX)
Chris@10: #  if defined(FFTW_SINGLE)
Chris@10: #    define ALIGNMENT 8     /* Alignment for the LD/ST macros */
Chris@10: #    define ALIGNMENTA 16   /* Alignment for the LDA/STA macros */
Chris@10: #  else
Chris@10: #    define ALIGNMENT 16    /* Alignment for the LD/ST macros */
Chris@10: #    define ALIGNMENTA 16   /* Alignment for the LDA/STA macros */
Chris@10: #  endif
Chris@10: #elif defined(HAVE_ALTIVEC)
Chris@10: #  define ALIGNMENT 8     /* Alignment for the LD/ST macros */
Chris@10: #  define ALIGNMENTA 16   /* Alignment for the LDA/STA macros */
Chris@10: #elif defined(HAVE_NEON)
Chris@10: #  define ALIGNMENT 8     /* Alignment for the LD/ST macros */
Chris@10: #  define ALIGNMENTA 8    /* Alignment for the LDA/STA macros */
Chris@10: #endif
Chris@10: 
Chris@10: #if HAVE_SIMD
Chris@10: #  ifndef ALIGNMENT
Chris@10: #  error "ALIGNMENT not defined"
Chris@10: #  endif
Chris@10: #  ifndef ALIGNMENTA
Chris@10: #  error "ALIGNMENTA not defined"
Chris@10: #  endif
Chris@10: #endif
Chris@10: 
Chris@10: /* rename for precision and for SIMD extensions */
Chris@10: #define XSIMD0(name, suffix) CONCAT(name, suffix)
Chris@10: #define XSIMD(name) XSIMD0(X(name), SIMD_SUFFIX)
Chris@10: #define XSIMD_STRING(x) x STRINGIZE(SIMD_SUFFIX)
Chris@10: 
Chris@10: /* TAINT_BIT is set if pointers are not guaranteed to be multiples of
Chris@10:    ALIGNMENT */
Chris@10: #define TAINT_BIT 1    
Chris@10: 
Chris@10: /* TAINT_BITA is set if pointers are not guaranteed to be multiples of
Chris@10:    ALIGNMENTA */
Chris@10: #define TAINT_BITA 2
Chris@10: 
Chris@10: #define PTRINT(p) ((uintptr_t)(p))
Chris@10: 
Chris@10: #define ALIGNED(p) \
Chris@10:   (((PTRINT(UNTAINT(p)) % ALIGNMENT) == 0) && !(PTRINT(p) & TAINT_BIT))
Chris@10: 
Chris@10: #define ALIGNEDA(p) \
Chris@10:   (((PTRINT(UNTAINT(p)) % ALIGNMENTA) == 0) && !(PTRINT(p) & TAINT_BITA))
Chris@10: 
Chris@10: #define SIMD_STRIDE_OK(x) (!(((x) * sizeof(R)) % ALIGNMENT))
Chris@10: #define SIMD_STRIDE_OKA(x) (!(((x) * sizeof(R)) % ALIGNMENTA))
Chris@10: #define SIMD_VSTRIDE_OK SIMD_STRIDE_OK
Chris@10: