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: 
Chris@10: #include "ifftw.h"
Chris@10: 
Chris@10: /* ``kernel'' malloc(), with proper memory alignment */
Chris@10: 
Chris@10: #if defined(HAVE_DECL_MEMALIGN) && !HAVE_DECL_MEMALIGN
Chris@10: #  if defined(HAVE_MALLOC_H)
Chris@10: #    include <malloc.h>
Chris@10: #  else
Chris@10: extern void *memalign(size_t, size_t);
Chris@10: #  endif
Chris@10: #endif
Chris@10: 
Chris@10: #if defined(HAVE_DECL_POSIX_MEMALIGN) && !HAVE_DECL_POSIX_MEMALIGN
Chris@10: extern int posix_memalign(void **, size_t, size_t);
Chris@10: #endif
Chris@10: 
Chris@10: #if defined(macintosh) /* MacOS 9 */
Chris@10: #  include <Multiprocessing.h>
Chris@10: #endif
Chris@10: 
Chris@10: #define real_free free /* memalign and malloc use ordinary free */
Chris@10: 
Chris@10: #define IS_POWER_OF_TWO(n) (((n) > 0) && (((n) & ((n) - 1)) == 0))
Chris@10: #if defined(WITH_OUR_MALLOC) && (MIN_ALIGNMENT >= 8) && IS_POWER_OF_TWO(MIN_ALIGNMENT)
Chris@10: /* Our own MIN_ALIGNMENT-aligned malloc/free.  Assumes sizeof(void*) is a
Chris@10:    power of two <= 8 and that malloc is at least sizeof(void*)-aligned.
Chris@10: 
Chris@10:    The main reason for this routine is that, as of this writing,
Chris@10:    Windows does not include any aligned allocation routines in its
Chris@10:    system libraries, and instead provides an implementation with a
Chris@10:    Visual C++ "Processor Pack" that you have to statically link into
Chris@10:    your program.  We do not want to require users to have VC++
Chris@10:    (e.g. gcc/MinGW should be fine).  Our code should be at least as good
Chris@10:    as the MS _aligned_malloc, in any case, according to second-hand
Chris@10:    reports of the algorithm it employs (also based on plain malloc). */
Chris@10: static void *our_malloc(size_t n)
Chris@10: {
Chris@10:      void *p0, *p;
Chris@10:      if (!(p0 = malloc(n + MIN_ALIGNMENT))) return (void *) 0;
Chris@10:      p = (void *) (((uintptr_t) p0 + MIN_ALIGNMENT) & (~((uintptr_t) (MIN_ALIGNMENT - 1))));
Chris@10:      *((void **) p - 1) = p0;
Chris@10:      return p;
Chris@10: }
Chris@10: static void our_free(void *p)
Chris@10: {
Chris@10:      if (p) free(*((void **) p - 1));
Chris@10: }
Chris@10: #endif
Chris@10: 
Chris@10: void *X(kernel_malloc)(size_t n)
Chris@10: {
Chris@10:      void *p;
Chris@10: 
Chris@10: #if defined(MIN_ALIGNMENT)
Chris@10: 
Chris@10: #  if defined(WITH_OUR_MALLOC)
Chris@10:      p = our_malloc(n);
Chris@10: #    undef real_free
Chris@10: #    define real_free our_free
Chris@10: 
Chris@10: #  elif defined(__FreeBSD__) && (MIN_ALIGNMENT <= 16)
Chris@10:      /* FreeBSD does not have memalign, but its malloc is 16-byte aligned. */
Chris@10:      p = malloc(n);
Chris@10: 
Chris@10: #  elif (defined(__MACOSX__) || defined(__APPLE__)) && (MIN_ALIGNMENT <= 16)
Chris@10:      /* MacOS X malloc is already 16-byte aligned */
Chris@10:      p = malloc(n);
Chris@10: 
Chris@10: #  elif defined(HAVE_MEMALIGN)
Chris@10:      p = memalign(MIN_ALIGNMENT, n);
Chris@10: 
Chris@10: #  elif defined(HAVE_POSIX_MEMALIGN)
Chris@10:      /* note: posix_memalign is broken in glibc 2.2.5: it constrains
Chris@10: 	the size, not the alignment, to be (power of two) * sizeof(void*).
Chris@10:         The bug seems to have been fixed as of glibc 2.3.1. */
Chris@10:      if (posix_memalign(&p, MIN_ALIGNMENT, n))
Chris@10: 	  p = (void*) 0;
Chris@10: 
Chris@10: #  elif defined(__ICC) || defined(__INTEL_COMPILER) || defined(HAVE__MM_MALLOC)
Chris@10:      /* Intel's C compiler defines _mm_malloc and _mm_free intrinsics */
Chris@10:      p = (void *) _mm_malloc(n, MIN_ALIGNMENT);
Chris@10: #    undef real_free
Chris@10: #    define real_free _mm_free
Chris@10: 
Chris@10: #  elif defined(_MSC_VER)
Chris@10:      /* MS Visual C++ 6.0 with a "Processor Pack" supports SIMD
Chris@10: 	and _aligned_malloc/free (uses malloc.h) */
Chris@10:      p = (void *) _aligned_malloc(n, MIN_ALIGNMENT);
Chris@10: #    undef real_free
Chris@10: #    define real_free _aligned_free
Chris@10: 
Chris@10: #  elif defined(macintosh) /* MacOS 9 */
Chris@10:      p = (void *) MPAllocateAligned(n,
Chris@10: #    if MIN_ALIGNMENT == 8
Chris@10: 				    kMPAllocate8ByteAligned,
Chris@10: #    elif MIN_ALIGNMENT == 16
Chris@10: 				    kMPAllocate16ByteAligned,
Chris@10: #    elif MIN_ALIGNMENT == 32
Chris@10: 				    kMPAllocate32ByteAligned,
Chris@10: #    else
Chris@10: #      error "Unknown alignment for MPAllocateAligned"
Chris@10: #    endif
Chris@10: 				    0);
Chris@10: #    undef real_free
Chris@10: #    define real_free MPFree
Chris@10: 
Chris@10: #  else
Chris@10:      /* Add your machine here and send a patch to fftw@fftw.org 
Chris@10:         or (e.g. for Windows) configure --with-our-malloc */
Chris@10: #    error "Don't know how to malloc() aligned memory ... try configuring --with-our-malloc"
Chris@10: #  endif
Chris@10: 
Chris@10: #else /* !defined(MIN_ALIGNMENT) */
Chris@10:      p = malloc(n);
Chris@10: #endif
Chris@10: 
Chris@10:      return p;
Chris@10: }
Chris@10: 
Chris@10: void X(kernel_free)(void *p)
Chris@10: {
Chris@10:      real_free(p);
Chris@10: }