cannam@127: Next: Multi-dimensional Array Format, Previous: Other Important Topics, Up: Other Important Topics [Contents][Index]
cannam@127:SIMD, which stands for “Single Instruction Multiple Data,” is a set of cannam@127: special operations supported by some processors to perform a single cannam@127: operation on several numbers (usually 2 or 4) simultaneously. SIMD cannam@127: floating-point instructions are available on several popular CPUs: cannam@127: SSE/SSE2/AVX/AVX2/AVX512/KCVI on some x86/x86-64 processors, AltiVec and cannam@127: VSX on some POWER/PowerPCs, NEON on some ARM models. FFTW can be cannam@127: compiled to support the SIMD instructions on any of these systems. cannam@127: cannam@127: cannam@127: cannam@127: cannam@127: cannam@127: cannam@127: cannam@127: cannam@127: cannam@127:
cannam@127: cannam@127:A program linking to an FFTW library compiled with SIMD support can
cannam@127: obtain a nonnegligible speedup for most complex and r2c/c2r
cannam@127: transforms. In order to obtain this speedup, however, the arrays of
cannam@127: complex (or real) data passed to FFTW must be specially aligned in
cannam@127: memory (typically 16-byte aligned), and often this alignment is more
cannam@127: stringent than that provided by the usual malloc (etc.)
cannam@127: allocation routines.
cannam@127:
In order to guarantee proper alignment for SIMD, therefore, in case
cannam@127: your program is ever linked against a SIMD-using FFTW, we recommend
cannam@127: allocating your transform data with fftw_malloc and
cannam@127: de-allocating it with fftw_free.
cannam@127:
cannam@127:
cannam@127: These have exactly the same interface and behavior as
cannam@127: malloc/free, except that for a SIMD FFTW they ensure
cannam@127: that the returned pointer has the necessary alignment (by calling
cannam@127: memalign or its equivalent on your OS).
cannam@127:
You are not required to use fftw_malloc. You can
cannam@127: allocate your data in any way that you like, from malloc to
cannam@127: new (in C++) to a fixed-size array declaration. If the array
cannam@127: happens not to be properly aligned, FFTW will not use the SIMD
cannam@127: extensions.
cannam@127:
cannam@127:
Since fftw_malloc only ever needs to be used for real and
cannam@127: complex arrays, we provide two convenient wrapper routines
cannam@127: fftw_alloc_real(N) and fftw_alloc_complex(N) that are
cannam@127: equivalent to (double*)fftw_malloc(sizeof(double) * N) and
cannam@127: (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * N),
cannam@127: respectively (or their equivalents in other precisions).
cannam@127:
cannam@127: Next: Multi-dimensional Array Format, Previous: Other Important Topics, Up: Other Important Topics [Contents][Index]
cannam@127: