c@174: If you are reading this, it means you think you may be interested in using the SIMD extensions in kissfft 
c@174: to do 4 *separate* FFTs at once.
c@174: 
c@174: Beware! Beyond here there be dragons!
c@174: 
c@174: This API is not easy to use, is not well documented, and breaks the KISS principle.  
c@174: 
c@174: 
c@174: Still reading? Okay, you may get rewarded for your patience with a considerable speedup 
c@174: (2-3x) on intel x86 machines with SSE if you are willing to jump through some hoops.
c@174: 
c@174: The basic idea is to use the packed 4 float __m128 data type as a scalar element.  
c@174: This means that the format is pretty convoluted. It performs 4 FFTs per fft call on signals A,B,C,D.
c@174: 
c@174: For complex data, the data is interlaced as follows:
c@174: rA0,rB0,rC0,rD0,      iA0,iB0,iC0,iD0,   rA1,rB1,rC1,rD1, iA1,iB1,iC1,iD1 ...
c@174: where "rA0" is the real part of the zeroth sample for signal A
c@174: 
c@174: Real-only data is laid out:
c@174: rA0,rB0,rC0,rD0,     rA1,rB1,rC1,rD1,      ... 
c@174: 
c@174: Compile with gcc flags something like
c@174: -O3 -mpreferred-stack-boundary=4  -DUSE_SIMD=1 -msse 
c@174: 
c@174: Be aware of SIMD alignment.  This is the most likely cause of segfaults.  
c@174: The code within kissfft uses scratch variables on the stack.  
c@174: With SIMD, these must have addresses on 16 byte boundaries.  
c@174: Search on "SIMD alignment" for more info.
c@174: 
c@174: 
c@174: 
c@174: Robin at Divide Concept was kind enough to share his code for formatting to/from the SIMD kissfft.  
c@174: I have not run it -- use it at your own risk.  It appears to do 4xN and Nx4 transpositions 
c@174: (out of place).
c@174: 
c@174: void SSETools::pack128(float* target, float* source, unsigned long size128)
c@174: {
c@174:    __m128* pDest = (__m128*)target;
c@174:    __m128* pDestEnd = pDest+size128;
c@174:    float* source0=source;
c@174:    float* source1=source0+size128;
c@174:    float* source2=source1+size128;
c@174:    float* source3=source2+size128;
c@174: 
c@174:    while(pDest<pDestEnd)
c@174:    {
c@174:        *pDest=_mm_set_ps(*source3,*source2,*source1,*source0);
c@174:        source0++;
c@174:        source1++;
c@174:        source2++;
c@174:        source3++;
c@174:        pDest++;
c@174:    }
c@174: }
c@174: 
c@174: void SSETools::unpack128(float* target, float* source, unsigned long size128)
c@174: {
c@174: 
c@174:    float* pSrc = source;
c@174:    float* pSrcEnd = pSrc+size128*4;
c@174:    float* target0=target;
c@174:    float* target1=target0+size128;
c@174:    float* target2=target1+size128;
c@174:    float* target3=target2+size128;
c@174: 
c@174:    while(pSrc<pSrcEnd)
c@174:    {
c@174:        *target0=pSrc[0];
c@174:        *target1=pSrc[1];
c@174:        *target2=pSrc[2];
c@174:        *target3=pSrc[3];
c@174:        target0++;
c@174:        target1++;
c@174:        target2++;
c@174:        target3++;
c@174:        pSrc+=4;
c@174:    }
c@174: }