I did some searching to find out just how far along automatic parallelizing compilers have come. Most of the results showed that, as expected, work centers around determining dependencies within loops, and I wondered if there was an efficient way to do that. I kept coming across the term "affine loop", so I found on page 13 in this powerpoint: http://groups.csail.mit.edu/cag/ps3/lectures/6.189-lecture11-compilers.pdf
that a loop nest is affine if the "loop bounds are integer linear functions of constants, loop constant variables and outer loop indexes" and array accesses follow the same prescription. That will be an interesting thing I'll keep in mind as we continue studying parallel code.
I did some searching to find out just how far along automatic parallelizing compilers have come. Most of the results showed that, as expected, work centers around determining dependencies within loops, and I wondered if there was an efficient way to do that. I kept coming across the term "affine loop", so I found on page 13 in this powerpoint: http://groups.csail.mit.edu/cag/ps3/lectures/6.189-lecture11-compilers.pdf that a loop nest is affine if the "loop bounds are integer linear functions of constants, loop constant variables and outer loop indexes" and array accesses follow the same prescription. That will be an interesting thing I'll keep in mind as we continue studying parallel code.