Re: Question on scalar replacement

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    So much for that idea.  Set a breakpoint in tree-sra.c:scalarize_stmt
    and continue until you hit this statement.  Start stepping from there to
    see why we don't rewrite the LHS of that assignment.

Will do in a shortly.  Let me address documentation issues now.


    OK.  Look in tree-sra.c:tree_sra().  If you want, you could just move
    that to the beginning of the file.

That's actually an on piece of documentation because it mentions a lot
of things that are parameters to that function that aren't.

But no, that's really not enough.  The only overview is basically the first
paragraph and that's too short.  I'd expect to see at least a half dozen
paragraphs.  As an example of something that is done properly, see
tree-nested.c.

    Each tree-ssa pass is (should be) documented at its main entry point. 
    Entry points are all declared in the main pass sequencer
    (tree-optimize.c:init_tree_optimization_passes).

Still, the convention is that a block of comments in front of a file should
describe what that file does and I don't think that having to do the above
search should be necessary.