[Bug tree-optimization/65426] New: Recognize byte-swaps when writing to buffers

[fuz at fuz dot su]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65426

            Bug ID: 65426
           Summary: Recognize byte-swaps when writing to buffers
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: minor
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: fuz at fuz dot su

Created attachment 35036
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35036&action=edit
Various instances of the aforementioned pattern

It would be great if gcc could recognize code like this:

    static inline void
    wm32(uint8_t a[4], uint32_t x)
    {
        a[0] = x >> 24;
        a[1] = x >> 16;
        a[2] = x >>  8;
        a[3] = x >>  0;
    }

and turn it into a single unaligned write (and perhaps a byte swap on platforms
where this is supported). This kind of code appears when one tries to write
marshalling code in a portable fashion.

Attached is a file containing various instances of this pattern.