Re: [PATCH]: Merge improvements from new-regalloc-branch: df.[ch] [1/2]

[Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>]

Dale Johannesen writes:

 > I'm in the process of working through a bug on ppc that involves this:
 > 
 > (set (zero_extract:SI (subreg:SI (reg/v:DI 136) 0)
 >              (const_int 8 [0x8])
 >              (const_int 0 [0x0]))
 >          (reg:SI 290)) 106 {insvsi}
 > 
 > which is setting a 1-byte field of an 8-byte (DI) struct.  You imply 
 > this is invalid; if that's true I'd certainly like to know it.

No, I believe this is valid.  Where did I imply it wasn't?  The
zero_extract operation is read-modify-write.  The subreg says which
register to access.

Michael.