On Sep 4, 2014 10:45 PM, "Matt Turner" <<a href="mailto:mattst88@gmail.com">mattst88@gmail.com</a>> wrote: > > On Thu, Sep 4, 2014 at 10:19 PM, Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> wrote: > > This will, eventually, allow us to manage execution widths of instructions > > in a much more natural way from the fs_visitor level. > > > > Signed-off-by: Jason Ekstrand <<a href="mailto:jason.ekstrand@intel.com">jason.ekstrand@intel.com</a>> > > Adding a width field to fs_inst might be fine, but I don't think I > like adding the parameter to every fs_inst constructor. It seems like > it's something that's rarely going to be non-default. With other > things like that we just set them after creating the instruction. Unfortunately, there's no good way to set a default in fs_inst when the constructor is unaware of the fs_visitor. The default is provided by the fs_visitor::emit functions. This is admittedly a little painful for blorp, but it works out OK for fs_visitor. I tried making a value of 0 mean "choose a default for me" but that didn't work well either. The only other real option would be to guess based on register widths but that gets us to a chicken-and-egg problem where register width depends on instruction width and vice-versa. > Also, am I correct that the width field isn't actually hooked up to > anything at this point? Yes, that's correct. Hooking it up comes later. --Jason