On Fri, Jul 08, 2022 at 11:36:06PM +0800, Tao Zhou wrote:
> On Wed, Jul 06, 2022 at 06:53:21PM +0100, Matthew Wilcox wrote:
>
> > On Thu, Jul 07, 2022 at 01:49:02AM +0800, Tao Zhou wrote:
> > > > +struct rv_monitor {
> > > > + const char *name;
> > > > + const char *description;
> > > > + bool enabled;
> > >
> > > Can the 'bool enabled;' be put at the end like the definition of
> > > structure rv_monitor_def. If '8+8+sizeof(bool)+8+8+8' not the same
> > > as '8+8+8+8+8+sizeof(bool)', I mean is it possible that after the
> > > end of stucture there is a int or char not require to align to 8 as
> > > an example from my nonsense.
> >
> > That will make no difference at all. C doesn't allow other variables
> > to "fill in the hole" at the end of the structure like that. For
> > example, one could legitimately do 'memset(&rvm, sizeof(rvm))',
> > and that would wipe out those other variables as well.
>
> I mean if it is possible that if @enabled placed at the end of the
> structure rv_monitor will save some bytes.
> If @enabled place in between, the next function pointer which is 8 bytes
> will align to be in x8 address and the size of structure rv_monitor is
> larger than been placed at the end of the structure.
> Or the compiler can do magic that I can not guess.
>
> Sorry for my late reply. I am not sure about this. But your reply is not
> about what I mean. You say that the size of structure is the same(I doute about this).
> But what my concert is that the other data next to the structure rv_monitor
> how to align if placing the @enabled at the end.
>
> Place in between, bytes:
> 8+8+8(padd)+8+8+8=48
>
> Place at the end, bytes:
> 8+8+8+8+8+sizeof(bool)=?
>
> ? is small than 48 and the data next to the data structure rv_monitor can use
> the saved byte which is the result of placing @enabled at the end of structure
> to place the data.
You don't need to take my word for it. You can try it yourself.
|