> David Brown wrote:
> >
> > Hi,
> >
> > I'm using bitfields more and more these days - they are much more
> > debugger-friendly than the traditional #define's and flag bytes, and let
the
> > compiler do more useful error-checking.
> >
> > However, one thing that would make them much nicer would be more
efficient
> > code for assignment. For example,
> >
> > #define true 1
> > #define false 0
> > typedef unsigned char byte;
> > typedef union {
> > struct {
> > byte a : 1; byte b : 1; byte c : 1; padding : 5;
> > };
> > byte raw;
> > } tFlags;
> >
> > tFlags x, y;
> >
> > void test1(void)
> > {
> > x.b = y.c;
> > }
> >
> > void test2(void)
> > {
> > if (y.c) x.b = true; else x.b = false;
> > }
> >
> > The two functions test1() and test2() have the same efffect, except that
> > test1() generates a range of shifts and masking operations, whereas
test2()
> > uses a smaller and faster arrangement of bit tests and jumps. In some
> > cases, the direct bit manipulation is best, such as when the same bit is
> > accessed in two structures. But would it be possible to automatically
> > generate the jump form when that is more efficient?
> >
> How about
>
> void test3(void)
> {
> x.b = (y.c == true);
> }
>
I haven't tried that one, but I don't think it will make any difference
compared to "x.b = y.c" or "x.b = y.c ? true : false;".
