On Sat, 6 Jun 2009, Andy Walls wrote:
> +Alternatively, you can create the unsigned short array dynamically:
> +
> +struct v4l2_subdev *sd = v4l2_i2c_subdev(v4l2_dev, adapter,
> + "module_foo", "chipid", 0, V4L2_I2C_ADDRS(0x10, 0x12));
>
> Strictly speaking, that's not "dynamically" in the sense of the
> generated machine code - everything is going to come from the local
> stack and the initialized data space. The compiler will probably be
> smart enough to generate an unnamed array in the initialized data space
> anyway, avoiding the use of local stack for the array. :)
No such luck, gcc will create an array on the stack and then initialize it
with a series of move word instructions. It isn't even smart enough to
turn:
movw $1, (%esp)
movw $2, 2(%esp)
movw $3, 4(%esp)
movw $-1, 6(%esp)
into:
movl $0x00020001, (%esp)
movl $0xffff0003, 4(%esp)
Now, if you use a different syntax, and change this:
#define V4L2_I2C_ADDRS(addr, addrs...) \
((const unsigned short []){ addr, ## addrs, -1 })
#define bar(addrs...) _bar(V4L2_I2C_ADDRS(addrs))
into this:
#define bar(addr, addrs...) \
({ const unsigned short _a[] = {addr, ## addrs, -1}; _bar(_a); })
If all the values are constants, then for the latter method only gcc will
will create an array in the initialized data segment and use that.
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
