[PATCH] add VRP for bitwise OR and AND: v3

Denis Vlasenko vda.linux@googlemail.com
Sun Aug 13 14:59:00 GMT 2006


On Saturday 12 August 2006 16:38, Richard Guenther wrote:
> On 8/12/06, Denis Vlasenko <vda.linux@googlemail.com> wrote:
> > * tree-vrp.c (extract_range_from_binary_expr):
> >   add value range propagation for bitwise AND/OR
> >
> > Patch is against gcc-4.2-20060805.
> > Since unmodified gcc-4.2-20060805 snapshot does not bootstrap
> > for me, I only verified that "make all-gcc" build
> > of gcc-4.2-20060805 on i386-pc-linux-gnu
> > with this patch succeeds.
> >
> > However, I successfully bootstrapped gcc-4.1.1 with this patch.
> 
> +/* Combines MIN and MAX as follows:
> +   MIN =   01010010101001010010
> +   MAX =   01010010100000010010
> +   result: 01010010100000000000.  */
> +
> +static tree
> +find_common_set_bits (tree min, tree max)
> 
> You should document these with an english sentence do clarify
> what this function (and find_common_unset_bits) does.  Certainly
> the function name doesn't match the workings if I apply my own
> common sense.

I am not native English speaker. I think my example in that
comment is perfectly adequate.

However, it is not relevant anymore. I am not going
to submit patches to the project with such sadistic
patch submission requirements:

On Saturday 12 August 2006 16:44, Richard Guenther wrote:
> > Like this?
> >
> > 2006-08-12 Denis Vlasenko <vda.linux@googlemail.com>
> >
> >         * tree-vrp.c (vrp_int_const_binop): deinline it
> 
> A tab before the '*', a captial letter after the ':' and a fullstop
> at the end.
> 
>       * tree-vrp.c (vrp_int_const_binop): Do not inline.
> 
> > Should I put a tab there or spaces? Exactly 8 spaces?
> > Do I need to have empty lines, where, and how many?

You probably didn't understand that I was sarcastic.

> if in doubt, follow existing changelogs in the gcc/ChangeLog
> file.  Two spaces after the date, for example.

Capital letter is mandatory! Two spaces are mandatory! Wow.
Gimme a break.
--
vda



More information about the Gcc-patches mailing list