Re: [PATCH 1/6] gxfb: create DC/VP/FP-specific handlers rather thanusing readl/writel

[Andrew Morton]

On Mon, 10 Mar 2008 17:35:44 -0400
Andres Salomon <dilinger@xxxxxxxxxx> wrote:

> On Mon, 10 Mar 2008 14:24:05 -0700
> Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> 
> > On Sat, 8 Mar 2008 20:48:26 -0500
> > Andres Salomon <dilinger@xxxxxxxxxx> wrote:
> > 
> > > +#define read_dc(reg)		readl(par->dc_regs + (reg))
> > > +#define write_dc(reg, val)	writel((val), par->dc_regs + (reg))
> > > +
> > > +#define read_vp(reg)		readl(par->vid_regs + (reg))
> > > +#define write_vp(reg, val)	writel((uint32_t) (val), \
> > > +					par->vid_regs + (reg))
> > > +
> > > +#define read_fp(reg)		readl(par->vid_regs + (reg))
> > > +#define write_fp(reg, val)	writel((uint32_t) (val), \
> > > +					par->vid_regs + (reg))
> > > +
> > 
> > Not very nice, sorry.  They're macros, and macros rather suck.  And they
> > implicitly rely upon the caller having some variable called "par" in scope.
> > 
> > It would be much nicer to do
> > 
> > /*
> >  * documentation goes here
> >  */
> > static inline u32 read_dc(struct geodefb_par *par, int reg)
> > {
> > 	return readl(par->dc_regs, reg);
> > }
> > 
> > no?
> 
> I can change it if you'd like (although.. sigh.)
> 
> However, it's a lot of extra passing around of the 'par' without any
> good reason.  Normal I prefer inline functions to macros as well, but
> I don't see the point here.
>  

It'll generate the same (or similar) code in both versions.

Code is written once and is read thousands of times, so we should optimise
for readers, not writers.  And I do think that being conventional here
helps readability, even if it does add a bit more source code.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/