Re: [PATCH] [4/6] kfifo: add kfifo_out_peek

[Stefani Seibold]

Am Montag, den 04.01.2010, 22:24 +0000 schrieb Alan Cox:
> > Macros are unpopular, for good reasons.  But the case for a
> > template-based container such as this is a good one.  However I worry
> > about the code bloat whcih the macro version might add.  We worry about
> > all this later on.
> 
> Its really a special case for structs and fixed objects.
> 
> Bloat is going to be a big issue if its macro and all the serial/tty
> stuff switches to it. Please keep the bytewise one none macro - even if
> its a lib/foo.c file that simply uses the macros to produce the existing
> new API.

Nope, currently the bytewise is the special case. Most of the user of
kfifo try to store other type then bytes in a fifo. 

The new macro based kfifo does not bloat the code, the opposite is the
truth. I checked the assembler output on intel and ppc and the generated
code is smaller and better.

The macro are written in a way that you get a useful single line compile
time error message.

Also "the power of two" thing will make the kfifo not very useable to
handle not multiple of 2 datas, in a fifo.

At last a type safe kfifo is 100% compatible to the current kfifo if the
the is a "unsigned char".

It would be better to discuss this in the "[PATCH] new kfifo API v.08"
threat from 28.12.2009.

Regards,
Stefani


--
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/