Re: [PATCH net-next v4 2/8] 6lowpan: add uncompress header size function

[Alexander Aring <alex.aring-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On Wed, Feb 26, 2014 at 10:24:45AM -0800, Joe Perches wrote:
> On Wed, 2014-02-26 at 16:10 +0000, David Laight wrote:
> > From: Alexander Aring
> []
> > > diff --git a/net/ieee802154/6lowpan.h b/net/ieee802154/6lowpan.h
> []
> > > +static inline u8 lowpan_addr_mode_size(const u8 addr_mode)
> > > +{
> > > +	switch (addr_mode) {
> > > +	case LOWPAN_IPHC_ADDR_00:
> > > +		return 16;
> > > +	case LOWPAN_IPHC_ADDR_01:
> > > +		return 8;
> > > +	case LOWPAN_IPHC_ADDR_02:
> > > +		return 2;
> > > +	default:
> > > +		return 0;
> > > +	}
> > > +}
> > 
> > The compiler will generate much better code if you index an array instead
> > of using a switch statement.
> 
> Are you sure of that?
> 
> Perhaps the compiler would inline the assignment
> anyway if addr_mode is __builtin_constant_p
> 
> gcc 4.8 here does the same thing with:
> 
> static inline unsigned char f1(unsigned char a)
> {
> 	switch (a) {
> 	case 0: return 16;
> 	case 1: return 8;
> 	case 2: return 2;
> 	default: return 0;
> 	}
> }
> 
> static inline unsigned char f2(unsigned char a)
> {
> 	static const unsigned char rtns[] = { 16, 8, 2, 0 };
> 
> 	return rtns[a & 3];
> }

to make a mask with & 3 is already done, so we don't need that.
I am not sure what I should do now, change or not change? :-)

- Alex

------------------------------------------------------------------------------
Flow-based real-time traffic analytics software. Cisco certified tool.
Monitor traffic, SLAs, QoS, Medianet, WAAS etc. with NetFlow Analyzer
Customize your own dashboards, set traffic alerts and generate reports.
Network behavioral analysis & security monitoring. All-in-one tool.
http://pubads.g.doubleclick.net/gampad/clk?id=126839071&iu=/4140/ostg.clktrk