Date: Tue, 12 Dec 2017 17:35:28 -0800
From: Jakub Kicinski <kubakici@wp.pl>
To: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [RFC][PATCH] new byteorder primitives -
 ..._{replace,get}_bits()
Message-ID: <20171212173528.340cd002@cakuba.netronome.com>
In-Reply-To: <20171213013056.GB21978@ZenIV.linux.org.uk>
References: <20171211053803.GW21978@ZenIV.linux.org.uk>
        <20171211155422.GA12326@ZenIV.linux.org.uk>
        <20171211200224.23bc5df4@cakuba.netronome.com>
        <20171212062002.GY21978@ZenIV.linux.org.uk>
        <20171212194532.GA7062@ZenIV.linux.org.uk>
        <20171212120409.64b6362e@cakuba.netronome.com>
        <20171212234856.GZ21978@ZenIV.linux.org.uk>
        <20171212155933.03c88eab@cakuba.netronome.com>
        <20171213003659.GA21978@ZenIV.linux.org.uk>
        <20171212170437.4b129e50@cakuba.netronome.com>
        <20171213013056.GB21978@ZenIV.linux.org.uk>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1979
Lines: 45

On Wed, 13 Dec 2017 01:30:56 +0000, Al Viro wrote:
> On Tue, Dec 12, 2017 at 05:04:37PM -0800, Jakub Kicinski wrote:
> > On Wed, 13 Dec 2017 00:36:59 +0000, Al Viro wrote:  
> > > On Tue, Dec 12, 2017 at 03:59:33PM -0800, Jakub Kicinski wrote:  
> > > > > +static __always_inline __##type type##_replace_bits(__##type old,	\
> > > > > +					base val, base mask)		\
> > > > > +{									\
> > > > > +	__##type m = to(mask);						\
> > > > > +        if (__builtin_constant_p(val) &&				\    
> > > > 
> > > > Is the lack of a __builtin_constant_p(mask) test intentional?  Sometimes
> > > > the bitfield is a packed array and people may have a helper to which
> > > > only the mask is passed as non-constant and the value is implied by the
> > > > helper, thus constant.    
> > > 
> > > If the mask in non-constant, we probably shouldn't be using that at all;
> > > could you show a real-world example where that would be the case?  
> > 
> > FIELD_* macros explicitly forbid this, since the code would be...
> > suboptimal with the runtime ffsl.  Real life examples are the hackish
> > macro NFP_ETH_SET_BIT_CONFIG() in
> > 
> > drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp_eth.c  
> 
> Why not simply make nfp_eth_set_bit_config() static inline and be
> done with that?  It's not that large and there are only few callers,
> so...

It used to be __always_inline, but apparently LLVM/clang doesn't
propagate constants :(  

4e59532541c8 ("nfp: don't depend on compiler constant propagation")

> > I remember there was also some Renesas code..  maybe this:
> > 
> > https://patchwork.kernel.org/patch/9881279/
> > 
> > it looks like cpg_z_clk_recalc_rate() and cpg_z2_clk_recalc_rate() only
> > differ in mask.  
> 
> *shrug*
> 
> That thing would expand into "reg &= 15" in one case and "reg >>= 8; reg &= 15"
> in another.  Either of which is cheaper than a function call, and definitely
> cheaper than any kind of dynamic calculation of shift, no matter how implemented.