Re: [PATCHv2 1/2] add basic register-field manipulation macros

[Jakub Kicinski]

On Tue, 14 Jun 2016 20:53:28 +0200, Arend van Spriel wrote:
> On 14-06-16 13:44, Jakub Kicinski wrote:
> > +#ifndef _LINUX_BITFIELD_H
> > +#define _LINUX_BITFIELD_H
> > +
> > +#include <asm/types.h>
> > +#include <linux/bug.h>
> > +#include <linux/log2.h>
> > +
> > +#define _bf_shf(x) (__builtin_ffsll(x) - 1)
> > +
> > +#define _BF_FIELD_CHECK(_mask, _val)					\
> > +	({								\
> > +		const u64 hi = (_mask) + (1ULL << _bf_shf(_mask));	\
> > +									\
> > +		BUILD_BUG_ON(!(_mask) || (hi && !is_power_of_2_u64(hi))); \
> > +		BUILD_BUG_ON(__builtin_constant_p(_val) ?		\
> > +			     ~((_mask) >> _bf_shf(_mask)) & (_val) :	\
> > +			     0);					\
> > +	})  
> 
> I am sceptic whether it is useful to have 64-bit used here and there is
> a price to pay on (many) 32-bit architectures for using 64-bit
> operations. Maybe it is not an issue because it is inside BUILD_BUG_ON()
> macro.

It's a trade-off between having a separate set of macros for 32-bit
machines and risking someone potentially loosing cycles when using the
macros in a non-standard way.  Note that all 64 bit operations are
performed on the mask which I expect to be compile time constant.

> > +#define FIELD_PUT(_mask, _val)					\
> > +	({							\
> > +		_BF_FIELD_CHECK(_mask, _val);			\
> > +		((u32)(_val) << _bf_shf(_mask)) & (_mask);	\
> > +	})
> > +
> > +#define FIELD_GET(_mask, _val)					\
> > +	({							\
> > +		_BF_FIELD_CHECK(_mask, 0);			\
> > +		(u32)(((_val) & (_mask)) >> _bf_shf(_mask));	\
> > +	})
> > +
> > +#define FIELD_PUT64(_mask, _val)				\
> > +	({							\
> > +		_BF_FIELD_CHECK(_mask, _val);			\
> > +		((u64)(_val) << _bf_shf(_mask)) & (_mask);	\
> > +	})
> > +
> > +#define FIELD_GET64(_mask, _val)				\
> > +	({							\
> > +		_BF_FIELD_CHECK(_mask, 0);			\
> > +		(u64)(((_val) & (_mask)) >> _bf_shf(_mask));	\
> > +	})  
> 
> Is there really hardware out there that exposes 64-bit wide hardware
> registers?

I need this for encoding 64-bit wide RISC instructions [1] to be honest.

[1] http://thread.gmane.org/gmane.linux.network/414538