[PATCH 2/2] bitops: rotate: Add riscv implementation using Zbb extension

[David Laight]

On Fri, 20 Jun 2025 12:20:47 -0400
Yury Norov <yury.norov at gmail.com> wrote:

> On Fri, Jun 20, 2025 at 07:16:10PM +0800, cp0613 at linux.alibaba.com wrote:
> > From: Chen Pei <cp0613 at linux.alibaba.com>
> > 
> > The RISC-V Zbb extension[1] defines bitwise rotation instructions,
> > which can be used to implement rotate related functions.
> > 
> > [1] https://github.com/riscv/riscv-bitmanip/
> > 
> > Signed-off-by: Chen Pei <cp0613 at linux.alibaba.com>
> > ---
> >  arch/riscv/include/asm/bitops.h | 172 ++++++++++++++++++++++++++++++++
> >  1 file changed, 172 insertions(+)
> > 
> > diff --git a/arch/riscv/include/asm/bitops.h b/arch/riscv/include/asm/bitops.h
> > index d59310f74c2b..be247ef9e686 100644
> > --- a/arch/riscv/include/asm/bitops.h
> > +++ b/arch/riscv/include/asm/bitops.h
> > @@ -20,17 +20,20 @@
> >  #include <asm-generic/bitops/__fls.h>
> >  #include <asm-generic/bitops/ffs.h>
> >  #include <asm-generic/bitops/fls.h>
> > +#include <asm-generic/bitops/rotate.h>
> >  
> >  #else
> >  #define __HAVE_ARCH___FFS
> >  #define __HAVE_ARCH___FLS
> >  #define __HAVE_ARCH_FFS
> >  #define __HAVE_ARCH_FLS
> > +#define __HAVE_ARCH_ROTATE
> >  
> >  #include <asm-generic/bitops/__ffs.h>
> >  #include <asm-generic/bitops/__fls.h>
> >  #include <asm-generic/bitops/ffs.h>
> >  #include <asm-generic/bitops/fls.h>
> > +#include <asm-generic/bitops/rotate.h>
> >  
> >  #include <asm/alternative-macros.h>
> >  #include <asm/hwcap.h>
> > @@ -175,6 +178,175 @@ static __always_inline int variable_fls(unsigned int x)
> >  	 variable_fls(x_);					\
> >  })  
> 
> ...
> 
> > +static inline u8 variable_ror8(u8 word, unsigned int shift)
> > +{
> > +	u32 word32 = ((u32)word << 24) | ((u32)word << 16) | ((u32)word << 8) | word;  
> 
> Can you add a comment about what is happening here? Are you sure it's
> optimized out in case of the 'legacy' alternative?

Is it even a gain in the zbb case?
The "rorw" is only ever going to help full word rotates.
Here you might as well do ((word << 8 | word) >> shift).

For "rol8" you'd need ((word << 24 | word) 'rol' shift).
I still bet the generic code is faster (but see below).

Same for 16bit rotates.

Actually the generic version is (probably) horrid for everything except x86.
See https://www.godbolt.org/z/xTxYj57To

unsigned char rol(unsigned char v, unsigned int shift)
{
    return (v << 8 | v) << shift >> 8;
}

unsigned char ror(unsigned char v, unsigned int shift)
{
    return (v << 8 | v) >> shift;
}

	David