RE: [PATCH] ext4: simplify 32bit calculation of lblk

[David Laight]

From: chi wu
> Sent: 05 April 2023 13:48
> 
> David Laight <David.Laight@xxxxxxxxxx> 于2023年4月5日周三 19:43写道：
> >
> > From: chi wu
> > > Sent: 05 April 2023 09:48
> > >
> > > Christoph Hellwig <hch@xxxxxxxxxxxxx> 于2023年4月5日周三 13:40写道：
> > > >
> > > > On Mon, Apr 03, 2023 at 09:53:04PM +0800, wuchi wrote:
> > > > > -                     if (block > ext_block)
> > > > > -                             return ext_pblk + (block - ext_block);
> > > > > -                     else
> > > > > -                             return ext_pblk - (ext_block - block);
> > > > > +                     return ext_pblk + ((signed long long)block - (signed long
> long)ext_block);
> > > >
> > > > And what exactly is the value add here, except for turning an easy
> > > > to parse statement into a complex expression using casts?
> > > >
> > > Yes，it will be more complex. the original intention is to reduce the
> > > conditional branch.
> >
> > What is wrong with just:
> >         return ext_pblk + block - ext_block;
> > (64bit + 32bit - 32bit)
> >

> oh, It's my fault. I am trapped in that ext_pblk + block may overflow,
> but it is ok here. thanks.

That doesn't matter, it will 'un-overflow' when ext_block is subtracted.
You do need the '+' to happen before the '-', ok because +/- group
left to right in C.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)