Re: x86/csum: Remove unnecessary odd handling

From: Noah Goldstein <goldstein.w.n@gmail.com>
To: David Laight <David.Laight@aculab.com>
Cc: "x86@kernel.org" <x86@kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"edumazet@google.com" <edumazet@google.com>,
	"tglx@linutronix.de" <tglx@linutronix.de>,
	"mingo@redhat.com" <mingo@redhat.com>,
	"torvalds@linux-foundation.org" <torvalds@linux-foundation.org>,
	"bp@alien8.de" <bp@alien8.de>,
	"dave.hansen@linux.intel.com" <dave.hansen@linux.intel.com>,
	"hpa@zytor.com" <hpa@zytor.com>
Subject: Re: x86/csum: Remove unnecessary odd handling
Date: Wed, 20 Sep 2023 14:20:24 -0500	[thread overview]
Message-ID: <CAFUsyfLPLQBs2gQwLDUZHudQrkHSXCcV=Wi0=Znb3gChDZvX2Q@mail.gmail.com> (raw)
In-Reply-To: <d4c6e6274bd14391a1e59f84f9ebce7a@AcuMS.aculab.com>

On Wed, Sep 6, 2023 at 9:38 AM David Laight <David.Laight@aculab.com> wrote:
>
> From: Noah Goldstein
> > Sent: 01 September 2023 23:21
> ...
> > +     return add32_with_carry(temp64 >> 32, temp64 & 0xffffffff);
>
> The generic C alternative:
>         return (temp64 + ror64(temp64, 32)) >> 32;
> is the same number of instructions but might get
> better scheduling.
>
Sorry, I missed this.
Bright idea :)
Adding in new version + you reviewed by tag. Then hopefully this can
get in...

> The C version of csum_fold() from arc/include/asm/checksum.h
> is also better than the x86 asm version.
> (And also pretty much all the other architecture dependant
> copies.)
>
>         David
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
>