Re: [PATCH 1/3] riscv: optimized memcpy

[Matteo Croce <mcroce@linux.microsoft.com>]

On Wed, Jun 16, 2021 at 10:24 AM David Laight <David.Laight@aculab.com> wrote:
>
> From: Matteo Croce
> > Sent: 16 June 2021 03:02
> ...
> > > > That's a good idea, but if you read the replies to Gary's original
> > > > patch
> > > > https://lore.kernel.org/linux-riscv/20210216225555.4976-1-gary@garyguo.net/
> > > > .. both Gary, Palmer and David would rather like a C-based version.
> > > > This is one attempt at providing that.
> > >
> > > Yep, I prefer C as well :)
> > >
> > > But if you check commit 04091d6, the assembly version was introduced
> > > for KASAN. So if we are to change it back to C, please make sure KASAN
> > > is not broken.
> > >
> ...
> > Leaving out the first memcpy/set of every test which is always slower, (maybe
> > because of a cache miss?), the current implementation copies 260 Mb/s when
> > the low order bits match, and 114 otherwise.
> > Memset is stable at 278 Mb/s.
> >
> > Gary's implementation is much faster, copies still 260 Mb/s when euqlly placed,
> > and 230 Mb/s otherwise. Memset is the same as the current one.
>
> Any idea what the attainable performance is for the cpu you are using?
> Since both memset and memcpy are running at much the same speed
> I suspect it is all limited by the writes.
>
> 272MB/s is only 34M writes/sec.
> This seems horribly slow for a modern cpu.
> So is this actually really limited by the cache writes to physical memory?
>
> You might want to do some tests (userspace is fine) where you
> check much smaller lengths that definitely sit within the data cache.
>

I get similar results in userspace, this tool write to RAM with
variable data width:

root@beaglev:~/src# ./unalign_check 1 0 1
size:           1 Mb
write size:      8 bit
unalignment:    0 byte
elapsed time:   0.01 sec
throughput:     124.36 Mb/s

# ./unalign_check 1 0 8
size:           1 Mb
write size:      64 bit
unalignment:    0 byte
elapsed time:   0.00 sec
throughput:     252.12 Mb/s

> It is also worth checking how much overhead there is for
> short copies - they are almost certainly more common than
> you might expect.
> This is one problem with excessive loop unrolling - the 'special
> cases' for the ends of the buffer start having a big effect
> on small copies.
>

I too believe that they are much more common than long ones.
Indeed, I wish to reduce the MIN_THRESHOLD value from 64 to 32 or even 16.
Or having it dependend on the word size, e.g. sizeof(long) * 2.

Suggestions?

> For cpu that support misaligned memory accesses, one 'trick'
> for transfers longer than a 'word' is to do a (probably) misaligned
> transfer of the last word of the buffer first followed by the
> transfer of the rest of the buffer (overlapping a few bytes at the end).
> This saves on conditionals and temporary values.
>
>         David
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
>

Regards,
-- 
per aspera ad upstream