Re: [PATCH v7] kunit: memcpy: Add benchmark

[David Laight]

On Mon, 9 Nov 2026 19:39:25 -0800
Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:

> On Tue, 10 Nov 2026 03:16:22 +0100 Matteo Croce <technoboy85@xxxxxxxxx> wrote:
> 
> > Add optional benchmarks for memcpy() and memmove(), enabled with
> > CONFIG_MEMCPY_KUNIT_BENCHMARK.  
> 
> hm, sorry, I must ask "why".  Nowadays optimizing memcpy() is a
> super-specialized thing undertaken by the bravest architecture
> specialists who are unlikely to rely upon a generic kunit test for the
> purpose!

You also want a comparison against the 'obvious loops' (which it can
take some effort to generate because the compiler tends to convert them
to memcpy()). Otherwise the numbers are pretty meaningless.

But kernel benchmarks are really the wrong place.
The 'bravest architecture specialists' are going to test the code
in userspace where multiple algorithms can be tested and quickly
tweaked/dismissed.
You also need a bank of systems with different cpu variants and run the
code on all on them - much easier if you only need to copy a program across.

I also (sometimes being one of the 'bravest architecture specialists')
try to work out how many bytes/clock the a copy (or ipcsum) is doing
and what the fixed costs are.
That actually tells you something useful.
(But you always need to check that the compiler generated something like
the expected copy - otherwise it is pretty pointless.)

David