Re: [PATCH v3] iomap: add allocation cache for iomap_dio

[Vlastimil Babka (SUSE)]

On 3/17/26 10:12, Christoph Hellwig wrote:
> On Tue, Mar 17, 2026 at 04:33:24PM +0800, changfengnan wrote:
>> > IO is not my area but getting from 1.19M to 1.20M doesn't look like it's
>> > worth the custom code? (possibly from 1.17M to 1.20M it also wasn't).
>> Yes, at least for now, there’s no need for a per-CPU.
>> It might be better to replace kmalloc with a new cache, but my tests so far
>> haven’t shown any performance improvements.  I’ll look into it further.
> 
> Does using a kmem_cache help?  That should generally be a nice win
> anyway due to keeping the objects together.

I think that's exactly what "It might be better to replace kmalloc with a
new cache" meant, and apparently with no improvements.
You might want to try create it with SLAB_NO_MERGE flag so it's really a
separate cache. Custom sheaf_capacity might also achieve that effect, but in
order to have deterministic results, the flag is a sure way.