| Message ID | 5e3fade0f17f0357684536d77bc75e0028f2b62e.1692119257.git.asml.silence@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | caching and SQ/CQ optimisations | expand |
On 8/15/23 11:31 AM, Pavel Begunkov wrote: > Queues heads and tails cache line aligned. That makes sq, cq taking 4 > lines or 5 lines if we include the rest of struct io_rings (e.g. > sq_flags is frequently accessed). > > Since modern io_uring is mostly single threaded, it doesn't make much > send to sread them as such, it wastes space and puts additional pressure "sense to spread". Can fix up while applying. Change itself looks good to me.
On 8/19/23 16:05, Jens Axboe wrote: > On 8/15/23 11:31 AM, Pavel Begunkov wrote: >> Queues heads and tails cache line aligned. That makes sq, cq taking 4 >> lines or 5 lines if we include the rest of struct io_rings (e.g. >> sq_flags is frequently accessed). >> >> Since modern io_uring is mostly single threaded, it doesn't make much >> send to sread them as such, it wastes space and puts additional pressure > > "sense to spread". Can fix up while applying. Change itself looks good > to me. I'll be resending as we agreed, will fix it up, thanks
diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index c0c03d8059df..608a8e80e881 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -69,8 +69,8 @@ struct io_uring_task { }; struct io_uring { - u32 head ____cacheline_aligned_in_smp; - u32 tail ____cacheline_aligned_in_smp; + u32 head; + u32 tail; }; /*
Queues heads and tails cache line aligned. That makes sq, cq taking 4 lines or 5 lines if we include the rest of struct io_rings (e.g. sq_flags is frequently accessed). Since modern io_uring is mostly single threaded, it doesn't make much send to sread them as such, it wastes space and puts additional pressure on caches. Put them all into a single line. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> --- include/linux/io_uring_types.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)