* [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path
@ 2017-08-14 12:29 Max Gurtovoy
2017-08-15 9:08 ` Sagi Grimberg
2017-08-16 8:10 ` Christoph Hellwig
0 siblings, 2 replies; 5+ messages in thread
From: Max Gurtovoy @ 2017-08-14 12:29 UTC (permalink / raw)
This patch slightly improves performance (mainly for small block sizes).
Signed-off-by: Max Gurtovoy <maxg at mellanox.com>
---
drivers/nvme/host/rdma.c | 12 ++++++------
1 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index da04df1..8540dbb 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -856,7 +856,7 @@ static void nvme_rdma_unmap_data(struct nvme_rdma_queue *queue,
if (req->mr->need_inval) {
res = nvme_rdma_inv_rkey(queue, req);
- if (res < 0) {
+ if (unlikely(res < 0)) {
dev_err(ctrl->ctrl.device,
"Queueing INV WR for rkey %#x failed (%d)\n",
req->mr->rkey, res);
@@ -921,7 +921,7 @@ static int nvme_rdma_map_sg_fr(struct nvme_rdma_queue *queue,
int nr;
nr = ib_map_mr_sg(req->mr, req->sg_table.sgl, count, NULL, PAGE_SIZE);
- if (nr < count) {
+ if (unlikely(nr < count)) {
if (nr < 0)
return nr;
return -EINVAL;
@@ -1057,7 +1057,7 @@ static int nvme_rdma_post_send(struct nvme_rdma_queue *queue,
first = ≀
ret = ib_post_send(queue->qp, first, &bad_wr);
- if (ret) {
+ if (unlikely(ret)) {
dev_err(queue->ctrl->ctrl.device,
"%s failed with error code %d\n", __func__, ret);
}
@@ -1083,7 +1083,7 @@ static int nvme_rdma_post_recv(struct nvme_rdma_queue *queue,
wr.num_sge = 1;
ret = ib_post_recv(queue->qp, &wr, &bad_wr);
- if (ret) {
+ if (unlikely(ret)) {
dev_err(queue->ctrl->ctrl.device,
"%s failed with error code %d\n", __func__, ret);
}
@@ -1443,7 +1443,7 @@ static blk_status_t nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
blk_mq_start_request(rq);
err = nvme_rdma_map_data(queue, rq, c);
- if (err < 0) {
+ if (unlikely(err < 0)) {
dev_err(queue->ctrl->ctrl.device,
"Failed to map data (%d)\n", err);
nvme_cleanup_cmd(rq);
@@ -1457,7 +1457,7 @@ static blk_status_t nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
flush = true;
err = nvme_rdma_post_send(queue, sqe, req->sge, req->num_sge,
req->mr->need_inval ? &req->reg_wr.wr : NULL, flush);
- if (err) {
+ if (unlikely(err)) {
nvme_rdma_unmap_data(queue, rq);
goto err;
}
--
1.7.1
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path
2017-08-14 12:29 [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path Max Gurtovoy
@ 2017-08-15 9:08 ` Sagi Grimberg
2017-08-15 9:15 ` Max Gurtovoy
2017-08-16 8:10 ` Christoph Hellwig
1 sibling, 1 reply; 5+ messages in thread
From: Sagi Grimberg @ 2017-08-15 9:08 UTC (permalink / raw)
> This patch slightly improves performance (mainly for small block sizes).
This looks fine,
Looks fine,
Reviewed-by: Sagi Grimberg <sagi at grimberg.me>
Since you mentioned it, can you quantify what that bought us?
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path
2017-08-15 9:08 ` Sagi Grimberg
@ 2017-08-15 9:15 ` Max Gurtovoy
2017-08-15 9:58 ` Sagi Grimberg
0 siblings, 1 reply; 5+ messages in thread
From: Max Gurtovoy @ 2017-08-15 9:15 UTC (permalink / raw)
On 8/15/2017 12:08 PM, Sagi Grimberg wrote:
>> This patch slightly improves performance (mainly for small block sizes).
>
> This looks fine,
>
> Looks fine,
>
> Reviewed-by: Sagi Grimberg <sagi at grimberg.me>
>
> Since you mentioned it, can you quantify what that bought us?
1 subsystem, 1 ns (null blk backing device)
bs (w/w.o)
--- ---------
512: 912k/900k
1k: 912k/900k
2k : 911k/901k
4k: 922k/906k
8k: 568k/569k
16k: 288k/288k
-Max
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path
2017-08-15 9:15 ` Max Gurtovoy
@ 2017-08-15 9:58 ` Sagi Grimberg
0 siblings, 0 replies; 5+ messages in thread
From: Sagi Grimberg @ 2017-08-15 9:58 UTC (permalink / raw)
> 1 subsystem, 1 ns (null blk backing device)
>
>
> bs (w/w.o)
> --- ---------
> 512: 912k/900k
> 1k: 912k/900k
> 2k : 911k/901k
> 4k: 922k/906k
That's somewhat surprising...
Thanks
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path
2017-08-14 12:29 [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path Max Gurtovoy
2017-08-15 9:08 ` Sagi Grimberg
@ 2017-08-16 8:10 ` Christoph Hellwig
1 sibling, 0 replies; 5+ messages in thread
From: Christoph Hellwig @ 2017-08-16 8:10 UTC (permalink / raw)
Thanks,
applied to nvme-4.14.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-08-16 8:10 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-08-14 12:29 [PATCH 1/1] nvme-rdma: Use unlikely macro in the fast path Max Gurtovoy
2017-08-15 9:08 ` Sagi Grimberg
2017-08-15 9:15 ` Max Gurtovoy
2017-08-15 9:58 ` Sagi Grimberg
2017-08-16 8:10 ` Christoph Hellwig
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.