[Qemu-devel] [PATCH 0/2] improve aio_poll performance

[Qemu-devel] [PATCH 0/2] improve aio_poll performance

[Yaowei Bai]

This patchset change to check epoll's enablement first before nodes
iteration to improve aio_poll()'s performance.

Also fix a wrong comment of mirror_start().

Yaowei Bai (2):
  block: mirror: fix wrong comment of mirror_start
  aio: improve aio_poll performance by checking epoll only once

 aio-posix.c               | 12 +++++++-----
 include/block/block_int.h |  2 +-
 2 files changed, 8 insertions(+), 6 deletions(-)

-- 
1.8.3.1

[Qemu-devel] [PATCH 1/2] block: mirror: fix wrong comment of mirror_start

[Yaowei Bai]

Obviously, we should write to '@target'.

Signed-off-by: Yaowei Bai <baiyaowei@cmss.chinamobile.com>
Reviewed-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
---
 include/block/block_int.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/block/block_int.h b/include/block/block_int.h
index 1e939de..77d4e65 100644
--- a/include/block/block_int.h
+++ b/include/block/block_int.h
@@ -730,7 +730,7 @@ void commit_active_start(const char *job_id, BlockDriverState *bs,
  * @errp: Error object.
  *
  * Start a mirroring operation on @bs.  Clusters that are allocated
- * in @bs will be written to @bs until the job is cancelled or
+ * in @bs will be written to @target until the job is cancelled or
  * manually completed.  At the end of a successful mirroring job,
  * @bs will be switched to read from @target.
  */
-- 
1.8.3.1

[Qemu-devel] [PATCH 2/2] aio: improve aio_poll performance by checking epoll only once

[Yaowei Bai]

As epoll whether enabled or not is a global setting, we can just
check it only once rather than checking it with every node iteration.
Through this we can avoid a lot of checks when epoll is not enabled.

Signed-off-by: Yaowei Bai <baiyaowei@cmss.chinamobile.com>
Reviewed-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
---
 aio-posix.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/aio-posix.c b/aio-posix.c
index 43162a9..4ef34dd 100644
--- a/aio-posix.c
+++ b/aio-posix.c
@@ -431,11 +431,13 @@ bool aio_poll(AioContext *ctx, bool blocking)
     assert(npfd == 0);
 
     /* fill pollfds */
-    QLIST_FOREACH(node, &ctx->aio_handlers, node) {
-        if (!node->deleted && node->pfd.events
-            && !aio_epoll_enabled(ctx)
-            && aio_node_check(ctx, node->is_external)) {
-            add_pollfd(node);
+
+    if (!aio_epoll_enabled(ctx)) {
+        QLIST_FOREACH(node, &ctx->aio_handlers, node) {
+            if (!node->deleted && node->pfd.events
+                && aio_node_check(ctx, node->is_external)) {
+                add_pollfd(node);
+            }
         }
     }
 
-- 
1.8.3.1

Re: [Qemu-devel] [PATCH 2/2] aio: improve aio_poll performance by checking epoll only once

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 991 bytes --]

On Wed, Sep 14, 2016 at 07:03:39AM -0400, Yaowei Bai wrote:
> As epoll whether enabled or not is a global setting, we can just
> check it only once rather than checking it with every node iteration.
> Through this we can avoid a lot of checks when epoll is not enabled.
> 
> Signed-off-by: Yaowei Bai <baiyaowei@cmss.chinamobile.com>
> Reviewed-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
> ---
>  aio-posix.c | 12 +++++++-----
>  1 file changed, 7 insertions(+), 5 deletions(-)

The commit message says "improve aio_poll performance" but no benchmark
results were provided.  Therefore I can't take this patch as a
performance optimization.  I'm still happy to merge the patch since it
makes the if statement simpler but I'll rename it to "aio-posix: avoid
unnecessary aio_epoll_enabled() calls".

I don't think this patch gives a measurable performance improvement.  If
you believe otherwise, please post benchmark results.  Please let me
know what you think.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]

Re: [Qemu-devel] [PATCH 1/2] block: mirror: fix wrong comment of mirror_start

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 1107 bytes --]

On 09/14/2016 06:03 AM, Yaowei Bai wrote:
> Obviously, we should write to '@target'.
> 
> Signed-off-by: Yaowei Bai <baiyaowei@cmss.chinamobile.com>
> Reviewed-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
> ---
>  include/block/block_int.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 

Reviewed-by: Eric Blake <eblake@redhat.com>

> diff --git a/include/block/block_int.h b/include/block/block_int.h
> index 1e939de..77d4e65 100644
> --- a/include/block/block_int.h
> +++ b/include/block/block_int.h
> @@ -730,7 +730,7 @@ void commit_active_start(const char *job_id, BlockDriverState *bs,
>   * @errp: Error object.
>   *
>   * Start a mirroring operation on @bs.  Clusters that are allocated
> - * in @bs will be written to @bs until the job is cancelled or
> + * in @bs will be written to @target until the job is cancelled or
>   * manually completed.  At the end of a successful mirroring job,
>   * @bs will be switched to read from @target.
>   */
> 

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]

Re: [Qemu-devel] [PATCH 2/2] aio: improve aio_poll performance by checking epoll only once

[Yaowei Bai]

On Wed, Sep 14, 2016 at 05:40:17PM +0100, Stefan Hajnoczi wrote:
> On Wed, Sep 14, 2016 at 07:03:39AM -0400, Yaowei Bai wrote:
> > As epoll whether enabled or not is a global setting, we can just
> > check it only once rather than checking it with every node iteration.
> > Through this we can avoid a lot of checks when epoll is not enabled.
> > 
> > Signed-off-by: Yaowei Bai <baiyaowei@cmss.chinamobile.com>
> > Reviewed-by: Xiubo Li <lixiubo@cmss.chinamobile.com>
> > ---
> >  aio-posix.c | 12 +++++++-----
> >  1 file changed, 7 insertions(+), 5 deletions(-)
> 
> The commit message says "improve aio_poll performance" but no benchmark
> results were provided.  Therefore I can't take this patch as a
> performance optimization.  I'm still happy to merge the patch since it
> makes the if statement simpler but I'll rename it to "aio-posix: avoid
> unnecessary aio_epoll_enabled() calls".

Sorry for replying late, i just came back from my vacation. And i'd like
to say it's ok for me to rename the commit message. Thank you for doing
it.

> 
> I don't think this patch gives a measurable performance improvement.  If

Yes, i agree with you at this point in consideration of there's no too many
fds to poll at most time.

> you believe otherwise, please post benchmark results.  Please let me
> know what you think.
> 
> Stefan

Re: [Qemu-devel] [PATCH 0/2] improve aio_poll performance

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 655 bytes --]

On Wed, Sep 14, 2016 at 07:03:37AM -0400, Yaowei Bai wrote:
> This patchset change to check epoll's enablement first before nodes
> iteration to improve aio_poll()'s performance.
> 
> Also fix a wrong comment of mirror_start().
> 
> Yaowei Bai (2):
>   block: mirror: fix wrong comment of mirror_start
>   aio: improve aio_poll performance by checking epoll only once
> 
>  aio-posix.c               | 12 +++++++-----
>  include/block/block_int.h |  2 +-
>  2 files changed, 8 insertions(+), 6 deletions(-)

Changed commit message as discussed.

Thanks, applied to my block tree:
https://github.com/stefanha/qemu/commits/block

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]