[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH 01/17] block/nvme: Avoid further processing if trace event no
From: |
Stefan Hajnoczi |
Subject: |
Re: [PATCH 01/17] block/nvme: Avoid further processing if trace event not enabled |
Date: |
Mon, 29 Jun 2020 14:02:22 +0100 |
On Fri, Jun 26, 2020 at 04:02:43PM +0200, Philippe Mathieu-Daudé wrote:
> On 6/26/20 12:36 PM, Stefan Hajnoczi wrote:
> > On Thu, Jun 25, 2020 at 08:48:22PM +0200, Philippe Mathieu-Daudé wrote:
> >> Avoid further processing if TRACE_NVME_SUBMIT_COMMAND_RAW is
> >> not enabled.
> >
> > Why?
> >
> > This saves 8 trace events, each with 8 arguments. I guess the intent is
> > to improve performance. Did you measure an improvement?
>
> No testing, I just tried to outsmart the compiler :/
Usually performance patches are accompanied with benchmark results.
Otherwise it's easy to modify code without making a difference or
accidentally introducing performance regressions. But this is a small
change and I'm not worried.
Please do explicitly state in the commit description that this is
intended as a performance optimization. I wasn't exactly sure about the
reason for this change.
Stefan
signature.asc
Description: PGP signature
- [PATCH 00/17] block/nvme: Various cleanups required to use multiple queues, Philippe Mathieu-Daudé, 2020/06/25
- [PATCH 02/17] block/nvme: Let nvme_create_queue_pair() fail gracefully, Philippe Mathieu-Daudé, 2020/06/25
- [PATCH 03/17] block/nvme: Define QUEUE_INDEX macros to ease code review, Philippe Mathieu-Daudé, 2020/06/25
- [PATCH 05/17] block/nvme: Replace qemu_try_blockalign0 by qemu_try_blockalign/memset, Philippe Mathieu-Daudé, 2020/06/25
- [PATCH 04/17] block/nvme: Be explicit we share NvmeIdCtrl / NvmeIdNs structures, Philippe Mathieu-Daudé, 2020/06/25
- [PATCH 06/17] block/nvme: Replace qemu_try_blockalign(bs) by qemu_try_memalign(pg_sz), Philippe Mathieu-Daudé, 2020/06/25