DPDK patches and discussions
 help / color / mirror / Atom feed
From: Artur Trybula <arturx.trybula@intel.com>
To: dev@dpdk.org, fiona.trahe@intel.com, shallyv@marvell.com,
	adamx.dybkowski@intel.com, arturx.trybula@intel.com,
	akhil.goyal@nxp.com
Subject: [dpdk-dev] [PATCH] doc: updates for cycle-count mode in compression perf tool
Date: Wed, 12 Feb 2020 13:06:49 +0100	[thread overview]
Message-ID: <20200212120649.5080-1-arturx.trybula@intel.com> (raw)

This commit adds release notes and updates documentation for
the cycle-count mode added to the compression performance tool.

Fixes: 2695db95a147 ("test/compress: add cycle-count mode to perf tool")

Signed-off-by: Artur Trybula <arturx.trybula@intel.com>
---
 doc/guides/rel_notes/release_20_02.rst |  5 +++++
 doc/guides/tools/comp_perf.rst         | 16 +++++++++++++---
 2 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/doc/guides/rel_notes/release_20_02.rst b/doc/guides/rel_notes/release_20_02.rst
index 786f9b119..2183a8aa8 100644
--- a/doc/guides/rel_notes/release_20_02.rst
+++ b/doc/guides/rel_notes/release_20_02.rst
@@ -188,6 +188,11 @@ New Features
   Add event device support for ``l3fwd`` sample application. It demonstrates
   usage of poll and event mode IO mechanism under a single application.
 
+* **Added cycle-count mode to the compression performance tool.**
+
+  Enhanced the compression performance tool by adding a cycle-count mode
+  which can be used to help measure and tune hardware and software PMDs.
+
 
 Removed Items
 -------------
diff --git a/doc/guides/tools/comp_perf.rst b/doc/guides/tools/comp_perf.rst
index 6c59c8403..ee585d102 100644
--- a/doc/guides/tools/comp_perf.rst
+++ b/doc/guides/tools/comp_perf.rst
@@ -14,8 +14,16 @@ which are passed to compress device with compression operations.
 Then, the output buffers are fed into the decompression stage, and the resulting
 data is compared against the original data (verification phase). After that,
 a number of iterations are performed, compressing first and decompressing later,
-to check the throughput rate
-(showing cycles/iteration, cycles/Byte and Gbps, for compression and decompression).
+to check the throughput rate (showing cycles/iteration, cycles/Byte and Gbps,
+for compression and decompression).
+Another option: ``pmd-cyclecount``, gives the user the opportunity to measure
+the number of cycles per operation for the 3 phases: setup, enqueue_burst and
+dequeue_burst, for both compression and decompression. An optional delay can be
+inserted between enqueue and dequeue so no cycles are wasted in retries while
+waiting for a hardware device to finish. Although artificial, this allows
+to measure the minimum offload cost which could be achieved in a perfectly
+tuned system. Comparing the results of the two tests gives information about
+the trade-off between throughput and cycle-count.
 
 .. Note::
 
@@ -56,7 +64,7 @@ See the DPDK Getting Started Guides for more information on these options.
 Application Options
 ~~~~~~~~~~~~~~~~~~~
 
- ``--ptest [benchmark/verify]``: set test type (default: benchmark)
+ ``--ptest [throughput/verify/pmd-cyclecount]``: set test type (default: throughput)
 
  ``--driver-name NAME``: compress driver to use
 
@@ -84,6 +92,8 @@ Application Options
 
  ``--external-mbufs``: allocate and use memzones as external buffers instead of keeping the data directly in mbuf areas
 
+ ``--cc-delay-us N``: delay between enqueue and dequeue operations in microseconds, valid only for the cyclecount test (default: 500 us)
+
  ``-h``: prints this help
 
 
-- 
2.17.1


             reply	other threads:[~2020-02-12 12:06 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-12 12:06 Artur Trybula [this message]
2020-02-12 12:12 ` Dybkowski, AdamX
2020-02-13  9:28   ` Akhil Goyal

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200212120649.5080-1-arturx.trybula@intel.com \
    --to=arturx.trybula@intel.com \
    --cc=adamx.dybkowski@intel.com \
    --cc=akhil.goyal@nxp.com \
    --cc=dev@dpdk.org \
    --cc=fiona.trahe@intel.com \
    --cc=shallyv@marvell.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).