From: Robin Jarry <rjarry@redhat.com>
To: dev@dpdk.org
Cc: Christophe Fontaine <cfontain@redhat.com>,
David Marchand <david.marchand@redhat.com>,
Jerin Jacob <jerinj@marvell.com>,
Kiran Kumar Kokkilagadda <kirankumark@marvell.com>,
Konstantin Ananyev <konstantin.ananyev@huawei.com>,
Maxime Leroy <maxime@leroys.fr>,
Nithin Kumar Dabilpuram <ndabilpuram@marvell.com>,
Vladimir Medvedkin <vladimir.medvedkin@intel.com>,
Zhirun Yan <yanzhirun_163@163.com>
Subject: [PATCH dpdk v2 0/2] graph: priority-based bitmap scheduling
Date: Tue, 19 May 2026 23:38:20 +0200 [thread overview]
Message-ID: <20260519213822.735891-1-rjarry@redhat.com> (raw)
In-Reply-To: <20260519101232.541102-2-rjarry@redhat.com>
Replace the FIFO circular buffer in rte_graph with a bitmap and a
priority-sorted schedule table. Nodes with lower priority values are
visited first during the graph walk. Source nodes are forced to
INT16_MIN. Topological depth from BFS ensures upstream-before-downstream
ordering when all priorities are equal.
In fan-out-then-converge topologies, setting a higher priority (lower
value) on branch nodes avoids redundant visits to the converge node.
The diamond perf test shows a ~10% throughput improvement (converge
visited once at 256 objs/call instead of twice at 128).
Changes v1 -> v2:
- added diamond perf test into separate preparatory patch
- added topological depth (topo_order) as secondary sort key to preserve
FIFO-like upstream-before-downstream ordering for default priorities,
preventing a regression in the reverse tree test
- restored idx == 0 guard on bitmap set in the enqueue path to avoid a
~15% throughput regression caused by touching the pending bitmap
cache line on every enqueue call
- use hiprio worker nodes in diamond test to demonstrate the actual
priority-based scheduling benefit
- added performance numbers measured before and after
Cc: Christophe Fontaine <cfontain@redhat.com>
Cc: David Marchand <david.marchand@redhat.com>
Cc: Jerin Jacob <jerinj@marvell.com>
Cc: Kiran Kumar Kokkilagadda <kirankumark@marvell.com>
Cc: Konstantin Ananyev <konstantin.ananyev@huawei.com>
Cc: Maxime Leroy <maxime@leroys.fr>
Cc: Nithin Kumar Dabilpuram <ndabilpuram@marvell.com>
Cc: Vladimir Medvedkin <vladimir.medvedkin@intel.com>
Cc: Zhirun Yan <yanzhirun_163@163.com>
Robin Jarry (2):
graph: add diamond topology performance test
graph: replace circular buffer with priority-based bitmap
app/test/test_graph_perf.c | 140 +-
doc/guides/prog_guide/graph_lib.rst | 37 +-
.../prog_guide/img/graph_mem_layout.svg | 1823 +++++++----------
lib/graph/graph.c | 27 +-
lib/graph/graph_debug.c | 12 +-
lib/graph/graph_ops.c | 46 +
lib/graph/graph_populate.c | 119 +-
lib/graph/graph_private.h | 42 +-
lib/graph/node.c | 2 +
lib/graph/rte_graph.h | 1 +
lib/graph/rte_graph_model_mcore_dispatch.h | 34 +-
lib/graph/rte_graph_model_rtc.h | 65 +-
lib/graph/rte_graph_worker.h | 2 +-
lib/graph/rte_graph_worker_common.h | 79 +-
14 files changed, 1194 insertions(+), 1235 deletions(-)
--
2.54.0
next prev parent reply other threads:[~2026-05-19 21:38 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-05-19 10:12 [PATCH dpdk] graph: replace circular buffer with priority-based bitmap scheduling Robin Jarry
2026-05-19 10:25 ` [EXTERNAL] " Jerin Jacob
2026-05-19 21:32 ` Robin Jarry
2026-06-09 4:52 ` Jerin Jacob
2026-05-19 21:38 ` Robin Jarry [this message]
2026-05-19 21:38 ` [PATCH dpdk v2 1/2] graph: add diamond topology performance test Robin Jarry
2026-05-19 21:38 ` [PATCH dpdk v2 2/2] graph: replace circular buffer with priority-based bitmap Robin Jarry
2026-06-16 6:30 ` kirankumark
2026-06-16 7:16 ` Robin Jarry
2026-06-16 7:57 ` kirankumark
2026-06-16 8:03 ` kirankumark
2026-06-10 10:51 ` [EXTERNAL] [PATCH dpdk] graph: replace circular buffer with priority-based bitmap scheduling Kiran Kumar Kokkilagadda
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=20260519213822.735891-1-rjarry@redhat.com \
--to=rjarry@redhat.com \
--cc=cfontain@redhat.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=kirankumark@marvell.com \
--cc=konstantin.ananyev@huawei.com \
--cc=maxime@leroys.fr \
--cc=ndabilpuram@marvell.com \
--cc=vladimir.medvedkin@intel.com \
--cc=yanzhirun_163@163.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