Fix latency issue with rl pacing by Sakoram · Pull Request #1649 · OpenVisualCloud/Media-Transport-Library

Sakoram · 2026-07-10T13:11:07Z

No description provided.

Hold the first content packet until its target TSC and derive default gates from the same media-clock instant as the RTP timestamp. Preserve exact-user TAI requests and cover ST20/ST40 timing boundaries. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

Release held mbufs on both ports before replacing shared pools while preserving sibling queue recovery requests. Cover repeated cleanup, simultaneous requests, and partial-burst ownership. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

Check every default, user, exact, jittered, and interlaced frame against a hardware-tested 300 us regression ceiling. Keep RTP timing and frame-count checks strict without percentage outlier state. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

Run build and test tools with bounded decoded output and checkout-local MTL and DPDK libraries. Keep gtest routing shell-free and report actionable NoCtx assertion failures. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

DawidWesierski4

Gonna do another PR for now this all

DawidWesierski4 · 2026-08-10T08:56:27Z

@@ -919,6 +919,17 @@ static uint64_t st_muldiv_u64_round_closest(uint64_t value, uint64_t multiplier,
  return (uint64_t)quotient;
 }

+uint64_t st_tai_round_to_media_clk_ns(uint64_t tai_ns, uint32_t sampling_rate) {
+  if (!sampling_rate) {
+    err("%s, invalid sampling rate\n", __func__);
+    return 0;
+  }
+
+  uint64_t tick = st_muldiv_u64_round_closest(tai_ns, sampling_rate, NS_PER_S);
+  if ((__uint128_t)tick * NS_PER_S / sampling_rate >= UINT64_MAX) return UINT64_MAX;
+  return st_muldiv_u64_round_closest(tick, NS_PER_S, sampling_rate);
+}
+
 uint32_t st10_tai_to_media_clk(uint64_t tai_ns, uint32_t sampling_rate) {
  if (!sampling_rate) {
    err("%s, invalid sampling rate\n", __func__);


NITPIC (it already is a mess here):
if we are making those public then they don't belong in st_fmt.c how are those funcitons connected to format conversion? Those should go to utils or st_api
nothing to do with fortmat conversion

DawidWesierski4 · 2026-08-10T09:07:08Z

+  if (!((s->ops.flags & ST20_TX_FLAG_EXACT_USER_PACING) && required_tai) &&
+      !(s->ops.flags &
+        (ST20_TX_FLAG_USER_TIMESTAMP | ST20_TX_FLAG_RTP_TIMESTAMP_EPOCH)) &&
+      !s->ops.rtp_timestamp_delta_us) {
+    start_time_tai =
+        st_tai_round_to_media_clk_ns(start_time_tai, s->fps_tm.sampling_clock_rate);


NITPIC(very corner corner case idk if you change it):
the required_tai could be 0, we are doing so in the tests
i think much more elegant approach would be to do somthing like this as required_tai being 0 and user pacing existing is an option

if (!(s->ops.flags & (ST20_TX_FLAG_EXACT_USER_PACING | ST20_TX_FLAG_USER_TIMESTAMP | ST20_TX_FLAG_RTP_TIMESTAMP_EPOCH)) && !s->ops.rtp_timestamp_delta_us) { start_time_tai = st_tai_round_to_media_clk_ns(start_time_tai, s->fps_tm.sampling_clock_rate); }

DawidWesierski4 · 2026-08-10T10:14:20Z

  }

-  return 0;
+  return ret;


NITPICK: we should propablly check the result then ?

DawidWesierski4 · 2026-08-10T10:19:17Z

+  if (s->inflight[i][0]) rte_pktmbuf_free_bulk(s->inflight[i], s->bulk);
+  if (s->trs_inflight_num[i])
+    rte_pktmbuf_free_bulk(&s->trs_inflight[i][s->trs_inflight_idx[i]],
+                          s->trs_inflight_num[i]);
+  if (s->trs_inflight_num2[i])
+    rte_pktmbuf_free_bulk(&s->trs_inflight2[i][s->trs_inflight_idx2[i]],
+                          s->trs_inflight_num2[i]);
+  for (unsigned int j = 0; j < s->trs_pad_inflight_num[i]; j++) {
+    if (s->pad[i][ST20_PKT_TYPE_NORMAL])
+      rte_pktmbuf_free(s->pad[i][ST20_PKT_TYPE_NORMAL]);


There should be a mention about a registered rte free safty here ?
What if the registered cleanup flow per packet gonna go when this clenup works ?
Is this safe ?

DawidWesierski4 · 2026-08-10T10:21:40Z

+  /* Payload pools are shared by redundant ports, so release all software-held
+   * mbufs before replacing them. */
+  st_tx_video_transmitter_state_cleanup(s);
+  uint64_t recovery_tsc = mt_get_tsc(impl);
+  for (uint8_t i = 0; i < s->ops.num_port; i++)
+    s->last_burst_succ_time_tsc[i] = recovery_tsc;


there is no test covrege for this, we should probs add some as this is a problematic flow

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from 02a56af to 9b04ad5 Compare July 16, 2026 11:31

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from 9b04ad5 to 9831aaf Compare July 23, 2026 14:10

Sakoram changed the base branch from main to Fix-complaince-checks July 23, 2026 14:12

Sakoram force-pushed the Fix-complaince-checks branch 2 times, most recently from dcfa843 to 617c0b6 Compare July 28, 2026 14:17

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from 9831aaf to 6c20f67 Compare July 29, 2026 09:21

Sakoram force-pushed the Fix-complaince-checks branch 5 times, most recently from 257fe07 to b67b50d Compare July 31, 2026 08:15

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from 6c20f67 to db1e27f Compare July 31, 2026 10:57

Base automatically changed from Fix-complaince-checks to main July 31, 2026 11:39

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from db1e27f to dc4e56b Compare July 31, 2026 11:44

Sakoram added 4 commits August 5, 2026 13:15

Fix: Quiesce redundant TX recovery

ed1e422

Release held mbufs on both ports before replacing shared pools while preserving sibling queue recovery requests. Cover repeated cleanup, simultaneous requests, and partial-burst ownership. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

Fix: Harden MCP test execution

ce013ef

Run build and test tools with bounded decoded output and checkout-local MTL and DPDK libraries. Keep gtest routing shell-free and report actionable NoCtx assertion failures. Signed-off-by: Kasiewicz, Marek <marek.kasiewicz@intel.com>

Sakoram force-pushed the fix-latency-issue-with-rl-pacing branch from dc4e56b to ce013ef Compare August 5, 2026 13:46

Sakoram marked this pull request as ready for review August 5, 2026 13:47

Sakoram requested review from DawidWesierski4, awilczyns, moleksy and soopel as code owners August 5, 2026 13:47

Sakoram mentioned this pull request Aug 5, 2026

Unexpected variable latency in interlaced mode #1653

Open

DawidWesierski4 reviewed Aug 10, 2026

View reviewed changes

Navigation Menu

Uh oh!

Fix latency issue with rl pacing - #1649

Fix latency issue with rl pacing#1649
Sakoram wants to merge 4 commits into
mainfrom
fix-latency-issue-with-rl-pacing

Sakoram commented Jul 10, 2026

Uh oh!

DawidWesierski4 left a comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Uh oh!

DawidWesierski4 Aug 10, 2026

Uh oh!

DawidWesierski4 Aug 10, 2026

Uh oh!

DawidWesierski4 Aug 10, 2026

Uh oh!

DawidWesierski4 Aug 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Sakoram commented Jul 10, 2026

Uh oh!

DawidWesierski4 left a comment

Choose a reason for hiding this comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

DawidWesierski4 Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants