Timing benchmark: SDIO vs USB (RTL8733BS vs 8733BU) — confirm/refute the transport-latency hypotheses
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 25/100
- Issue type
- Feature
- Clarity
- Clearly specified
- Activity status
- Quiet
- Tech stack
- c
- Domain
- networking, performance
Research direction
Start with docs/timing-accuracy.md, tests/reglat.cpp, and tests/txegress_witness.cpp; first confirm dependencies #233 and #234, then settle the SDIO access model and comparison hardware. Run the USB and SDIO measurements with the shared witness and timesync tools, including ReadTsf under RX load and TA corrections. Done means a verdict on H1–H5 and the USB/SDIO/PCIe results documented in docs/timing-accuracy.md.
Written by the indexing model from the issue text.
Description
Goal
Once the RTL8733B port lands on both buses (#233 USB, #234 SDIO), the 8733BU vs 8733BS pair gives a clean apples-vs-apples experiment: same chip, same MAC, same firmware, same PHY — only the transport differs. Run it on one host with one common witness receiver to isolate the bus as the sole variable, and confirm or refute the SDIO position we predicted this session.
This extends the USB-vs-PCIe transport-latency work already in docs/timing-accuracy.md (tests/reglat.cpp, PR #228) with the SDIO middle point.
Depends on
- #233 — RTL8733BU USB HAL merged
- #234 — RTL8733BS SDIO transport merged
Background (measured this session)
| Transport | Register op | Jitter (p99−p50) |
|---|---|---|
| USB (8822EU) | ~68 µs | ~64 µs |
| PCIe 8821CE (MMIO) | ~2.2 µs | ~0.08 µs |
| SDIO | ? (predict single-digit µs) | ? |
Also measured: host-pushed data-frame submit-to-air jitter ~117 µs on a quiet 5 GHz channel over USB — almost entirely the USB + MAC-pipeline (CSMA minimal), i.e. the part a faster/more-deterministic bus should reduce.
Hypotheses to confirm/refute
- H1 — register-op latency. An SDIO CMD52/53 register round-trip is single-digit µs, sitting between USB (~68 µs) and PCIe (~2.2 µs). Measure with an SDIO
reglatequivalent. - H2 — no bulk-IN starvation.
ReadTsf()is reliable during active RX on SDIO, unlike USB (where the continuous bulk-IN flood starves the control read, forcing the timesync UE to use per-frametsfl). Measure ReadTsf success + latency under RX load, full-duplex. - H3 — TA actuator tightens. The closed-loop uplink timing-advance per-correction jitter drops from ~0.5–1.2 ms (USB) toward tens of µs on SDIO (the
AdjustBeaconTimingFineregister sequence). Re-runtimesync_ta_demoover SDIO. - H4 — TX-egress wall unchanged. Submit-to-air jitter for host-pushed data frames stays MAC/CSMA-bound with no TX-egress timestamp — SDIO improves the submit path but does not unlock µs-precise scheduled TX (a MAC/firmware limit, transport-independent). Measure with the
txegress_witnessbench. - H5 — clean isolation. 8733BU vs 8733BS on the same host + same witness isolates the bus as the only variable, so any delta is purely transport (not chip/firmware/PHY).
Method
- Port both buses (#233, #234).
- Build an SDIO
reglat+ reusetests/txegress_witness.cpp/timesync/timesync_ta_demo.sh. - Run 8733BU (USB) TX and 8733BS (SDIO) TX against a common witness receiver on the same host; collect register-op, ReadTsf-under-RX, TA-jitter, and submit-to-air numbers.
- Fold the resulting USB | SDIO | PCIe table into
docs/timing-accuracy.md.
Open questions
- SDIO userspace access model (per #234) determines how
reglatis even run — a kernel shim oversdio_func, or in-kernel measurement. This affects what "userspace latency" means for SDIO and must be settled with #234. - Which host provides both a USB port and an SDIO slot for the same-host comparison (an OpenIPC SoC dev board, or a desktop SDIO reader)?
Success criteria
A confirmed-or-refuted verdict on H1–H5, with a transport-latency table (register op, ReadTsf-under-RX reliability, TA per-correction jitter, submit-to-air jitter) across USB / SDIO / PCIe — the empirical basis for the transport recommendation in the LTE/5G-parity discussion.
- Dominant language
- C
- Stars
- 56
- Forks
- 29
- Avg merge
- 3h 43m
- Merged PRs (30d)
- 75
Getting set up
This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from OpenIPC/devourer
-
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
Maintainers usually reply within 1 day
-
Follow-ups left open by the 8822B fast bandwidth toggle fix (#584)Possibly taken @josephnef claimed this today. Openenhancement
Difficulty 4/5 3-5 days Newbie friendliness 42/100
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 12/100
OpenIPC/devourer#579 · 1 comment ·
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 12/100
Maintainers usually reply within 1 day
-
enhancement
Difficulty 5/5 Over a week Newbie friendliness 8/100
Maintainers usually reply within 1 day
All issues in OpenIPC/devourer
Similar issues
-
needs triage
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
389ds/389-ds-base#8157 ·
Maintainers usually reply within 2 days
-
Difficulty 2/5 1-3 hours Newbie friendliness 67/100
coolsnowwolf/lede#14305 ·
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Maintainers usually reply within 1 day
-
Needs-Triage
Difficulty 2/5 1-3 hours Newbie friendliness 66/100
Maintainers usually reply within 2 days