P12 Hummer H3 XDF Release

[LSxHunter]

Hey guys! I've been actively working on a P12 XDF after noticing one hadn't been put online. I'm a software dev and AI engineer by trade.

My biggest ask is for more information, logs, ect. I do have a custom logger and dashboard I'm building for the P12 as well to accomplish this myself but obviously more data is better.

https://lsxhunter.com/p12/

The website will be updated on a regular cadence until 100% complete.

Hope this helps some of you!

[LSxHunter]

The numbers

• 565 items (310 tables/curves, 255 scalars), up from 359 in v3.2. I skipped publishing v3.3 so I could do a cleanup pass first.
• About 41% of the cal region (0x20000-0x3FFFB) by bytes is now in the download. Around 73% is defined in total; the rest of the defined share is held back on purpose (reference-only models, fault handling, or not proven yet).
• TunerPro cross-check: loaded on a stock read and exported, 565 of 565 match our own decoder, zero failures.
• 38 items confirmed on the truck with logs. Everything else is traced in the PCM's code and exercised in an emulator.

How it's proven
Every item is traced from the code that reads it, not guessed from hex. The emulator runs the whole OS scheduler at real task rates, and I replay my truck's logs through it. On a real drive it tracks the PCM's own logged values closely (desired throttle r=0.99, load 0.98, delivered torque 0.95, rpm exact). All 464 diagnostic PIDs are mapped to the RAM they report. Every finding has a numbered entry in an evidence ledger (700+ now) that a script re-proves from the stock bin.
The open E40 definition here was a big help. Matching routines between the two OSes turned up a lot of leads, but nothing ships on the E40's word. Each one gets re-traced in the P12's own code first.
What's new since v3.2

• Trans: more shift/TCC schedules, cold-engine shift limits, TCC slip targets, garage engagement pressure in psi, TAP learn limits, the trans torque model.
• Spark: the hot high-load spark ceiling (drops to an 18-24 deg table after ~18 s hot at high load), spark floors, knock detection threshold gains.
• Airflow: cam phasing logic (load-based target and region-based path), learned MAF/VE ratio.
• Fuel / idle / torque: stoich AFR, open-loop fuel gain, PE delay countdown, idle acquire logic, ETC torque request multipliers.
• Diagnostics: the per-code DTC status table plus monitor thresholds and timers. Standard warning: disabling emissions codes on a road vehicle may be illegal where you live. The readiness override and cat over-temp confirm switch are left out on purpose.

Heads-ups

• Five released tables are all zero on the stock H3 (DFCO entry spark, fuel cutoff rpm reduction, 3-4 upshift pressure temp modifier, LTFT learn min MAP, knock transient threshold adder). A non-zero edit switches that feature on rather than adjusting it.
• Correction: the v3.4 VE bank item text said a 9-11 deg desired cam blends toward banks 2-3. By the code it's banks 4-5 (the PCM halves its actual cam reading before comparing it with desired). Fixed in v3.4.1 (text only, same data).
• Main spark is a high/low octane pair blended by the learned octane scaler. Change both.

What would help
If you've got a 12627884 truck, send your OS ID, what you changed (old and new values) and a log. Corrections help as much as confirmations. Two logs in particular would promote held-back items: one drive in 4-Lo, and one shift after a tire size change.

[kur4o]

Nice work already, do you have the 12627884 bin to share, I miss this file, found 885, but it is 2mb so not a near close.

[NSFW]

Bravo!

[LSxHunter]

Nice work already, do you have the 12627884 bin to share, I miss this file, found 885, but it is 2mb so not a near close.

Version 3.10 just went up. About 50% through it. https://lsxhunter.com/p12/