Skip to content

docs: clarify the source of the Borderlands 4 RR timing - #281

Merged
phantoslayer21 merged 2 commits into
mainfrom
docs/bl4-attribution
Nov 18, 2026
Merged

phantoslayer21 merged 2 commits into
mainfrom
docs/bl4-attribution

Conversation

@phantoslayer21

@phantoslayer21 phantoslayer21 commented Nov 18, 2026 •

Copy link
Copy Markdown
Owner

Why

The exit summary of the Borderlands 4 log can mislead a reader. clean/installer/tests/logs/bl4-rr-only.log:5 ends with:

diag: summary (process exit): support queries 13 (RR: yes 0 / no 0), RR lookups 0, RR instances created 0, RR evaluates 0

while lines 1-4 show Ray Reconstruction (RR) timing and destroy instance 1 after 36044 frames (0 failed).

The two do not conflict:

  • The summary counters (d_rr_instances_, d_rr_evals_, d_rr_lookups_) are incremented only in the game-interface proxy: clean/rr_f/proxy/src/proxy_core.cpp:511, :923, :957, :1174. The line is written at proxy_core.cpp:261-269.
  • The direct (NGX) route creates and evaluates instances in clean/rr_f/proxy/src/ngx_rr.cpp:334 and :563 and does not touch those counters.
  • Borderlands 4 creates Ray Reconstruction itself, so it uses the direct route (clean/installer/README.md:974-976).
  • The timing lines (1.589 ms, 1.592 ms, 1.589 ms) are written by the RR runtime, clean/rr_f/runtime/src/rr_runtime.cpp:2404. The destroy line is rr_runtime.cpp:2461.

So the zeros are expected for this game. The COMPATIBILITY.md text said the figure was "Ray Reconstruction's own GPU time per frame" without saying where it comes from, which left that question open.

What changed

Only COMPATIBILITY.md (the paragraph under the table). It now says:

  • the figure is the GPU time per frame that WrongGPU's Ray Reconstruction runtime logged for its instance 1: about 1.59 ms per frame in the logged timing windows (300, 300 and 44 frames), render 642x361, output 1920x1080; the instance was destroyed after 36044 frames with 0 failed;
  • the log's last line, "RR instances created 0, RR evaluates 0", counts only the game-interface route;
  • Borderlands 4 uses the direct route, so those zeros are expected and do not mean Ray Reconstruction did not run;
  • the Ray Reconstruction mode and the WrongGPU version of that run are still not recorded in the excerpt.

The numbers and the source file are unchanged.

Checks

  • Numbers check (3 rows: 1.589 at COMPATIBILITY.md:15, 1.59 at :30, 36044 at :31, each against the log): number gate: 3 row(s), 0 failing.
  • tools/ci/check_line_endings.py: 1460 files checked, 0 problems.
  • clean/installer/tools/check_source_size.py: 125 files, 0 at or over the limit. check_wording.py: 84 files, 0 hits. (Neither reads COMPATIBILITY.md.)
  • Leak scan of clean/ at HEAD: scanned 1393 files at HEAD: 0 hit(s). Scanning COMPATIBILITY.md itself reports 8 name hits, the same 8 as on main. They are existing mentions of the DLSS product name in the requirements text. This change adds none.

Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
@phantoslayer21

Copy link
Copy Markdown
Owner Author

The second reader's point is fixed in deb74d9. "Over 36044 frames" read as a whole-run average. The log has three separate timing windows (bl4-rr-only.log:1-3: 300, 300 and 44 frames, 1.589, 1.592 and 1.589 ms) and, separately, the instance's lifetime count (:4: destroyed after 36044 frames, 0 failed). COMPATIBILITY.md now says "about 1.59 ms per frame in the logged timing windows" and, as its own sentence, "The instance was destroyed after 36044 frames with 0 failed." The numbers check is unchanged: 3 rows, 0 failing. I also updated the PR description to match.

@phantoslayer21
phantoslayer21 merged commit 0a3a4f9 into main Nov 18, 2026
10 checks passed
@phantoslayer21
phantoslayer21 deleted the docs/bl4-attribution branch November 18, 2026 18:07
phantoslayer21 pushed a commit that referenced this pull request Nov 18, 2026
* docs: README: players first, a shorter developer overview

What you need, install, tested games, Model L and Model K and problems come first; the developer overview links out to the detailed docs instead of repeating them. Every fact kept or moved (links and anchors checked).

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: COMPATIBILITY: one section per game instead of long table cells

Keeps the Borderlands 4 timing paragraph of #281 word for word, so #281 then this applies cleanly.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: CONTRIBUTING: shorter sentences around an unchanged clean-room statement

The clean-room statement is byte-identical to main.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: ROADMAP: drop items 0.6.5 already shipped

Preset K and the 0.6.1 heading are done; the open Hogwarts Legacy and RESIDENT EVIL items stay.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: SECURITY: clearer scope and reporting steps

Same scope and contacts as before.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: BUILDING: steps as lists and command blocks, commands for 0.6.5

The tool versions are still the ones recorded for 0.6.0 (an open question in the pull request).

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: RELEASING: the release gates as a checklist

Every gate and step of main's text is kept.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: TESTING-OTHER-GAMES: steps for 0.6.5, Model L and Model K testing

Model K is converted from a DLSS file from one of the player's games, as the release notes say.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

* docs: CLEAN-ROOM: say the tree now carries 0.6.5

It was first published as the 0.6.0 snapshot with one commit and no private history; it now carries 0.6.5. "Left out:" became "Not included:". Legal text: the maintainer should check these two sentences.

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>

---------

Signed-off-by: Jonathan Shin <85975340+j0nsh1n@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant