feat(debug): engine-fed FPS counter + sorted per-unit profiler tables (labelle-engine#380) by apotema · Pull Request #640 · labelle-toolkit/labelle-assembler

apotema · 2026-07-19T16:39:59Z

What

Rewires the debug inspector's Performance section onto the engine's profiler APIs — the UI half of labelle-engine#380 (engine data side shipped in labelle-engine#733 + labelle-engine#784).

Real FPS: the header now reads Game.frameStats() / frameTimeMs() (engine's always-on FrameProfiler). The previous code was a Zig-0.16-migration stub faking a fixed 16ms counter, so the FPS readout had been hardwired to ~62.5 since std.time.milliTimestamp went away. The mini frame-time graph now renders Game.frameHistory() (real dt ring, newest 40 frames).
Sorted per-unit tables: Game.scriptProfileRows() / pluginProfileRows() typed rows replace the stale hand-rolled @ptrCast layouts (which silently mismatched the engine's current Stat-based rows). Scripts and Plugins render as tables sorted by per-frame cost (tick+postTick+drawGui; one-shot setup excluded from the sort key but shown in its own column), - for never-ran phases, severity markers (* >1ms, ! >5ms) and per-group totals.
Live capture without the env var: arms Game.setProfilingCapture(true) while the panel + Show Performance are visible, hands back null on close so a user's LABELLE_PROFILE headless dump keeps running.
LABELLE_DEBUG_OPEN=1 opens the inspector at boot (headless screenshot verification, skips the F12).

All engine reads are @hasDecl/@hasField-gated: against an engine without the new API the section degrades to a hint label, and older row layouts read missing phases as 0 — no compile break for games pinning older engines.

Verification

Ran on flying-platform-labelle (bgfx desktop) via local: overrides for this plugin + the engine PR branch: inspector opens at boot, FPS header live, per-script/per-plugin tables populate, and the profiler's 120-frame log dumps fire purely from the panel arming capture (no LABELLE_PROFILE set). Capture overhead measured at ~51 ns per dispatched unit while the panel is open (~2.3 µs/frame at FP scale), ~2 ns when closed.

Plugin tests: sorting excludes setup, severity marks, phase-cell formatting, old-row tolerance (zig build test in plugins/debug, exit-checked).

Pins

Full panel needs a labelle-engine release cut from labelle-engine#784 (next minor). Degrades gracefully below that. Follow-ups tracked on the engine PR: colored severity labels (needs labelColored on core's GuiInterface + imgui adapter), per-language-script breakdown via labelle-scripting.

https://claude.ai/code/session_011szWvquoss1yNX7KWSKCaM

^{Need help on this PR? Tag /codesmith with what you need. Autofix is disabled.}

… (labelle-engine#380) Rewires the inspector's Performance section onto the engine's profiler APIs (engine PR labelle-engine#380 follow-up of #733): - FPS header now reads Game.frameStats()/frameTimeMs() — the always-on engine FrameProfiler — replacing the 0.16-migration stub that faked a 16ms counter (FPS was hardwired ~62.5 since std.time.milliTimestamp went away). Mini-graph reads Game.frameHistory() (real dt ring). - Per-unit tables: Game.scriptProfileRows()/pluginProfileRows() typed rows replace the stale hand-rolled @ptrCast layouts (which no longer matched the engine's Stat-based rows). Scripts and Plugins render as tables sorted by per-frame cost (tick+post+gui, setup excluded), columns per lifecycle phase, '-' for never-ran, severity markers (* >1ms, ! >5ms — text until GuiInterface grows colored labels), per-group ms totals. - Live capture without env var: arms Game.setProfilingCapture(true) while the panel + Show Performance are visible, hands back null on close so a user's LABELLE_PROFILE headless dump keeps running. - LABELLE_DEBUG_OPEN=1 opens the inspector at boot (headless screenshot verification; skips the F12). - All engine reads @hasDecl/@hasField-gated: against an engine without the API (< 2.5) the section degrades gracefully, older row layouts read missing phases as 0. Tests: sorting excludes setup, severity marks, phase-cell formatting, old-row tolerance. Requires labelle-engine with Game.frameStats/scriptProfileRows for the full panel (2.4.0 + labelle-engine#380 follow-up; degrades below that). Claude-Session: https://claude.ai/code/session_011szWvquoss1yNX7KWSKCaM

apotema · 2026-07-19T16:40:11Z

Engine API side: labelle-toolkit/labelle-engine#784.

coderabbitai · 2026-07-19T16:40:11Z

Warning

Review limit reached

You’ve reached a temporary PR review limit under our Fair Usage Limits Policy.

Your recent review volume is higher than typical usage, so adaptive limits are currently applied.

Next review available in: 43 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 1bc949ca-60d9-4df2-b08d-4765eafe177a

📥 Commits

Reviewing files that changed from the base of the PR and between 6563822 and 2b927f6.

📒 Files selected for processing (1)

plugins/debug/src/root.zig

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch feat/380-debug-inspector-perf

_{Comment @coderabbitai help to get the list of available commands.}

gemini-code-assist

Code Review

This pull request refactors the performance and FPS tracking in plugins/debug/src/root.zig to utilize the engine's built-in profilers, introducing structured performance tables for scripts and plugins along with corresponding unit tests. The reviewer identified three compatibility and UI issues: first, if a performance table is collapsed, the total time incorrectly displays as 0.00ms because the summation is skipped; second, game.frameTimeMs() is called without a comptime check, risking compilation failures on older engines; and third, game.pluginProfileRows() similarly lacks a @hasDecl check, which could also break builds on older engine versions.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

gemini-code-assist · 2026-07-19T16:42:59Z

+    var total_ns: u64 = 0;
+    if (Gui.beginTable(title, cols)) {
+        Gui.tableNextRow();
+        _ = Gui.tableNextColumn();
+        Gui.label("name");
+        _ = Gui.tableNextColumn();
+        Gui.label("tick");
+        if (show_post) {
+            _ = Gui.tableNextColumn();
+            Gui.label("post");
        }
+        _ = Gui.tableNextColumn();
+        Gui.label("gui");
+        _ = Gui.tableNextColumn();
+        Gui.label("setup");
+
+        for (rows) |r| {
+            total_ns += r.frameNs();
+            Gui.tableNextRow();
+            var name_buf: [96]u8 = undefined;
+            _ = Gui.tableNextColumn();
+            Gui.label(std.fmt.bufPrintZ(&name_buf, "{s}", .{r.name}) catch "?");
+            var cell: [24]u8 = undefined;
+            _ = Gui.tableNextColumn();
+            Gui.label(fmtPhase(&cell, r.tick_ns));
+            if (show_post) {
+                _ = Gui.tableNextColumn();
+                Gui.label(fmtPhase(&cell, r.post_ns));
+            }
+            _ = Gui.tableNextColumn();
+            Gui.label(fmtPhase(&cell, r.gui_ns));
+            _ = Gui.tableNextColumn();
+            Gui.label(fmtPhase(&cell, r.setup_ns));
+        }
+        Gui.endTable();
    }


If Gui.beginTable returns false (e.g., when the table is collapsed or clipped), the loop that sums total_ns is skipped entirely. This causes the "Total" label printed outside the table to incorrectly display 0.00ms. Summing the total_ns before the beginTable check ensures the total remains accurate regardless of the table's visibility state.

var total_ns: u64 = 0; for (rows) |r| total_ns += r.frameNs(); if (Gui.beginTable(title, cols)) { Gui.tableNextRow(); _ = Gui.tableNextColumn(); Gui.label("name"); _ = Gui.tableNextColumn(); Gui.label("tick"); if (show_post) { _ = Gui.tableNextColumn(); Gui.label("post"); } _ = Gui.tableNextColumn(); Gui.label("gui"); _ = Gui.tableNextColumn(); Gui.label("setup"); for (rows) |r| { Gui.tableNextRow(); var name_buf: [96]u8 = undefined; _ = Gui.tableNextColumn(); Gui.label(std.fmt.bufPrintZ(&name_buf, "{s}", .{r.name}) catch "?"); var cell: [24]u8 = undefined; _ = Gui.tableNextColumn(); Gui.label(fmtPhase(&cell, r.tick_ns)); if (show_post) { _ = Gui.tableNextColumn(); Gui.label(fmtPhase(&cell, r.post_ns)); } _ = Gui.tableNextColumn(); Gui.label(fmtPhase(&cell, r.gui_ns)); _ = Gui.tableNextColumn(); Gui.label(fmtPhase(&cell, r.setup_ns)); } Gui.endTable(); }

Fixed in 2544c55. total_ns is now summed unconditionally from the rows BEFORE the beginTable branch, so the Total footer is correct even when the table body is collapsed/clipped (beginTable returns false).

gemini-code-assist · 2026-07-19T16:42:59Z

+    const st = game.frameStats();
+    var fps_buf: [64]u8 = undefined;
+    Gui.label(std.fmt.bufPrintZ(&fps_buf, "FPS: {d:.0} | Frame: {d:.1}ms", .{ st.fps, game.frameTimeMs() }) catch "?");


While frameStats is gated at comptime, game.frameTimeMs() is called directly without a check. If an older engine version implements frameStats but lacks frameTimeMs, compilation will fail. Adding a comptime @hasDecl check for frameTimeMs ensures graceful degradation and prevents compile breaks.

const st = game.frameStats(); const frame_ms = if (comptime @hasDecl(Game, "frameTimeMs")) game.frameTimeMs() else 0.0; var fps_buf: [64]u8 = undefined; Gui.label(std.fmt.bufPrintZ(&fps_buf, "FPS: {d:.0} | Frame: {d:.1}ms", .{ st.fps, frame_ms }) catch "?");

Fixed in 2544c55. frameStats and frameTimeMs are now folded under a single capability check — !(@hasDecl(Game, "frameStats") and @hasDecl(Game, "frameTimeMs")) — so an engine exposing one but not the other still compiles (degrades to the 'n/a' label).

This call site IS gated — it's the same single frameTimeMs() call the :134 fix folded; the line moved to 159 after the insertTopRow addition shifted the file. drawFpsHeader early-returns at the top under one capability check: if (comptime !(@hasDecl(Game, "frameStats") and @hasDecl(Game, "frameTimeMs"))) { Gui.label("FPS: n/a …"); return; } — so by line 159 both accessors are guaranteed present. grep -n frameTimeMs shows exactly one call (line 159) plus the comment on 152; there is no second ungated site. (2544c55)

Re-anchor of the single frameTimeMs() call already gated (round 2, 2544c55): drawFpsHeader early-returns under one check gating BOTH frameStats and frameTimeMs, so the call is only reached when both are present. grep -n frameTimeMs shows exactly one call site. No change needed.

gemini-code-assist · 2026-07-19T16:42:59Z

+    {
+        const src = game.pluginProfileRows();
+        const n = @min(src.len, MAX_PERF_ROWS);
+        for (src[0..n], 0..) |r, i| rows_buf[i] = .{
+            .name = r.name,
+            .setup_ns = phaseNs(r, "setup"),
+            .tick_ns = phaseNs(r, "tick"),
+            .post_ns = phaseNs(r, "post_tick"),
+            .gui_ns = phaseNs(r, "draw_gui"),
+        };
+        drawPerfGroup(Gui, "Plugins", rows_buf[0..n], true);
+    }


The outer function drawPerfTables only checks for the presence of scriptProfileRows. Calling game.pluginProfileRows() without a corresponding @hasDecl check can cause compilation failures on older engines that might only support script profiling. Wrapping the plugin profiling block in a comptime check ensures robust compatibility.

if (comptime @hasDecl(Game, "pluginProfileRows")) { const src = game.pluginProfileRows(); const n = @min(src.len, MAX_PERF_ROWS); for (src[0..n], 0..) |r, i| rows_buf[i] = .{ .name = r.name, .setup_ns = phaseNs(r, "setup"), .tick_ns = phaseNs(r, "tick"), .post_ns = phaseNs(r, "post_tick"), .gui_ns = phaseNs(r, "draw_gui"), }; drawPerfGroup(Gui, "Plugins", rows_buf[0..n], true); }

Fixed in 2544c55. drawPerfTables now gates each accessor independently: the scripts block is under @hasDecl(Game, "scriptProfileRows") and the plugins block under @hasDecl(Game, "pluginProfileRows"), so an engine exposing only one still compiles.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 27fbba2311

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-19T16:50:08Z

+        const n = @min(src.len, MAX_PERF_ROWS);
+        for (src[0..n], 0..) |r, i| rows_buf[i] = .{


Select top profiler rows before truncating

When a game has more than 64 scripts, this slices to the first 64 rows before sorting, so an expensive script registered later is never copied into rows_buf and the table is only sorted within that prefix; the same pattern below affects plugins. Please sort/select across the full source rows and only then cap the displayed list.

Useful? React with 👍 / 👎.

Fixed in 2544c55. Replaced the truncate-before-sort with a streaming insertTopRow that keeps the top MAX_PERF_ROWS by per-frame cost in one pass, so the priciest rows survive regardless of source order (late-registered hot scripts included). Added a regression test: fill the cap with cheap rows, then insert the priciest ones AFTER the buffer is full — they land at the front and the smallest is evicted.

apotema · 2026-07-19T17:08:50Z

Verification (on-screen panel)

Built a sokol demo with three busy-work scripts + the debug plugin, ran headless surfaceless (sokol true-surfaceless capture), overlay opened via LABELLE_DEBUG_OPEN=1, captured with the engine screenshot facility. The panel renders:

FPS: 60 | Frame: 16.7ms header + min/avg/max + frame-time mini-graph [####…] — all from the engine FrameProfiler.
Scripts table sorted by per-frame cost: physics_u_sim 0.75, worker_u_ai 0.16, needs_u_manager 0.03, Total Scripts: 0.94ms — correct per-script names, setup/gui columns shown.
Plugins table with the debug plugin's gui 0.26ms, Total Plugins: 0.26ms.
Live capture armed purely by opening the panel — the profiler's log ranking fired ([script] physics_u_sim: worst=4.82ms avg=1.36ms, …) with no LABELLE_PROFILE set, confirming setProfilingCapture works end to end.

Known wart → follow-up (assembler codegen)

The per-plugin rows label the plugin by its tuple index (2) rather than its name (debug). Root cause is in this repo's codegen: registries.zig emits engine.SystemRegistry(.{ @import("labelle-gfx"), @import("<plugin>"), … }) — an anonymous tuple, so SystemRegistry's field.name is "0"/"1"/"2". The per-SCRIPT table (the issue's primary ask) is correct because ScriptRunner iterates named decls. Fix is to emit a named struct (.{ .labelle_gfx = …, .debug = … }, sanitizing -→_) so field.name carries the real name — small change, but it touches the bit-identical golden-file suite, so it's deliberately split out rather than folded into this PR. Filing as a follow-up.

Overhead: ~51 ns per dispatched unit while capturing, ~2 ns when off (engine PR has the microbench).

…pluginProfileRows Round-1 review fixes on the inspector Performance section: - Top-N truncation before sort (codex P2): drawPerfTables sliced the first MAX_PERF_ROWS source rows THEN sorted, so with >64 scripts the priciest (registered later) never appeared. Replaced with a streaming insertTopRow that keeps the top-N by per-frame cost regardless of source order — sort-then-truncate semantics in one pass. Regression test: cap+ rows with the priciest arriving AFTER the buffer is full lands at the front. - Total skipped when table collapses (gemini): total_ns was summed inside the beginTable body but the Total label prints outside, so a collapsed/clipped table showed a stale/zero total. Total is now summed unconditionally from the rows before the render branch. - frameTimeMs unguarded (gemini): game.frameTimeMs() was called while only frameStats was @hasDecl-gated. Both now fold under one capability check. - pluginProfileRows unguarded (gemini): drawPerfTables gated only on scriptProfileRows but called pluginProfileRows. Each accessor is now gated independently, so an engine exposing only one still compiles. Plugin zig build test green. Claude-Session: https://claude.ai/code/session_011szWvquoss1yNX7KWSKCaM

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2544c55f62

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-19T17:40:25Z

+    // if `beginTable` returns false (collapsed / clipped) the row loop
+    // is skipped, and a total computed inside it would print stale/zero.
+    var total_ns: u64 = 0;
+    for (rows) |r| total_ns += r.frameNs();


Sum profiler totals before capping rows

When a game has more than MAX_PERF_ROWS scripts or plugins, rows is already the capped top-64 slice built by insertTopRow, so the Total Scripts/Total Plugins footer omits every hidden row. That under-reports total frame cost exactly in larger projects; accumulate the total while iterating all game.*ProfileRows() before truncating the displayed rows.

Useful? React with 👍 / 👎.

Fixed in dd66d1d. drawPerfTables now accumulates the full total (and full unit count) across EVERY source row, before the top-N cap, and passes both to drawPerfGroup; the footer prints that full total and, when truncated, notes (top N of M). drawPerfGroup no longer re-sums its capped rows. Regression test perf total sums ALL units, not just the displayed top-N: MAX_PERF_ROWS+10 equal-cost units → full total counts all of them and strictly exceeds the displayed top-N sum.

Round-2 review (codex, root.zig:235): the Total footer summed only the top-N rows kept by insertTopRow, so with more than MAX_PERF_ROWS scripts/plugins the total undercounted (it dropped the cost of every truncated unit). drawPerfTables now accumulates the full total (and full unit count) across EVERY source row, before the top-N cap, and passes both to drawPerfGroup. The footer prints the full total and, when truncated, notes "(top N of M)" so the display set vs the accounted set is unambiguous. drawPerfGroup no longer re-sums its (capped) rows. Regression test: MAX_PERF_ROWS+10 equal-cost units -> full total counts all of them and strictly exceeds the sum of the displayed top-N. Plugin zig build test green. Claude-Session: https://claude.ai/code/session_011szWvquoss1yNX7KWSKCaM

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dd66d1de00

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-19T18:34:24Z

+    const Game = @TypeOf(game.*);
+    var rows_buf: [MAX_PERF_ROWS]PerfRow = undefined;
+
+    if (comptime @hasDecl(Game, "scriptProfileRows")) {


Gate profiler tables when capture cannot be armed

When running with the repo's default engine 2.5.0 and no LABELLE_PROFILE, Game already has scriptProfileRows/pluginProfileRows but lacks setProfilingCapture, so syncProfilingCapture returns without enabling recording. This check still enters the table-rendering path and shows - cells and 0.00ms totals as if they were real live timings, even though nothing is being sampled. Please gate the tables on capture support or LABELLE_PROFILE in this compatibility path, or show the fallback hint instead.

Useful? React with 👍 / 👎.

Fixed in 2b927f6. drawPerfTables now gates up front on capture being genuinely armable AND active:

Engine lacking setProfilingCapture/profilingCaptureActive (< 2.6, e.g. 2.5 which has the row accessors but not the capture-control API): prints profiling capture unavailable - requires engine >= 2.6.0 and returns — no more wall of -/0.00ms.

Capture-capable but recording currently off: prints profiling capture inactive and returns.

Only when game.profilingCaptureActive() is true does it render the tables.

Test drawPerfTables gates on capture armability: a label-capturing stub Gui drives it against three mock engines (old/no-API → 'unavailable', new/active → tables shown, new/idle → 'inactive'), asserting the tables appear only when active.

Round-3 review (codex P2, root.zig:198): against an engine that exposes scriptProfileRows/pluginProfileRows (>= 2.4) but NOT setProfilingCapture (< 2.6), syncProfilingCapture is a no-op, nothing enables recording, and every row reads back 0 — so the tables rendered a wall of `-`/`0.00ms` that reads as a broken panel. drawPerfTables now gates up front: - No setProfilingCapture/profilingCaptureActive (< 2.6): print "profiling capture unavailable - requires engine >= 2.6.0" and return. - Capture-capable but recording currently off: print "profiling capture inactive" and return. - Only when capture is genuinely active does it render the tables. Test: a label-capturing stub Gui drives drawPerfTables against three mock engines (old/no-API, new/active, new/idle) asserting the right line shows and the tables appear only when active. Plugin zig build test green. Claude-Session: https://claude.ai/code/session_011szWvquoss1yNX7KWSKCaM

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2b927f6795

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-19T18:49:36Z

+    if (comptime !@hasDecl(Game, "setProfilingCapture") or !@hasDecl(Game, "profilingCaptureActive")) {
+        Gui.label("profiling capture unavailable - requires engine >= 2.6.0");
+        return;


Preserve env-enabled profiling on 2.5 engines

When a game is pinned to the released engine that already has scriptProfileRows/pluginProfileRows but not the new capture-control methods, launching with LABELLE_PROFILE=1 still produces real per-unit rows; this unconditional early return now hides them behind the “requires engine >= 2.6.0” message. Please allow the no-capture-control compatibility path to render when the env gate is already enabled, and only suppress the zero-row/no-env case.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-07-19T18:49:36Z

-        updateFpsTracking();
+        // Keep the engine's live per-unit capture in step with panel
+        // visibility (runs even when hidden, so closing disarms it).
+        syncProfilingCapture(game, debug_visible and show_perf);


Skip capture on wasm profiler builds

On wasm/editor-preview builds, the engine profiler clock returns 0, but this still forces setProfilingCapture(true) whenever the panel is open; profilingCaptureActive() then reports true and drawPerfTables renders rows of -/0.00ms instead of the fallback, even though no real per-unit timings can be sampled. Gate this override to targets with a usable profiler clock, or avoid rendering the tables in that context.

Useful? React with 👍 / 👎.

apotema mentioned this pull request Jul 19, 2026

feat(#380): runtime profiling-capture toggle + setup/plugin-drawGui phase timing labelle-toolkit/labelle-engine#784

Merged

gemini-code-assist Bot reviewed Jul 19, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Jul 19, 2026

View reviewed changes

apotema mentioned this pull request Jul 19, 2026

feat: language capability rows — assembler language-agnosticism (#619) #643

Merged

chatgpt-codex-connector Bot reviewed Jul 19, 2026

View reviewed changes

apotema merged commit 875369f into main Jul 19, 2026
4 checks passed

chatgpt-codex-connector Bot reviewed Jul 19, 2026

View reviewed changes

apotema deleted the feat/380-debug-inspector-perf branch July 19, 2026 18:55

		const n = @min(src.len, MAX_PERF_ROWS);
		for (src[0..n], 0..) \|r, i\| rows_buf[i] = .{

Uh oh!

Conversation

apotema commented Jul 19, 2026 • edited by blacksmith-sh Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Verification

Pins

Uh oh!

apotema commented Jul 19, 2026

Uh oh!

coderabbitai Bot commented Jul 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review limit reached

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

apotema commented Jul 19, 2026

Verification (on-screen panel)

Known wart → follow-up (assembler codegen)

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Jul 19, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

apotema commented Jul 19, 2026 •

edited by blacksmith-sh Bot

Loading

coderabbitai Bot commented Jul 19, 2026 •

edited

Loading