From 64b6dcb37b3c70d6bf7601f4bb195346ca3a69c3 Mon Sep 17 00:00:00 2001 From: Manan Khurana <86416632+Mkhurana3@users.noreply.github.com> Date: Mon, 9 Nov 2026 19:47:26 +0530 Subject: [PATCH 1/3] Ocarina of Time: note the Saria-cutscene GPU hang and the cost of task isolation on an M2 MacBook Air Adds one "Open" bullet to the game's note. On an M2 MacBook Air (16 GB), 6 of 6 runs from the autosave just before Saria gives Link the ocarina died within 158-241 s on main 1794a35a, in a GPU hang and then a revoked GPU. With DM_NVN_ISOLATE_TASK=1 from #1842 (not on main) two runs of about 8 and 12 minutes had no ErrorHang. Over the same 3,600 frames of the cutscene at render scale 0.75, the switch took 4.5 ms (9 %) more GPU time a frame, one pair played by hand. The same runs show first-use compiles of the foliage task's pipelines of 7.0, 7.1 and 10.5 s, and two pipelines of task a7f22b4a795ff4d1 that failed with XPC_ERROR_CONNECTION_INTERRUPTED. Only the note changes: no code, no front-end action. Co-Authored-By: Claude Sonnet 5.5 Claude-Session: https://claude.ai/code/session_01964yimrHXRP82dKYhi1BKj --- dm/games/0400486024DEC000/README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/dm/games/0400486024DEC000/README.md b/dm/games/0400486024DEC000/README.md index d880588ffe..4ebe6005b0 100644 --- a/dm/games/0400486024DEC000/README.md +++ b/dm/games/0400486024DEC000/README.md @@ -106,6 +106,7 @@ From an empty save: at about 45 s `A` (title), `A` (New Game), then the movie. F - The "Navi" name label in the Great Deku Tree scene was missing on an earlier chain and has not been checked since. Not played yet: the Deku Tree dungeon, combat, later areas. Sound opens with no red line; movies and speech were not listened to. Physics issues reported in PR 1299 were not seen on this route (walking, the ladder, the camera). - Resolved: dotted horizontal rows across the picture in play, in fog and light (GitHub issue 1537): the agl_ndsr_vr networks' HMMA calls outside a chain overwrote an accumulator a chain held ([An HMMA outside a chain overwrote the chain's accumulator](../../../docs/reference/shaders.md#an-hmma-outside-a-chain-overwrote-the-chains-accumulator)), since chain 23a1ea1db092; found by bisecting chains on the Kokiri Forest autosave (0f510fd5's chain clean, bd9f422d's not). Under the game's upscaler they had also grown into white blotches: [LDSM reads rows other lanes stored](../../../docs/reference/shaders.md#ldsm-reads-rows-other-lanes-stored). - No iOS shader build or iPhone run has been done. Its memory use makes an iPhone unlikely as it stands. +- Open: GPU hang after the Saria cutscene on an M2 MacBook Air (16 GB, macOS 27.2, main `1794a35a`, 2026-11-09; [#1326](https://github.com/seedds/nxnative/issues/1326), [#1690](https://github.com/seedds/nxnative/issues/1690)). From the autosave just before Saria gives Link the ocarina, 6 of 6 runs died within 158–241 s (a GPU hang, then macOS revoked GPU access). With `DM_NVN_ISOLATE_TASK=1` from [#1842](https://github.com/seedds/nxnative/pull/1842) (not on main), two runs of about 8 and 12 minutes (100,001 and 480,001 isolated draws of task `a5288d0a6c4971fd`) had no `ErrorHang`; that is two runs, not a controlled count. Over the same 3,600 frames of the cutscene at render scale 0.75, the switch took 4.5 ms (9 %) more GPU time a frame (one pair, played by hand); neither setting held 30 fps there (GPU p50 about 35–70 ms). First use of three `a5288d0a6c4971fd` pipelines took 7.0, 7.1 and 10.5 s to compile, each a freeze of that length, and two pipelines of task `a7f22b4a795ff4d1` failed with `XPC_ERROR_CONNECTION_INTERRUPTED`. ## Found on this game From 87c81a6dc48f5e1fb3a90761bd943a2a733a33ee Mon Sep 17 00:00:00 2001 From: Manan Khurana <86416632+Mkhurana3@users.noreply.github.com> Date: Mon, 9 Nov 2026 20:18:21 +0530 Subject: [PATCH 2/3] Ocarina of Time note: say the off/on cost is not clean, as the other ten windows match The A/B windows show the whole +4.5 ms (9 %) GPU gap coming from two slow ON windows (ending at frames 1500 and 1800, one with pipeline compiles and one without); in the other ten the GPU p50 differs by about 0.01 ms each. The bullet now says that instead of giving the gap as the cost of the extra cuts. --- dm/games/0400486024DEC000/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dm/games/0400486024DEC000/README.md b/dm/games/0400486024DEC000/README.md index 4ebe6005b0..ff4a0d77f3 100644 --- a/dm/games/0400486024DEC000/README.md +++ b/dm/games/0400486024DEC000/README.md @@ -106,7 +106,7 @@ From an empty save: at about 45 s `A` (title), `A` (New Game), then the movie. F - The "Navi" name label in the Great Deku Tree scene was missing on an earlier chain and has not been checked since. Not played yet: the Deku Tree dungeon, combat, later areas. Sound opens with no red line; movies and speech were not listened to. Physics issues reported in PR 1299 were not seen on this route (walking, the ladder, the camera). - Resolved: dotted horizontal rows across the picture in play, in fog and light (GitHub issue 1537): the agl_ndsr_vr networks' HMMA calls outside a chain overwrote an accumulator a chain held ([An HMMA outside a chain overwrote the chain's accumulator](../../../docs/reference/shaders.md#an-hmma-outside-a-chain-overwrote-the-chains-accumulator)), since chain 23a1ea1db092; found by bisecting chains on the Kokiri Forest autosave (0f510fd5's chain clean, bd9f422d's not). Under the game's upscaler they had also grown into white blotches: [LDSM reads rows other lanes stored](../../../docs/reference/shaders.md#ldsm-reads-rows-other-lanes-stored). - No iOS shader build or iPhone run has been done. Its memory use makes an iPhone unlikely as it stands. -- Open: GPU hang after the Saria cutscene on an M2 MacBook Air (16 GB, macOS 27.2, main `1794a35a`, 2026-11-09; [#1326](https://github.com/seedds/nxnative/issues/1326), [#1690](https://github.com/seedds/nxnative/issues/1690)). From the autosave just before Saria gives Link the ocarina, 6 of 6 runs died within 158–241 s (a GPU hang, then macOS revoked GPU access). With `DM_NVN_ISOLATE_TASK=1` from [#1842](https://github.com/seedds/nxnative/pull/1842) (not on main), two runs of about 8 and 12 minutes (100,001 and 480,001 isolated draws of task `a5288d0a6c4971fd`) had no `ErrorHang`; that is two runs, not a controlled count. Over the same 3,600 frames of the cutscene at render scale 0.75, the switch took 4.5 ms (9 %) more GPU time a frame (one pair, played by hand); neither setting held 30 fps there (GPU p50 about 35–70 ms). First use of three `a5288d0a6c4971fd` pipelines took 7.0, 7.1 and 10.5 s to compile, each a freeze of that length, and two pipelines of task `a7f22b4a795ff4d1` failed with `XPC_ERROR_CONNECTION_INTERRUPTED`. +- Open: GPU hang after the Saria cutscene on an M2 MacBook Air (16 GB, macOS 27.2, main `1794a35a`, 2026-11-09; [#1326](https://github.com/seedds/nxnative/issues/1326), [#1690](https://github.com/seedds/nxnative/issues/1690)). From the autosave just before Saria gives Link the ocarina, 6 of 6 runs died within 158–241 s (a GPU hang, then macOS revoked GPU access). With `DM_NVN_ISOLATE_TASK=1` from [#1842](https://github.com/seedds/nxnative/pull/1842) (not on main), two runs of about 8 and 12 minutes (100,001 and 480,001 isolated draws of task `a5288d0a6c4971fd`) had no `ErrorHang`; that is two runs, not a controlled count. Switch off against on over the same 3,600 frames of the cutscene at render scale 0.75 (one pair, played by hand): the GPU p50 per 300 frames averaged 47.7 ms off and 52.2 ms on, but the other ten windows differ by about 0.01 ms each and the whole gap sits in two slow ON windows (one with pipeline compiles, one without), so it is not a clean cost of the cuts; neither setting held 30 fps there (GPU p50 about 35–70 ms). First use of three `a5288d0a6c4971fd` pipelines took 7.0, 7.1 and 10.5 s to compile, each a freeze of that length, and two pipelines of task `a7f22b4a795ff4d1` failed with `XPC_ERROR_CONNECTION_INTERRUPTED`. ## Found on this game From 84e88361ea24c0c723668a3a5108bb5ff580df0c Mon Sep 17 00:00:00 2001 From: Manan Khurana <86416632+Mkhurana3@users.noreply.github.com> Date: Mon, 9 Nov 2026 21:26:42 +0530 Subject: [PATCH 3/3] README: drop the stale compiler-failure note, cite the in-game compile The a7f22b4a795ff4d1 XPC failures are a compiler crash already fixed on main (37f076a), per the maintainers on #1326, and are not part of the hang. The three a5288d0a6c4971fd first-use compiles are explained by the task and its meshes being absent from the prebuilt mesh lists. --- dm/games/0400486024DEC000/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dm/games/0400486024DEC000/README.md b/dm/games/0400486024DEC000/README.md index ff4a0d77f3..845d5bd3e6 100644 --- a/dm/games/0400486024DEC000/README.md +++ b/dm/games/0400486024DEC000/README.md @@ -106,7 +106,7 @@ From an empty save: at about 45 s `A` (title), `A` (New Game), then the movie. F - The "Navi" name label in the Great Deku Tree scene was missing on an earlier chain and has not been checked since. Not played yet: the Deku Tree dungeon, combat, later areas. Sound opens with no red line; movies and speech were not listened to. Physics issues reported in PR 1299 were not seen on this route (walking, the ladder, the camera). - Resolved: dotted horizontal rows across the picture in play, in fog and light (GitHub issue 1537): the agl_ndsr_vr networks' HMMA calls outside a chain overwrote an accumulator a chain held ([An HMMA outside a chain overwrote the chain's accumulator](../../../docs/reference/shaders.md#an-hmma-outside-a-chain-overwrote-the-chains-accumulator)), since chain 23a1ea1db092; found by bisecting chains on the Kokiri Forest autosave (0f510fd5's chain clean, bd9f422d's not). Under the game's upscaler they had also grown into white blotches: [LDSM reads rows other lanes stored](../../../docs/reference/shaders.md#ldsm-reads-rows-other-lanes-stored). - No iOS shader build or iPhone run has been done. Its memory use makes an iPhone unlikely as it stands. -- Open: GPU hang after the Saria cutscene on an M2 MacBook Air (16 GB, macOS 27.2, main `1794a35a`, 2026-11-09; [#1326](https://github.com/seedds/nxnative/issues/1326), [#1690](https://github.com/seedds/nxnative/issues/1690)). From the autosave just before Saria gives Link the ocarina, 6 of 6 runs died within 158–241 s (a GPU hang, then macOS revoked GPU access). With `DM_NVN_ISOLATE_TASK=1` from [#1842](https://github.com/seedds/nxnative/pull/1842) (not on main), two runs of about 8 and 12 minutes (100,001 and 480,001 isolated draws of task `a5288d0a6c4971fd`) had no `ErrorHang`; that is two runs, not a controlled count. Switch off against on over the same 3,600 frames of the cutscene at render scale 0.75 (one pair, played by hand): the GPU p50 per 300 frames averaged 47.7 ms off and 52.2 ms on, but the other ten windows differ by about 0.01 ms each and the whole gap sits in two slow ON windows (one with pipeline compiles, one without), so it is not a clean cost of the cuts; neither setting held 30 fps there (GPU p50 about 35–70 ms). First use of three `a5288d0a6c4971fd` pipelines took 7.0, 7.1 and 10.5 s to compile, each a freeze of that length, and two pipelines of task `a7f22b4a795ff4d1` failed with `XPC_ERROR_CONNECTION_INTERRUPTED`. +- Open: GPU hang after the Saria cutscene on an M2 MacBook Air (16 GB, macOS 27.2, main `1794a35a`, 2026-11-09; [#1326](https://github.com/seedds/nxnative/issues/1326), [#1690](https://github.com/seedds/nxnative/issues/1690)). From the autosave just before Saria gives Link the ocarina, 6 of 6 runs died within 158–241 s (a GPU hang, then macOS revoked GPU access). With `DM_NVN_ISOLATE_TASK=1` from [#1842](https://github.com/seedds/nxnative/pull/1842) (not on main), two runs of about 8 and 12 minutes (100,001 and 480,001 isolated draws of task `a5288d0a6c4971fd`) had no `ErrorHang`; that is two runs, not a controlled count. Switch off against on over the same 3,600 frames of the cutscene at render scale 0.75 (one pair, played by hand): the GPU p50 per 300 frames averaged 47.7 ms off and 52.2 ms on, but the other ten windows differ by about 0.01 ms each and the whole gap sits in two slow ON windows (one with pipeline compiles, one without), so it is not a clean cost of the cuts; neither setting held 30 fps there (GPU p50 about 35–70 ms). First use of three `a5288d0a6c4971fd` pipelines took 7.0, 7.1 and 10.5 s to compile, each a freeze of that length; the maintainers say the task and its meshes are not in the prebuilt mesh lists, so they compile in-game (#1326). ## Found on this game