From 2f60019cbd417cdbdaa6da0e90d8515fbccb9bbf Mon Sep 17 00:00:00 2001 From: Ed_ Date: Mon, 27 Jul 2026 01:53:21 -0400 Subject: [PATCH] docs(twitter): add 1868436190447432064 corpus (NOTimothyLottes Vulkan 'buffer zoo' problem) 2-post thread. The real 'buffer zoo' problem with Vulkan: need HW instruction emulation macros with overcomplete both {offset, index} inputs so emulation can choose the right path based on whatever {SSBO, TEXEL_BUFFER, future pointer}. TEXEL_BUFFER for formats one can't load from SSBOs, both need 'indexes', and whenever IHVs actually correctly optimize the pointer extension, one needs byte offsets. 'Deadcode removal nightmare land wins today.' Sibling to the 1870351985855119449 STORAGE_TEXEL_BUFFER aliasing thread. --- docs/twitter/1868436190447432064/thread.md | 23 +++++++++++ .../1868436190447432064/thread_data.json | 38 +++++++++++++++++++ 2 files changed, 61 insertions(+) create mode 100644 docs/twitter/1868436190447432064/thread.md create mode 100644 docs/twitter/1868436190447432064/thread_data.json diff --git a/docs/twitter/1868436190447432064/thread.md b/docs/twitter/1868436190447432064/thread.md new file mode 100644 index 00000000..67c0c99a --- /dev/null +++ b/docs/twitter/1868436190447432064/thread.md @@ -0,0 +1,23 @@ +--- +title: "TEXEL_BUFFER is needed for whatever formats one cannot load from SSBOs, and both" +author: "NOTimothyLottes" +handle: "@NOTimothyLottes" +post_url: "https://x.com/NOTimothyLottes/status/1868436190447432064" +post_id: "1868436190447432064" +timestamp: "2024-12-15 23:21:16" +post_count: 2 +reply_count: 0 +repost_count: 0 +like_count: 3 +view_count: 461 +--- + +# @NOTimothyLottes — TEXEL_BUFFER is needed for whatever formats one cannot load from SSBOs, and both + +## Post 1 (2024-12-15 23:18:42) + +The real "buffer zoo" problem with Vulkan is needing to make your HW instruction emulation macros with overcomplete both {offset, and index} inputs so the emulation can choose the correct one based on whatever crap path is required {SSBO, TEXEL_BUFFER, future pointer} + +## Post 2 (2024-12-15 23:21:16) — reply to Post 1 + +TEXEL_BUFFER is needed for whatever formats one cannot load from SSBOs, and both of those require 'indexes', and whenever the IHVs actually correctly optimize the pointer extension, one needs the byte offsets instead. So deadcode removal nightmare land wins today :( diff --git a/docs/twitter/1868436190447432064/thread_data.json b/docs/twitter/1868436190447432064/thread_data.json new file mode 100644 index 00000000..517c408c --- /dev/null +++ b/docs/twitter/1868436190447432064/thread_data.json @@ -0,0 +1,38 @@ +{ + "root_post_id": "1868436190447432064", + "posts": [ + { + "post_id": "1868435545736782009", + "author": "NOTimothyLottes", + "handle": "NOTimothyLottes", + "text": "The real \"buffer zoo\" problem with Vulkan is needing to make your HW instruction emulation macros with overcomplete both {offset, and index} inputs so the emulation can choose the correct one based on whatever crap path is required {SSBO, TEXEL_BUFFER, future pointer}", + "timestamp": "2024-12-15 23:18:42", + "media_urls": [], + "reply_to_id": null, + "quote_of_id": null, + "metrics": { + "reply_count": 1, + "repost_count": 1, + "like_count": 10, + "view_count": 1400 + } + }, + { + "post_id": "1868436190447432064", + "author": "NOTimothyLottes", + "handle": "NOTimothyLottes", + "text": "TEXEL_BUFFER is needed for whatever formats one cannot load from SSBOs, and both of those require 'indexes', and whenever the IHVs actually correctly optimize the pointer extension, one needs the byte offsets instead. So deadcode removal nightmare land wins today :(", + "timestamp": "2024-12-15 23:21:16", + "media_urls": [], + "reply_to_id": "1868435545736782009", + "quote_of_id": null, + "metrics": { + "reply_count": 0, + "repost_count": 0, + "like_count": 3, + "view_count": 461 + } + } + ], + "source_url": "https://x.com/NOTimothyLottes/status/1868436190447432064" +} \ No newline at end of file