Skip to content

Lower direct 1D surface atomic add instructions - #842

Closed
monsterlag wants to merge 3 commits into
NotRexed:mainfrom
monsterlag:fix/native-suatom-direct
Closed

monsterlag wants to merge 3 commits into
NotRexed:mainfrom
monsterlag:fix/native-suatom-direct

Conversation

@monsterlag

@monsterlag monsterlag commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

Lower direct-descriptor native 0x593 in its 1D U32 ADD/IGN/Strong(GPU) form to the existing Maxwell surface atomic. Preserve the returned old value, descriptor bank/offset, predication and scalar aliases. Unsupported forms, descriptor bounds and bank conflicts still fail.

Strong(GPU) reuses the existing Device-scoped relaxed atomic; explicit MEMBARs remain independent. Encoding and memory-order references are documented with the fixture.

Updated on main b2765ba6 through a normal merge, preserving its allocator, depth, mesh and cache changes. At f92f2c41, Linux source checks, Windows source checks, SPIR-V checks, Windows renderer/builder and Linux AppImage all pass. Separately, all12 unchanged authored programs passed full REGMAP/Maxwell CFG translation, IR checks and Vulkan1.3 SPIR-V validation against the combined private 2.0.6 integration; the released lowerer refused the supported authored form.

These checks do not establish GPU numerical results or a particular visible game fix. No game-derived programs or caches are included.

monsterlag and others added 2 commits November 8, 2026 17:55
Preserve the old U32 image value, descriptor location and scalar register aliases for the supported native SUATOM form. Keep Strong(GPU) atomic coherence on the existing Device/Relaxed path and leave explicit MEMBAR instructions separate.

Exercise authored instructions through production lowering, Maxwell CFG translation and SPIR-V emission, including strict mode rejection and high-register mapping.

Assisted-by: OpenAI Codex

Co-Authored-By: OpenAI Codex <noreply@openai.com>
Normally merge b2765ba, preserving all upstream allocator, depth, mesh,
cache and workflow changes. Retain this topic's exact authored assertions.

Assisted-by: OpenAI Codex
Co-Authored-By: OpenAI Codex <noreply@openai.com>
NotRexed pushed a commit that referenced this pull request Nov 9, 2026
@NotRexed

NotRexed commented Nov 9, 2026 •

Copy link
Copy Markdown
Owner

This is in dev and main now (landed in c7e0058, "Merge PR #842"; first shipped in v2.0.8-rc.1), so closing. Thanks!

@NotRexed NotRexed closed this Nov 9, 2026
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.

2 participants