49 Commits
Author SHA1 Message Date
ayo 7c5d3c9157 feat: add d4r inference engines and Windows parity tooling
Add native K/M/RR inference engines, HIP kernels, replay and validation
harnesses, and Windows DLSS setup and runtime support. Include GPU
profiling tools, build and packaging scripts, platform integration
patches, and implementation documentation.
2026-10-16 08:54:47 -04:00
ayo 73994d4634 Clear windows branch; preserve previous contents in win-old 2026-10-16 05:20:08 -04:00
xdfnx-dev df415ec828 Support distributor DLL packages and validate Windows player mode 2026-10-03 11:28:21 +03:00
xdfnx-dev 6da1e79f9b Record Cyberpunk K and M 4K gameplay validation 2026-10-03 00:16:18 +03:00
xdfnx-dev bea15d3754 Record issue 10 regression results and use the Windows fixture host 2026-10-02 23:36:02 +03:00
xdfnx-dev 17121ccbf6 Fix Windows numeric NGX feature discovery and quick-test launches 2026-10-02 23:36:02 +03:00
xdfnx-dev 21ed0a8c88 Add automatic native Windows GPU quick-test packaging 2026-10-02 18:19:15 +03:00
xdfnx-dev 3d5f42865f Backport gfx1154 support into ZLUDA's Windows LLVM compiler
Adapt official LLVM 7a0829e41228 to the pinned LLVM 22 parser and ELF structures. Apply and audit the complete source delta before building, export its provenance, and package the source patch. Add 33 offline public PTX compilation checks across all eleven GPU targets without hardware execution. Preserve strict arithmetic and document RX9070XT K/M regressions.
2026-10-02 17:45:41 +03:00
xdfnx-devandrealdody 3fa963fd1b Enable experimental native Windows RDNA3 and RDNA4 targets
Share explicit target metadata across build/runtime/package guards, preserve native gfx11 WMMA with a Windows LDS lane exchange, select HIP devices by D3D12 LUID, and reject wrong-target/vkd3d installations before changing game files. Review all issue #10 logs and document compile-only coverage plus RX9070XT regressions.

Co-authored-by: realdody <[email protected]>
2026-10-02 17:10:42 +03:00
xdfnx-devandrealdody 98de8e3a3e Build native Windows packages for both RDNA4 targets
Add explicit gfx1200/gfx1201 builds, runtime selection and per-target
code objects. Guard staging, packaging and installation against mixed
architectures; retain gfx1201 compatibility and strict K/M arithmetic.

Adapt realdody's target configuration and MSVC binary16/BOOL portability
fixes from ae530b0. Do not import its forced gfx12 layout on RDNA3.
Add exhaustive binary16 and C++/PowerShell target contract regressions.

Validation: 20 CTest gates on RX 9070 XT; four software-only gfx1200 gates;
MSVC portability compile checks; exact 4K K async and M control outputs;
isolated package install/restore and mixed-target rejection tests.
No physical gfx1200 validation is claimed.

Co-authored-by: realdody <[email protected]>
2026-10-02 16:18:18 +03:00
xdfnx-dev 8f7058d2ba Add exact input-kernel experiments and preserve translated controls 2026-10-02 15:34:22 +03:00
xdfnx-dev 4f008b3ed1 Record bounded K game results for shared V and private WGP outputs 2026-10-02 15:21:34 +03:00
xdfnx-dev 96650ed29d Add private WGP output validation and record measured K experiments 2026-10-02 15:03:32 +03:00
xdfnx-dev bf4923aa4c Reproduce gfx12 FP16 accumulator differences and preserve K baseline 2026-10-02 13:03:19 +03:00
xdfnx-dev 9dece89599 Record controlled K timings and add read-only AMD GPU telemetry 2026-10-02 12:39:14 +03:00
xdfnx-dev 973eb5ecd2 Batch Windows DLSS GPU inputs and add deferred kernel diagnostics 2026-10-02 12:06:15 +03:00
Codex 77a5c008f9 Record upstream Windows integration draft PR 2026-10-02 10:53:12 +03:00
Codex 7eaeb8c3db Capture PresentMon metrics and record Windows release validation 2026-10-02 10:44:09 +03:00
Codex 541c9838f7 Merge upstream lifecycle fixes into native Windows RDNA4 port 2026-10-02 10:11:10 +03:00
Codex caecad0897 Measure Windows interop GPU boundaries without added completion waits 2026-10-02 10:08:47 +03:00
Codex 0b471be024 Queue native Windows DLSS asynchronously and drain feature consumers 2026-10-02 10:00:26 +03:00
Codex df44d8aeae Record final K production and package validation results 2026-10-01 21:23:44 +03:00
Codex 0c18df6253 Validate Windows native M FP8 and reduce operand packing cost 2026-10-01 21:16:28 +03:00
Codex 74aad12938 Cache D3D12 recording lookups without retaining game resources 2026-10-01 20:23:08 +03:00
Codex 73ee2e5ecc Measure Windows submission costs and validate faster private K output stores 2026-10-01 19:54:41 +03:00
Codex 40d9a55e10 Remove game log backpressure and measure Windows K/M kernels 2026-10-01 18:37:14 +03:00
Codex 74395b6aef Validate native Windows DLSS output on GPU and record actual M gameplay 2026-10-01 17:54:24 +03:00
Codex 5d33bddcd0 Fix game startup resource lifetime and D32S8 depth-plane interop 2026-10-01 17:27:06 +03:00
Codex 19d859e43d Capture pre-NGX command signatures through early OptiScaler device discovery 2026-10-01 16:48:12 +03:00
Codex e4c1c87b66 Resolve texture states in queue order across DLSS command lists 2026-10-01 16:41:52 +03:00
Codex e14b4cb54c Run native Windows DLSS K and M through an explicit OptiScaler source backend 2026-10-01 16:13:07 +03:00
Codex 27d6a1230a Forward Windows NGX project identity and validate frontend native network use 2026-10-01 15:45:55 +03:00
Codex 6ddf740877 Preserve initial PSO and enhanced texture layouts in Windows interop 2026-10-01 15:04:58 +03:00
Codex e9eec6d4e0 Convert native Windows D3D12 formats on gfx1201 with GPU validation 2026-10-01 14:48:42 +03:00
Codex 312c6bb50d Track Windows D3D12 indirect state and newer command interfaces 2026-10-01 14:15:01 +03:00
Codex 63607298ea Execute Windows DLSS K and M at recorded D3D12 command-list boundaries 2026-09-30 22:21:06 +03:00
Codex b8d801ae4c Correct M attention precision and validate every temporal transformer block 2026-09-30 21:30:05 +03:00
Codex e15f6a73a6 Run native Windows D3D12 DLSS K and M through shared VRAM 2026-09-30 20:22:03 +03:00
Codex 2db93adb1a Validate native Windows M against real PTX and correct patch-merge bounds 2026-09-30 19:01:18 +03:00
Codex 5592b7a921 Execute native Windows K DLSS and validate real transformer weights 2026-09-30 18:23:01 +03:00
Codex a0c36a6d6f Validate all eleven native gfx1201 K layers with Windows replay 2026-09-30 16:55:41 +03:00
Codex 3c28a82e34 Build patched Windows ZLUDA and execute native NGX DLSS diagnostics 2026-09-30 16:35:30 +03:00
Codex 64472e5a61 Merge upstream RDNA4 layouts and validate native Windows K integration 2026-09-30 14:54:38 +03:00
Codex 2dea066d6a Execute gfx1201 K enc2 and generalize layer reference diagnostics 2026-09-30 14:40:49 +03:00
Codex 1b5d4c9db7 Execute gfx1201 K enc1 and validate against numpy reference 2026-09-29 20:04:49 +03:00
Codex 6760294f60 Validate gfx1201 WMMA layout and compile first K kernel 2026-09-29 19:50:50 +03:00
Codex 64669c2c36 Initialize Windows NGX CUDA with HIP-backed NVAPI topology 2026-09-29 19:32:58 +03:00
Codex 239a644270 Validate native D3D12 HIP interop and isolate Windows mapped-view leak 2026-09-29 19:13:20 +03:00
Codex e95e061e81 Add native Windows gfx1201 HIP and ZLUDA diagnostics 2026-09-29 18:50:09 +03:00