wouter/d4r
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Commit Graph

  • 3f66081add engine: presets E/F AutoExposure, the game's exposure fields and NGX's jitter order linux woutermans 2026-10-19 02:31:58 +02:00
  • 916b0d1467 release: engine models skip L/M on RDNA2 and render in parallel woutermans 2026-10-18 23:03:58 +02:00
  • 4064117244 zluda: skip compiling modules whose kernels are all native (0014) woutermans 2026-10-18 23:03:58 +02:00
  • e6e3d18ea2 release: install.sh covers the engine's CNN models e and f woutermans 2026-10-18 21:27:26 +02:00
  • b9254f7f96 engine: Super Resolution presets E and F, the DLSS 3 CNN, on RDNA2 woutermans 2026-10-18 20:38:06 +02:00
  • 7da2f9ac39 Fix Windows NVAPI loading the gfx1101 build on every other AMD GPU v0.1.5a windows ayo 2026-10-18 12:19:37 -04:00
  • c29076b38f engine/rr: run dec0_1's translated tail on two waves pr37 itchyfeetleech 2026-10-18 22:03:34 +13:00
  • 69a4f6ae2f engine/rr: cheaper bounds checks and conversion guards in the translated passes itchyfeetleech 2026-10-18 21:47:38 +13:00
  • e46c589c68 engine/rr: hardware FP8 saturation for the f32 conversions itchyfeetleech 2026-10-18 21:19:00 +13:00
  • 2fa4561bc1 engine/rr: lighter binary16 conversion guards in the hybrid stages itchyfeetleech 2026-10-18 20:46:26 +13:00
  • 49439afb81 engine/rr: preset F in less shared memory (encoder 0 in one shader, enc1, dec0_1) itchyfeetleech 2026-10-18 20:20:07 +13:00
  • 0e851b7f90 release: prepare 0.1.5 with RDNA2 kernels and in-game DLSS/FSR switching ayo 2026-10-17 16:05:09 -04:00
  • df9f2cd012 files again v0.1.5 ayo 2026-10-17 07:56:00 -04:00
  • 78a3880c44 file ayo 2026-10-17 07:55:42 -04:00
  • ac4ee6c225 Expand Windows GPU support, add native FSR RR, and improve SR interop ayo 2026-10-17 07:40:34 -04:00
  • a617c818de TL;DR: Add an experimental FSR Ray Regeneration backend on Windows, run the K network natively on HIP with Vulkan->HIP shared memory, extend K kernels to gfx11 (RDNA3), and add dependency/packaging/test tooling around them. ayo 2026-10-17 01:43:06 -04:00
  • 247fe1092d Advance native Windows RR integration, kernel tooling and local packaging ayo 2026-10-16 12:53:26 -04:00
  • 3345fffcd3 readme ayo 2026-10-16 08:58:54 -04:00
  • 7c5d3c9157 feat: add d4r inference engines and Windows parity tooling ayo 2026-10-16 08:54:47 -04:00
  • 25493b3198 feat: add d4r inference engines and Windows parity tooling ayo 2026-10-16 08:54:39 -04:00
  • 73994d4634 Clear windows branch; preserve previous contents in win-old ayo 2026-10-16 05:20:08 -04:00
  • 34e6b0fe71 rr: add FSR Ray Regeneration, exact RDNA3 inference and motion experiments ayo 2026-10-16 04:55:25 -04:00
  • 17c1828f97 engine/rr, vulkan: honour PTX alignment for 32-bit and shared vector accesses woutermans 2026-10-15 02:49:16 +02:00
  • cf96e58bd9 engine/rr: bind the game colour and the exposure values for preset F's downp stage woutermans 2026-10-14 21:11:42 +02:00
  • d9e4627e60 release: install.sh accepts any y/n answer, and a non-terminal install builds the engine models woutermans 2026-10-14 20:50:35 +02:00
  • 747ac8538d release: install.sh asks what to install, caches the NVIDIA DLLs and the engine models, and [Engine] defaults to auto woutermans 2026-10-14 20:25:06 +02:00
  • d1559ca5fa release: install.sh downloads DLSS RR 310.9.1 and builds the engine's models woutermans 2026-10-14 20:09:46 +02:00
  • 11a3511120 engine: Super Resolution models from a one-frame weight dump woutermans 2026-10-14 20:09:22 +02:00
  • 5f520a186d engine/rr: build the Ray Reconstruction models from nvngx_dlssd.dll alone, without a capture woutermans 2026-10-14 20:09:08 +02:00
  • e3606cedf9 shim: defer CUDA feature creation for every engine feature and start the RR model at CreateFeature, waiting for it instead of compiling a CUDA fallback woutermans 2026-10-14 19:49:14 +02:00
  • 51afe5ea39 engine: fix preset F frame index at byte 1032 to f & 7u woutermans 2026-10-14 19:42:39 +02:00
  • af450a92b9 engine: cut the models' and the engines' memory woutermans 2026-10-14 15:14:28 +02:00
  • 162bbd77c8 shim: adaptive marker polling, opt-in GPU-side input wait/completion and RGBA8 output ayo 2026-10-13 22:30:41 -04:00
  • f0b55a850a engine: this GPU's own arithmetic and shared weight tiles as the builders' defaults woutermans 2026-10-14 00:52:41 +02:00
  • 8eb7a94533 engine/rr: native preset F (DLSS 4.5 Ray Reconstruction) and the reference's fused roundings woutermans 2026-10-13 13:49:46 +02:00
  • 065fe2c2a1 Engine: drop the layer-loop memory flush in preset M woutermans 2026-10-12 23:19:39 +02:00
  • d5c8be25c8 ASI add-on for existing OptiScaler installs ayo 2026-10-12 17:36:01 -04:00
  • 85175e0b27 Merge pull request #17 from woutermans/rr/cuda-ray-reconstruction Woutermans 2026-10-12 19:51:31 +02:00
  • db24805160 rr: Cyberpunk's extra inputs (SSS guide, colour before particles) and hardware depth woutermans 2026-10-12 13:06:59 +02:00
  • 50fa65d87b engine/rr: native Ray Reconstruction stages with capture/replay and shim integration woutermans 2026-10-12 12:36:35 +02:00
  • 30472ae195 Fix the rebase onto the native engine: Vulkan frontend staging, RR gating, lazy RR debug stages woutermans 2026-10-11 00:06:49 +02:00
  • 417abc64b5 zluda: renumber the RR patches to 0009-0013 woutermans 2026-10-08 00:44:03 +02:00
  • b9d61015dc packaging: version the RR work 0.1.4 woutermans 2026-10-05 13:56:50 +02:00
  • 0fe3a28cb7 kernels/rr: staged-image Swin layer engine prototypes woutermans 2026-10-05 13:06:06 +02:00
  • 798e25611c zluda: coalesce pair shadow stores, signed native FP8 conversion woutermans 2026-10-05 13:06:00 +02:00
  • 406c0286c0 zluda: gfx12 native FP8 requantization and packed operand gathers woutermans 2026-10-05 13:05:42 +02:00
  • 2e81a6bbfa docs: the denoiser layers' measured launch shape and parameter sizes woutermans 2026-10-04 12:47:24 +02:00
  • f87df1cbc5 kernels/rr: recover a denoiser layer's phase table from its PTX woutermans 2026-10-04 12:44:33 +02:00
  • 4d45dda4a8 docs: the denoiser's Swin block is the L/M template's block woutermans 2026-10-04 12:40:56 +02:00
  • 1a9762bb0e bridge: let the launch-stats dump take a kernel filter woutermans 2026-10-04 12:19:11 +02:00
  • 708b158332 kernels: compare two per-launch dumps byte for byte woutermans 2026-10-04 12:15:04 +02:00
  • dbef6f794e docs: pair the output-kernel numbers from the same run woutermans 2026-10-04 03:37:55 +02:00
  • 0dc2397e2b shim/harness: capture and replay a frame's Ray Reconstruction guides exactly woutermans 2026-10-04 03:32:15 +02:00
  • 74ce9d8f9a kernels: build the denoiser's output kernel natively, authorized separately woutermans 2026-10-04 03:32:09 +02:00
  • 56e1b76622 Restore behavioral render preset override coverage woutermans 2026-10-04 00:13:23 +02:00
  • bd343a7be1 Drop brittle shim source-text checks after rebase woutermans 2026-10-03 23:37:20 +02:00
  • 5461c161f3 Keep RR guides and separate alpha in VRAM woutermans 2026-10-03 23:21:46 +02:00
  • 7117c38674 Add CUDA NGX Ray Reconstruction support woutermans 2026-10-03 18:14:45 +02:00
  • 503061d568 x4r ayo 2026-10-12 05:37:36 -04:00
  • 37d1055228 Engine: support dynamic input resolution for presets K, L and M woutermans 2026-10-11 20:45:26 +02:00
  • 653460f7b2 Engine: native preset L (rrlite unfolded input, expansion and reconstruction stages) woutermans 2026-10-10 23:58:18 +02:00
  • de37ca5869 Merge pull request #33 from woutermans/engine-native-pr Woutermans 2026-10-10 21:27:41 +02:00
  • 2088422682 Add redistribution guard woutermans 2026-10-10 20:16:33 +02:00
  • e0d1549a21 Engine: native presets K and M in the D3D12 shim, wider game coverage woutermans 2026-10-10 20:16:33 +02:00
  • 16cfc1f063 mt ayo 2026-10-09 04:32:45 -04:00
  • 694a7a0126 Add experimental native Vulkan DLSS frontend ayo 2026-10-09 02:57:23 -04:00
  • f0c881985b Change d4r license to GPL-3.0-only win-old ayo 2026-10-08 17:17:44 -04:00
  • 18c694eecb Change d4r license to GPL-3.0-only ayo 2026-10-08 17:17:44 -04:00
  • fb244c1ef9 i love gitignore ayo 2026-10-08 16:04:26 -04:00
  • 4d5952bdb4 no idea how that got there 👀 ayo 2026-10-08 16:04:08 -04:00
  • 12782b3ff6 Merge pull request #31 from lunar-me/fix-docs/issue-012 ayo 2026-10-08 20:26:31 +01:00
  • c276548dab Merge pull request #32 from lunar-me/fix-docs/issue-013 ayo 2026-10-08 20:26:24 +01:00
  • e86131d396 Merge pull request #30 from lunar-me/fix-docs/issue-011 ayo 2026-10-08 20:26:10 +01:00
  • ce7d0ec34a docs: fix documentation (grammar-minor-003) pi 2026-10-09 05:11:57 +13:00
  • 77886fd07a docs: fix documentation (grammar-minor-002) pi 2026-10-09 05:11:48 +13:00
  • 9060d3c076 docs: fix documentation (grammar-minor-001) pi 2026-10-09 05:11:38 +13:00
  • 934fb486c7 Clamp the patch merge to ±2π in the native Swin encoders woutermans 2026-10-08 02:21:41 +02:00
  • 62738a01cc Merge pull request #29 from woutermans/sr-native-layout Woutermans 2026-10-08 02:08:06 +02:00
  • 8813ea723c kernels/tex: native token-lane parts for preset L's enc0 and dec0 woutermans 2026-10-08 01:08:34 +02:00
  • 0a12b6ce16 Add NVIDIA DLL downloader and bundle install.sh in release ZIPs ayo 2026-10-07 16:43:27 -04:00
  • 6ec70b3ee9 Preserve upstream accuracy-package adjustment during PR integration woutermans 2026-10-07 22:39:37 +02:00
  • 69b284b8b7 Fix PR integration: token prep slots, input ordering and output lifetime woutermans 2026-10-07 22:39:37 +02:00
  • e3d32a62a3 small package change ayo 2026-10-07 16:33:29 -04:00
  • 84e6048066 Merge PR #27: experimental packed and in-place DLSS output woutermans 2026-10-07 22:26:30 +02:00
  • 607035aa98 Merge PR #26: RDNA4 token-lane Swin and hardware FP8 texture tails woutermans 2026-10-07 22:26:30 +02:00
  • e5a8a4afb1 Merge pull request #28 from Tk-Glitch/R16_FLOAT_depth Woutermans 2026-10-07 22:13:38 +02:00
  • a627439025 Add MicroCUDA backend, adaptive output wait, wider native K coverage (improve perf) ayo 2026-10-07 15:54:38 -04:00
  • 40e0094c19 Add R16_FLOAT depth input support Tk-Glitch 2026-10-07 13:12:28 +02:00
  • 283738d626 Document the linear-texture input investigation woutermans 2026-10-07 10:59:56 +02:00
  • 87365077e0 Add experimental packed and in-place DLSS output woutermans 2026-10-07 10:59:53 +02:00
  • 1038825225 Hardware e4m3 decode in dec0's tail on the native-FP8 build woutermans 2026-10-07 01:15:36 +02:00
  • 5198fd38b3 Hardware e4m3 in the texture tails: enc0 tail and dec0 head woutermans 2026-10-06 23:25:14 +02:00
  • 3490023668 Token lanes for enc2 and dec2 too woutermans 2026-10-06 23:00:03 +02:00
  • a7d565d1cd Token-lane block for the RDNA4 native-FP8 layers: enc1, dec1 and the enc3 tube woutermans 2026-10-06 21:58:28 +02:00
  • 7c60d53cd0 Merge pull request #25 from woutermans/rdna4-fp8-byte-activations Woutermans 2026-10-06 19:27:30 +02:00
  • 8537c3132f Merge pull request #23 from dhaern/fix/launcher-config-and-packaging Woutermans 2026-10-06 19:20:35 +02:00
  • 17012ea588 Merge pull request #21 from dhaern/fix/native-kernel-hardening Woutermans 2026-10-06 19:20:31 +02:00
  • 9f96c3e57e Merge pull request #20 from dhaern/fix/ngx-feature-cleanup Woutermans 2026-10-06 19:20:27 +02:00
  • 1db8df252f Merge pull request #22 from dhaern/fix/shim-log-path-and-ranges Woutermans 2026-10-06 19:20:23 +02:00
  • 3381127259 Merge pull request #24 from Tk-Glitch/R8G8B8A8_UNORM_motion Woutermans 2026-10-06 19:20:20 +02:00