wouter/d4r - d4r - Stoopjes Git

46 Commits 4 Branches 7 Tags

3 Commits

Include renames

Author	SHA1	Message	Date
ayo	22026ede12	Port K kernel tuning and M gfx1101 flags Preset K, byte-identical to the previous output in all captured buffers and the harness image, measured on an RX 7700 XT: - LDS row padding so WMMA operand loads hit distinct banks - WGP mode for enc1/enc2/dec1/dec2, ds_bpermute half-wave exchange for enc0/dec0/dec3, more waves per window in enc4/dec4/dec5 - cheaper operand plumbing in common/wmma_layout.h (packed f16 conversions, no dynamic vector indexing), raw buffer weight loads, accumulator-init vectors prepared once - enc1 as a single-buffer recompute layer, dec1 Q/K/V chains split - dec5 as four dependent phase kernels	2026-10-05 14:48:37 -04:00
ayo	f66cccaf28	Add experimental preset L and fix DLSS resource lifetimes and waits - Build L texture variants and complete older release kernel bundles. - Add selective Swin encoder fallback to reduce L/M motion trails. - Add opt-in GPU motion-vector dilation for thin geometry and sky gaps. - Replace GPU input waits with cancellable readiness polling and use sleeping event queries for completion. - Retain recorded resources through command allocator completion, defer shutdown until cleanup, and preserve unfinished output synchronization. - Extend lifecycle harness checks and add event, model and motion tests. - Update configuration defaults, build scripts and release documentation.	2026-10-02 00:30:46 -04:00
ayo	b645ef9096	Add DLSS accuracy mode and GLIBC 2.41 release builds - Add default-off PreferAccuracy for all native kernel families, including RDNA4 FP8 variants. - Restore intermediate FP8 quantization, per-MMA FP16 rounding, denormal handling and NGX synchronization in accuracy mode. - Select separately marked accuracy kernels and use a separate JIT cache; reject mixed sets and never silently substitute fast kernels. - Add a pinned GLIBC 2.41 build environment for ZLUDA/LLVM, the DLLs and both kernel sets, with ABI checks and toolchain provenance. - Support parallel target builds and pre-extracted PTX; correct ROCm header paths and Clang-CL input handling. - Bump to 0.1.3 and update release notes, documentation and tests. Validation: 11 tests passed; all 912 kernels built; E/K/M GPU harness runs completed with accuracy off/on. Packaged libraries loaded under GLIBC 2.41 and require no GLIBC version newer than 2.38. RTX image-quality parity remains unverified. Fixes #6 Refs #6	2026-10-01 01:35:39 -04:00

Author

SHA1

Message

Date

ayo

22026ede12

Port K kernel tuning and M gfx1101 flags

Preset K, byte-identical to the previous output in all captured buffers
and the harness image, measured on an RX 7700 XT:

- LDS row padding so WMMA operand loads hit distinct banks

- WGP mode for enc1/enc2/dec1/dec2, ds_bpermute half-wave exchange for
  enc0/dec0/dec3, more waves per window in enc4/dec4/dec5

- cheaper operand plumbing in common/wmma_layout.h (packed f16
  conversions, no dynamic vector indexing), raw buffer weight loads,
  accumulator-init vectors prepared once

- enc1 as a single-buffer recompute layer, dec1 Q/K/V chains split

- dec5 as four dependent phase kernels

2026-10-05 14:48:37 -04:00

ayo

f66cccaf28

Add experimental preset L and fix DLSS resource lifetimes and waits

- Build L texture variants and complete older release kernel bundles.
- Add selective Swin encoder fallback to reduce L/M motion trails.
- Add opt-in GPU motion-vector dilation for thin geometry and sky gaps.
- Replace GPU input waits with cancellable readiness polling and use
  sleeping event queries for completion.
- Retain recorded resources through command allocator completion, defer
  shutdown until cleanup, and preserve unfinished output synchronization.
- Extend lifecycle harness checks and add event, model and motion tests.
- Update configuration defaults, build scripts and release documentation.

2026-10-02 00:30:46 -04:00

ayo

b645ef9096

Add DLSS accuracy mode and GLIBC 2.41 release builds

- Add default-off PreferAccuracy for all native kernel families,
  including RDNA4 FP8 variants.

- Restore intermediate FP8 quantization, per-MMA FP16 rounding,
  denormal handling and NGX synchronization in accuracy mode.

- Select separately marked accuracy kernels and use a separate JIT
  cache; reject mixed sets and never silently substitute fast kernels.

- Add a pinned GLIBC 2.41 build environment for ZLUDA/LLVM, the DLLs
  and both kernel sets, with ABI checks and toolchain provenance.

- Support parallel target builds and pre-extracted PTX; correct
  ROCm header paths and Clang-CL input handling.

- Bump to 0.1.3 and update release notes, documentation and tests.

Validation: 11 tests passed; all 912 kernels built; E/K/M GPU harness
runs completed with accuracy off/on. Packaged libraries loaded under
GLIBC 2.41 and require no GLIBC version newer than 2.38.

RTX image-quality parity remains unverified.

Fixes #6
Refs #6

2026-10-01 01:35:39 -04:00