- [Interop] MarkerPollUs = auto is the new default: the input-marker wait learns when
each feature's inputs become ready, sleeps in <=200 us chunks, yields near the
prediction and backs off when late (d4r_input_wait.h); a number keeps fixed pacing
(0 yields). d4r_config.py no longer exports 200 when the key is unset
- D4R_SHIM_GPU_INPUT_WAIT=1 (experimental, split VRAM, age 0): NGX's null stream
waits on a coherent Vulkan word imported into HIP (d4rStreamWaitValue32Eq) instead
of CPU polling. The original readback marker stays the authority; a CPU watchdog
releases discarded lists and timeouts. Denoiser, verify, capture and dilation
frames keep the CPU wait
- D4R_SHIM_GPU_COMPLETION=1 (needs GPU_INPUT_WAIT): the bridge writes a completion
word right after the native output kernel (d4rNgxWriteValue32), and the game's queue
polls it with a bounded compute wait (kernels/support/vulkan_wait.comp), replacing
the split list's CPU timeline signal. frame_retired writes the same word on every
failure path
- D4R_SHIM_NATIVE_COMPLETION=1: a null-stream callback in the bridge signals the
split timeline (d4rNgxTimeline*); D4rFrameCompletion::can_signal only releases the
oldest outstanding frame
- RGBA8 output: tex_common.h adds an 'R4' redirect tag, so K's native output kernel
stores packed RGBA8 directly (D4R_SHIM_OUTPUT_RGBA8=1, checked against the kernel
manifest by d4rOutputKernelRgba8). Host packing handles RGBA8 too.
patches/vkd3d-proton/0004 (VKD3D_D4R_LINEAR_RGBA8=WxH) makes such outputs linear and
exportable so the result can be written in place
- D4R_SHIM_FUSE_RGBA8_INPUT=1: one Vulkan pass converts and copies RGBA8 color input
(vulkan_read.comp mode 5 matches the blit's toward-zero halves, plus a row stride)
- profiling: thread CPU time from the Unix bridge (Wine's GetThreadTimes ticks at
10 ms); game-queue D3D12 timestamps (vk_input/vk_dlss_wait/vk_output/vk_total);
gpu_eval split into input_gpu_wait and dlss_gpu; D4R_PROFILE_GPU_EVENTS,
_TIMESTAMPS, _LOG and _CPU_TIME=0 switch off individual parts for bisection
- bridge: D4R_NGX_STREAM_PRIORITY=high|normal runs NGX on a priority stream
(off by default; no gain measured)
- release_feature: a feature already marked retiring by a failed drain can still
be released normally
- harness: RGBA8 with motion and jitter scenes (re-quantized each frame),
D4R_HARNESS_RGBA8_RAMP, D4R_HARNESS_THRASH_MB to evict L2/Infinity Cache before
each evaluation
- kernels/native: in_hoist.py and split_output_paths.py (opt-in scheduling
experiments); scripts/build_k_ldr_output.sh builds K's LDR output with an
optional D4R_LDR_OUTPUT_MAX_VGPR cap
- tests: test_input_wait.py and test_latency_optimizations.py; MarkerPollUs and
frame-completion tests updated
Ray Reconstruction's eleven DLSS-D Swin layers still run ZLUDA's compile of
NVIDIA's PTX; at 1280x720 -> 3840x2160 on gfx1201 they are 30.2 of a 31.6 ms
frame. The denoiser's high-fidelity output kernel is the first one with a native
replacement: make_ptx.py replaces its whole-texel RGBA16F store
(sust.b.2d.v4.b16.zero) through d4r_sust_v4b16, kernels/build.sh builds the pair
as the `rr` family, and packaging builds it into every target folder including
the RDNA4 FP8 ones.
Both libraries define modules named cuda_dldn_engine_*, with different text, so
one manifest cannot take the union of their hashes: kernel_manifest.py now takes
the denoiser DLL as --rr-dll and lets only its hashes authorize a
cuda_dldn_engine_* binary, and package_release.sh passes it when the denoiser is
bundled.
Verified in the D3D12 harness with Ray Reconstruction at 1280x720 -> 3840x2160:
the output kernel drops from 1.44 to 1.00 ms/frame and the 66 MB raw output is
byte-identical to the run with every kernel translated.
docs/native-kernels.md gains the denoiser family, the per-kernel cost table and
the instruction counts that show why the Swin layers are the remaining work.
- Tag redirected surfaces as R10G10B10A2 ('R3') as well as RGBA16F ('R2'), and
let the native output kernel store the game's R10G10B10A2 texels itself.
- Import linear, exportable game textures (vkd3d-proton patch 0003) into HIP
so the output kernel writes the game's output texture with no game-side copy.
- Add ID3D12DXVKInteropDeviceD4R3 with GetVulkanLinearImageInfo.
- Gate both paths behind D4R_SHIM_OUTPUT_PACKED and D4R_SHIM_INPLACE, and use
them only for split frames of preset M whose output kernel honours the redirect.
- Convert on the host when a recorded packed frame finds the wrong kernel.
- Give the harness an R10G10B10A2 output, and add D4R_SHIM_SPIN_POLL so
frame timing is not quantised to the poll sleep.
- Document the measured removal of the output copies in performance.md.
make_ptx.py turns the 32 cvt.rn.f16x2.e4m3x2 of rrlite_dec0_4x4_folded into calls of d4r_dec_e4m3x2 (tex_common.h) when D4R_TEX_FP8=1: v_cvt_pk_f32_fp8 and two conversions to f16 instead of ZLUDA's integer lowering (about 25 instructions each; the kernel goes from 4058 to 3418 instructions). The conversion equals ZLUDA's e4m3_to_f16_bits for every code except the NaN codes 0x7f / 0xff, checked over all 65536 code pairs on an RX 9070 XT, and the head never stages those codes (it clamps to +-448 before encoding). Harness frames are byte-identical in the accuracy and the fast set, over a 16-frame moving sequence in the accuracy set. dec0 at 2227x1253 render: 0.43 -> 0.38 ms (accuracy), 0.36 -> 0.30 ms (fast). -DD4R_TEX_SOFT_E4M3 gives ZLUDA's lowering; builds without D4R_TEX_FP8 keep the PTX instruction.
On the gfx12 native-FP8 path the enc0 tail and the dec0 head encode their results with v_cvt_pk_fp8_f32 on the f32 copies of the f16 values (e4m3x4_f32 in tex_common.h) instead of ZLUDA's integer e4m3x2. The clamps keep the conversion in range: +-2pi for enc0 (the f16 bound hi[0], same in f32) and +-448 for dec0, both as a median of three, where a NaN gives the lower bound. The codes equal e4m3x2 for every non-NaN f16, checked over all 65536 patterns on an RX 9070 XT; a NaN in dec0 now encodes as -448 instead of the NaN code. Harness frames are byte-identical in both sets (LDR and HDR, motion vectors at render and display resolution). dec0: 0.21 -> 0.15 ms at 1280x720 render and 0.74 -> 0.53 ms at 2560x1440; enc0 changes by 1-2% (its time is ZLUDA's compile of the feature code). -DD4R_TEX_SOFT_E4M3 (D4R_TEX_CFLAGS) keeps e4m3x2.
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
- 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
RDNA4 (experimental; untested on RDNA4 hardware)
- kernels/common/wmma_layout.h picks the WMMA operand/accumulator layout
per target; K (pwin_*) and M (swin_*) index through it, so the same
sources build for gfx11 and gfx12. D4R_WMMA_LAYOUT=12 on gfx11 is a test
shim that runs the gfx12 indexing on gfx11 WMMA.
- M can use native e4m3 FP8 WMMA on gfx12 (D4R_NATIVE_FP8=1). Texture
tails (enc0_tail, dec0_head) get gfx12 f16 and FP8 paths that match
ZLUDA's gfx12 lowering.
- ZLUDA patch 0007: gfx12 WMMA helpers, optional native FP8 MMA
(D4R_ZLUDA_WMMA_FP8_NATIVE), a layout override for testing, and an
architecture argument for d4r_emit. Texture kernels are now compiled
offline per target through d4r_emit (D4R_ZLUDA_EMIT) instead of on the
local GPU.
- New d4r.ini option [Kernels] NativeFp8 (default on), wired through the
shim and d4r_config.py. On gfx12 the bridge serves kernels/<arch>-fp8
when it is on.
- The bridge picks the KFD GPU with the most SIMDs (so an iGPU is not
chosen over a discrete GPU); D4R_GPU_ARCH overrides it.
- kernels/build.sh and package_release.sh build gfx1100-gfx1103 and
gfx1200/gfx1201 (plus -fp8 variants), and D4R_BUNDLE_TEX takes
per-target folders.
Performance (RDNA3; harness byte-identical unless noted)
- ZLUDA patch 0006: m16n8k8 f16 MMAs on WMMA (E; not bit-exact, 52-55 dB
per layer, D4R_ZLUDA_WMMA_K8=0 disables it), per-key weight-image slots
for native prep kernels (d4r_prep_key_at, d4r_prep_key_slots), and
wave64 builds of MMA-free kernels (D4R_ZLUDA_WAVE64).
- M: packed FP8 encoding, conflict-free patch-merge reads, scalar weight
loads in enc1, and weight prep once per weight set for the six enc3
tube blocks.
- K: CU mode for the six non-wide pwin layers.
- Texture kernels: every DLSS flag variant is built natively (24 K output,
8 M post, 4 M downsample, 4 M enc0), not only Townfall's; post and K
output run as wave64.
- M direct output: rrlite_downsample_kernel writes the shim's output
buffer; the shim's default direct-output presets are now 11,13.
Tools
- dump_runner takes the prep grid from d4r_prep_blocks and refuses
texture/surface dumps.
- New pwin_check.py, swin_check.py, model_enc3.py and tail_check.hip.
Docs: README GPU table and RDNA4 notes, building (patches 0002-0007,
d4r_emit), native-kernels (RDNA4 section), performance (new numbers),
packaging README. Version 0.1.2.
ayo