Replace the per-channel Reinhard display transform with a parameterized
soft clip T_p(x) = tanh(x^p)^(1/p), default softclip p=2, exposed through
--tone-map and --tone-map-p. Keep --tone-map reinhard bit-compatible with
the previous x/(1+x) curve for existing images and reject combining it
with an explicit --tone-map-p.
Route the primary image and the mesh overlay through the same
ToneMapSettings; the linear HDR FITS writer stays pre-tone-map. Add a
focused optics-linked tone-map test target, CLI success/error coverage,
and document the display operator versus the Moffat effective PSF.
--draw-mesh no longer modifies the primary render. The clean HDR FITS and
tone-mapped image are written first, then the final image-plane mesh is
overlaid in place on the already-consumed HDR buffer and saved as a
<stem>_mesh sibling. This applies uniformly to single-frame, movie-frame,
and imported lens-map renders, with derived paths validated before catalog,
spacetime, and render initialization.
Dummy PSF backends still write nothing and report ignoring --draw-mesh. The
reference-image target now uses the configured image extension so the
ENABLE_PNG=0 test path passes, and the CLI tests cover clean-main identity,
sibling naming, PATH_MAX overflow, and zero-tolerance HDR comparison.
README, README.zh-CN, and usage.md describe the new naming; the mesh reference
asset is renamed to match.
Add the production-density 4K Galactic-center comparison to the 2026-09-25 benchmark record, including the per-image error decomposition, and refine the fast-mode positioning in usage.md, the design document, and README.
- Document that nearest/bilinear deposit accuracy is set by the deposit mode and N rather than output resolution, and that the still-frame error is dominated by a small number of resolved bright cores while the flux-dominant faint texture is much better than the global relative L2.
- Add the temporal-coherence caveat: nearest deposition jumps by 1/N output pixel per axis across supersampled-cell boundaries, so fast mode is not qualified as a temporally coherent movie path; bilinear removes the centroid jump but broadens the profile.
- Add scripts/fast_mode_error_decomposition.py, which ranks the largest |fast-base| RGB samples and partitions the error by a disjoint base-value range.
Replace the nested spatial global convolution in fast_psf_accumulator_resolve with a reusable double-precision FFTW linear convolution on the CPU PSF backend.
- Add the private src/fast_psf_fftw.{c,h} module: zero-padded R2C/C2R plans, cached kernel spectrum and planar scratch, exact 1/(Pwidth*Pheight) and 1/N^2 normalization, (R,R) crop, and additive HDR output.
- Keep the previous nested loops as fast_psf_accumulator_resolve_spatial_reference for tests/benchmarks only; it is not a runtime fallback.
- Cache the circular row spans on FastPsfAccumulator and report one-time plan, kernel transform, scratch, and per-frame stage timings.
- Require fftw3_omp for CPU builds; HIP and dummy builds do not link FFTW.
- Namespace test/helper binaries by spacetime and build tag, and reject make test / psf-capture for non-CPU backends.
- Add tests/test_fast_psf_fftw.c (FFTW versus spatial), tests/benchmark_fast_psf_fftw.c, an FFTW CLI smoke check, and the 2026-09-25 benchmark record.
Add an optional CPU preview path that deposits each point-source image as a
supersampled delta and resolves the whole frame with one global Moffat
convolution plus an N x N box average, instead of splatting a per-event PSF.
- optics: FastPsfAccumulator builds the pixel-area-integral kernel of the
target Moffat at the supersampled scale (width N*alpha, same beta). The
1/N^2 box average then reproduces the final pixel-area integral, so the
requested FWHM and beta are preserved without renormalisation. Deposits are
per-cell atomic adds; resolve accumulates into the caller's HDR buffer.
- frame: fast branch in frame_splat_catalog with one shared supersampled
buffer and a single resolve per frame; the accumulator is reused across
movie frames and built from the map dimensions on lens-map import.
- main: --fast-mode, --fast-supersample N (1..8, default 2) and
--fast-deposit nearest|bilinear (default nearest). CPU-only and rejected in
the HIP/dummy backends; --psf-min-y still applies per event while
--max-cache-psf-flux does not.
- The deposition scheme was chosen by scripts/fast_mode_deposit_error.py:
nearest keeps the PSF shape exactly with <= 0.5/N px position quantization;
bilinear keeps the exact centroid but broadens FWHM and beta. Recorded in
benchmarks/fast_mode_deposit_2026-09-18.md.
- tests/test_frame.c covers fast nearest vs the direct evaluator at the snapped
centre, flux conservation, bilinear centroid, min-Y discard, frame plumbing,
and HDR accumulation onto a non-zero background.
- benchmarks/fast_mode_cpu_2026-09-18.md records a ~10x speedup on the 2MASS
galactic-centre field with small tone-mapped differences.
Document the producer-private adaptive path, the dummy support workload
diagnostics, the negative result of the old serial binning and the
2026-09-13 full-day comparison (splat 282.944 s vs atomic 481.417 s).
Add the corresponding raw logs, hotspot fixtures and runner evidence,
and lay out the staged plan toward a sub-240 s full-day render without
changing the default atomic path.
Pin OMP_DYNAMIC=FALSE so thread counts stay deterministic, record it in
the command line, and locate the binary hash correctly when the command
is wrapped in env with inline assignments.
Add production, production-prepared, production-parallel and
production-busy modes plus an optional 1..512 round count. Repeated
rounds validate against a round-scaled CPU double reference and report
per-round throughput; production-parallel exercises concurrent
prepared-chunk submission and production-busy adds host spin workers to
probe contention.
Add an opt-in DUMMY_PSF_SUPPORT_STATS mode to the diagnostic-only dummy
consumer. It records per-chunk cache-limited 16px support references,
touched tiles, 256-reference tasks, maximum tile list and first-pass
counting time, then reports percentiles. No HDR is produced and no
references are materialized or uploaded.
Move adaptive tile16 selection and support-expanded CPU binning into a
per-producer HipPsfPreparedChunk computed before acquiring the shared
sink lock. The sink keeps sole ownership of the GPU HDR, stream, device
scratch and upload staging; submit_prepared runs under the lock and
preserves the existing single-stream ordering and direct fallback path.
The legacy hip_psf_sink_submit entry point prepares an internal chunk so
single-threaded replay stays compatible, and default accumulation remains
atomic.
Replace the in-tree Wyman analytic CIE fit with a repository CIE 1931
2-degree LUT derived from the 360-830 nm, 1 nm-linear CSV reference. The
GRBBLUT3 table stores 1024 uniform log(T) XYZ nodes over
[670.146556, 101408.88] K with four-point cubic Lagrange interpolation, and
the same CSV reference supplies a three-term Rayleigh-Jeans form above the
table and an inverse-temperature/log-XYZ crossover with channel-specific
endpoint forms below it. The loader validates the header and payload
checksum and rejects legacy formats and runtime generation.
Initialize the backend in the renderer, frame test, and PSF capture tool so
the new default is active wherever colors are produced.
Preserve the protected full-catalog dummy run, its initial harness failure, input and output hashes, distribution quantiles, and the resulting sparse-tile scheduling conclusions.
Reuse production catalog mapping and per-worker chunk boundaries to report triangle and spatial distributions without initializing HIP or writing image outputs. Include a protected full-catalog runner and build documentation.
Add a bounded per-chunk center-tile selector to the real-event replay prototype and retain the paired timing, mixed-boundary, CPU, HIP, and sandbox-failure evidence.
Archive fixed catalog subsets and captured events, manifests, reproducible commands, raw timing and regression logs, and device resource evidence. Document distribution-dependent tile reduction gains and independent indexed producer throughput.
Update the GPU plan and renderer design while retaining the production atomic path. Preserve historical measurement metadata and exclude generated Python bytecode.
Capture prepared events through a test-only producer consumer and measure serial and indexed parallel generation independently. Compare the production atomic kernel with bounded pixel-owned tile reductions while preserving double HDR, cache support, and direct fallback ordering.
Validate captured inputs, edge and multichunk fixtures, CPU/HIP renderer regressions, and 45 isolated final replays. Keep the production HIP algorithm unchanged because gains depend on event distribution.
Replace the artifact-affected Schwarzschild image with the corrected 4K render. Link both READMEs to the paired CPU/HIP benchmark and clarify that this run had no PSF wing clipping or direct fallbacks.
Preserve user-provided complete CPU and HIP logs and associate the optimized implementation with e1ec480669. Record matching event counts, user-confirmed bit-identical PNG output, and the 2.77x workflow speedup with tracing/import differences noted.
Mark the pre-artifact-fix CPU benchmark as historical and link the new paired result.
Cooperate across 32 lanes per PSF and use two completion-protected staging slots with complete batch timing. Restore coarse OpenMP event production while serializing shared GPU submissions and direct fallback boundaries.
Add bounded benchmarks, streaming and renderer regressions, and preserve validation evidence and ownership documentation.
Check spherical area weight sums against a fixed 1e-8 tolerance and normalize accepted weights before interpolation. Skip invalid images in both catalog paths.
Add a thin-triangle regression for both parities, document the measured tolerance margin, and refresh HDR fixtures with strict comparisons restored.
Schedule ray-pool advancement in dynamic chunks of 32 slots to distribute frame-grouped active rays across workers while preserving stable endpoint destinations.
Add 1/4/16-thread lens-map equivalence coverage and document schedule comparisons with complete commands, input data and original logs. The 393-frame 64x36 production-source benchmark improves from 62.13 to 38.69 seconds with byte-identical PNGs and lens maps. Release and Debug test suites pass.
Allow frames to finish refinement independently without aborting the movie. Report genuine refinement errors with generation and frame identifiers, and clarify the sample batch contract.
Add a two-frame Schwarzschild regression covering unequal refinement progress. Validated Release and Debug make test suites and 640x360 rendering of the final eight free-fall frames in both builds.
Integrate timelike geodesics and Fermi-Walker tetrads in ingoing Kerr-Schild coordinates, sampled at a configurable proper-time cadence.
Add movie-track-samples to preserve CSV events as frames. Document usage and singularity guards, and cover analytic orbits, transport convergence, sampling, and CSV rendering.
wyj