fix(scheduler): return midpoint for zero-range remap instead of silent to_min#1
Open
martial wants to merge 1 commit into
Open
fix(scheduler): return midpoint for zero-range remap instead of silent to_min#1martial wants to merge 1 commit into
martial wants to merge 1 commit into
Conversation
…t to_min When from_min == from_max, apply_remap() was silently returning to_min for any input value. This caused wires with a collapsed input range to always output the minimum value with no warning, making the issue very hard to diagnose in a live graph. Changed the degenerate-range fallback from t=0.0f to t=0.5f (output midpoint), which is a more neutral default and makes the failure mode visible in the UI rather than silently biasing outputs low. A future follow-up could add a stderr warning or a graph-load diagnostic for wires where from_min == from_max.
jeffcrouse
added a commit
that referenced
this pull request
Jul 1, 2026
Visuals params were global per op-type — every Plasma read the same warp/hue/ density/glow, ports drew only on first_node_of(op). Now each node owns its params. - VisualNode gains a stable `id` (monotonic, persisted) + `params[4]`; mapping dests move from "uniform.<name>" to "node:<id>.<name>". render() drops the global pu/decay/radius args and reads each node's params (Feedback decay scaled 0.82..0.98, Blur radius). - NodeGraph: param ports render on EVERY node; param_port/nearest_param are per-(node,local); connect/disconnect use node_param_dest. New apply_params() resolves each node's params from the registry and republishes the canonical six viz.<name> return-path sources from the first node of each op-type (so P27's static map-source catalog is unchanged). fill_uniforms removed. The out-of-box master.level->glow seed moved to set_visual_graph (needs the default ids). - Persistence: chain saves/loads node `id` so per-node mappings reattach. Old files degrade gracefully (default id=index; stale uniform.* mappings go unread). - main: graph.apply_params() replaces fill_uniforms + the viz publish + pu/decay/ radius; mapping overview decodes "node:<id>.<name>". Builds clean; launches with no validation errors. Two Plasma nodes can now be driven by different sources independently. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
jeffcrouse
added a commit
that referenced
this pull request
Jul 1, 2026
First platform step of P1 (ADR-0011): lift vivid-classic's operator ABI onto our trunk so we can build an operator-based visuals model. Placed at app/src/operator_api/ so classic's `#include "operator_api/..."` resolves verbatim. - Lifted headers (in-process GPU-op subset): types.h, operator.h, value_model.h, value_view.h, gpu_operator.h, gpu_types.h, type_id.h, metronome_sync.h, gpu_common.h. Verified they compile against our pinned wgpu-native (the #1 risk — no WGPUStringView/vivid_sv incompatibility; ABI v10). - Lifted operator_descriptor_validation.{h,cpp} (header+stdlib only → headless); compiled into both vivid_poc and the test lib. - tests/test_descriptor_validation.cpp (HEADLESS, 17 checks): well-formed → clean; null/missing-name/missing-capability/null-params/param-missing-name/dup-param/ port-missing-name/dup-port (per-direction) all fire their stable named codes. Spike (P1.0) cleared its two scariest unknowns: the ABI compiles against our wgpu, and a PlasmaOp authored in our toolchain yields the right collect_params/ports metadata + GpuProcessable capability. Headless ctest 4/4; full app build green. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bug
In
apply_remap()(src/runtime/scheduler.cpp), whenfrom_min == from_max(zero input range), the fallback was:This silently forces
t = 0.0f, which means the output is alwaysto_minregardless of input. A wire with a collapsed input range will bias all downstream values to the minimum with no warning — very hard to diagnose in a live graph.Fix
Change the fallback to
t = 0.5f(output midpoint):This is a more neutral default and makes the failure mode visible rather than hiding it.
A follow-up could add a
stderrwarning or a graph-load diagnostic for wires wherefrom_min == from_max.— Michelle