Expose selected decoder plugin and backend per decoded sample

Open
  • Is something wrong?
#64 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
38/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
cpp, python
Domain
backend-api-design

Research direction

Start by reading the generic decoder API around nvimgcodecDecoderDecode, nvimgcodecFutureGetProcessingStatus, and nvimgcodecExecutionParams_t, then compare the Python get_decoder_ids() API and the cuCIM workaround described in the issue. Define how per-sample plugin and backend provenance, fallback information, and results after completion should fit the C and Python APIs. Done means callers can query those details for successful and failed samples while preserving generic per-sample dispatch.

Written by the indexing model from the issue text.

Description

Request

Please expose the decoder plugin identifier and backend kind that actually handled each sample submitted through the generic decoder API.

Motivation

Applications that audit heterogeneous decode pipelines need to distinguish GPU-only, hybrid CPU/GPU, and CPU fallback execution without inferring it from the codec, output buffer, or installed extensions. The generic decoder can fall through registered implementations in priority order, and a batch can contain per-sample choices.

Current API boundary

With nvImageCodec 0.9, nvimgcodecDecoderDecode plus nvimgcodecFutureGetProcessingStatus reports only processing status. nvimgcodecExecutionParams_t::backends can constrain backend kinds, but the selected nvimgcodecDecoderDesc_t::id and backend_kind are not returned.

Current documentation also provides plugin enumeration/selection (get_decoder_ids() in Python and plugin selections/framework queries in the newer C API), which is useful for configuration but still does not identify the implementation selected after generic fallback.

Suggested shape

A future/result query returning, for each submitted sample:

  • selected decoder plugin ID;
  • selected backend kind;
  • whether fallback occurred (and optionally rejected candidates/statuses).

The query should remain valid after the decode completes and should work for both success and failure. An equivalent Python result/property would be useful.

Workaround and limitation

cuCIM can create separate generic decoders restricted to one backend kind and probe/decode in an explicit order. That provides truthful backend provenance but changes selection to a caller policy, cannot identify the plugin within that backend, and cannot preserve arbitrary per-sample mixed-backend dispatch in a single batch.

Dominant language
Jupyter Notebook
Stars
164
Forks
26
Avg merge
27m
Merged PRs (30d)
1

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from NVIDIA/nvImageCodec

All issues in NVIDIA/nvImageCodec

Similar issues

More Backend & API Design issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.