Upgrade nvimgcodec to 0.9.0

[mzient]

Category:

Other (e.g. Documentation, Tests, Configuration)

Description:

This PR updates the nvimgcodec dependency to version 0.9 and removes the workarounds needed for 0.8.

Additional information:

Affected modules and functionalities:

Key points relevant for the review:

Tests:

• Existing tests apply
• New tests added
  • Python tests
  • GTests
  • Benchmark
  • Other
• N/A

Checklist

Documentation

• Existing documentation applies
• Documentation updated
  • Docstring
  • Doxygen
  • RST
  • Jupyter
  • Other
• N/A

DALI team only

Requirements

• Implements new requirements
• Affects existing requirements
• N/A

REQ IDs: N/A

JIRA TASK: DALI-4811

[greptile-apps]

Greptile Summary

This PR bumps the nvimgcodec dependency from 0.8.x to 0.9.x and removes the ROI/EXIF-orientation workaround that was required for the older release. Version bounds are updated consistently across CMake, conda recipes, and ExternalProject_Add; the workaround code paths in image_decoder.h and the associated comment in convert_test.cc are cleanly excised.

• cmake/Dependencies.common.cmake: Minimum version raised to 0.9.0, FetchContent fallback updated to 0.9.0.20 tarballs (split by CUDA major version 12/13), and ExternalProject_Add GIT_TAG moved from v0.8.0 to v0.9.0. The include/ directory path in the tarball is now a direct child of the source root rather than inside a per-CUDA subdirectory.
• dali/operators/imgcodec/image_decoder.h: roi_orient_war_ flag, post_decode_orientation per-sample state, OutputToRawROI() helper, and all related decode-path branches are removed; apply_exif_orientation is now set unconditionally from use_orientation_.
• conda recipes: libnvimgcodec* version pins updated to >=0.9.0,<0.10.0 in both dali_native_libs and dali_python_bindings.

Confidence Score: 5/5

Safe to merge — all three build paths now consistently target nvimgcodec 0.9.0, the workaround removal is correct given the updated minimum version, and the conda recipes are updated in lock-step.

The workaround removal is well-scoped: gated behind the 0.9.0 minimum enforced at runtime by EnforceMinimumNvimgcodecVersion. The static-link ExternalProject_Add was updated to v0.9.0 and the FetchContent fallback now fetches the correct 0.9.0.20 tarballs per CUDA major version.

cmake/Dependencies.common.cmake — the bare message() added in the FetchContent block should carry a STATUS prefix to match the rest of the file.

Important Files Changed

Filename	Overview
cmake/Dependencies.common.cmake	Version bounds, FetchContent URL/hash, and ExternalProject GIT_TAG all updated to 0.9.0; one bare message() without STATUS prefix is a leftover debug print.
dali/operators/imgcodec/image_decoder.h	ROI/orientation workaround removed cleanly; apply_exif_orientation now unconditionally reflects use_orientation_; no residual branching on roi_orient_war_ remains.
dali/operators/imgcodec/util/convert_test.cc	Comment updated to remove the reference to the now-deleted WAR; no logic changes.
conda/dali_native_libs/recipe/meta.yaml	All four libnvimgcodec* version pins updated from >=0.8.0,<0.9.0 to >=0.9.0,<0.10.0 consistently.
conda/dali_python_bindings/recipe/meta.yaml	All four libnvimgcodec* version pins updated from >=0.8.0,<0.9.0 to >=0.9.0,<0.10.0 consistently.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[BUILD_NVIMAGECODEC] --> B{WITH_DYNAMIC_NVIMGCODEC?}
    B -- Yes --> C{find_package nvimgcodec 0.9.0..0.10.0}
    C -- Found --> D[Use system headers]
    C -- Not Found --> E{CUDA_VERSION_MAJOR?}
    E -- 13 --> F[FetchContent cuda13 tarball 0.9.0.20]
    E -- 12 --> G[FetchContent cuda12 tarball 0.9.0.20]
    E -- Other --> H[FATAL_ERROR]
    F --> I[include_dirs source_dir/include]
    G --> I
    B -- No --> J[ExternalProject_Add GIT_TAG v0.9.0]
    J --> K[Static link nvimgcodec 0.9.0]
    D --> L[Dynamic load nvimgcodec >= 0.9.0 < 0.10.0]
    I --> L

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
    A[BUILD_NVIMAGECODEC] --> B{WITH_DYNAMIC_NVIMGCODEC?}
    B -- Yes --> C{find_package nvimgcodec 0.9.0..0.10.0}
    C -- Found --> D[Use system headers]
    C -- Not Found --> E{CUDA_VERSION_MAJOR?}
    E -- 13 --> F[FetchContent cuda13 tarball 0.9.0.20]
    E -- 12 --> G[FetchContent cuda12 tarball 0.9.0.20]
    E -- Other --> H[FATAL_ERROR]
    F --> I[include_dirs source_dir/include]
    G --> I
    B -- No --> J[ExternalProject_Add GIT_TAG v0.9.0]
    J --> K[Static link nvimgcodec 0.9.0]
    D --> L[Dynamic load nvimgcodec >= 0.9.0 < 0.10.0]
    I --> L

Loading

_{Reviews (3): Last reviewed commit: "Use upstream nvimgcodec." | Re-trigger Greptile}

[dali-automaton]

CI MESSAGE: [58774048]: BUILD STARTED

[mzient]

@greptileai Please review again

[dali-automaton]

CI MESSAGE: [58785890]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [58785874]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [58774048]: BUILD PASSED

[dali-automaton]

CI MESSAGE: [58785890]: BUILD PASSED

[dali-automaton]

CI MESSAGE: [58785874]: BUILD PASSED

[dali-automaton]

CI MESSAGE: [58896425]: BUILD STARTED

[mzient]

Now there are separate archives for CUDA 12 and 13.

[dali-automaton]

CI MESSAGE: [58896425]: BUILD PASSED