Skip to content

Update CUDA to 13.3U1 and related libraries for CTK 12 & 13#6416

Open
JanuszL wants to merge 1 commit into
NVIDIA:mainfrom
JanuszL:cuda_13.3u1
Open

Update CUDA to 13.3U1 and related libraries for CTK 12 & 13#6416
JanuszL wants to merge 1 commit into
NVIDIA:mainfrom
JanuszL:cuda_13.3u1

Conversation

@JanuszL

@JanuszL JanuszL commented Jul 1, 2026

Copy link
Copy Markdown
Contributor
  • Bump cuFile to 1.18.1.6 for CTK 13
  • Remove installation of nvJPEG and nvJPEG2k for CUDA 12 & 13
    (replaced by nvImgCodec; no longer needed as build dependencies)

Category:

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

Description:

  • Bump cuFile to 1.18.1.6 for CTK 13
  • Remove installation of nvJPEG and nvJPEG2k for CUDA 12 & 13
    (replaced by nvImgCodec; no longer needed as build dependencies)

Additional information:

Affected modules and functionalities:

  • docker toolchain files for CUDA 12 and 13

Key points relevant for the review:

  • NA

Tests:

  • Existing tests apply
    • build process and all tests
  • 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: N/A

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56467734]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56467756]: BUILD STARTED

@greptile-apps

greptile-apps Bot commented Jul 1, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

This PR updates the CUDA 13.3 toolchain to the 13.3U1 point release (bumping CUDA_SUBVERSION to 13.3.1-1 and cuFile to 1.18.1.6-1), and removes libnvjpeg / libnvjpeg2k from all four Dockerfiles since nvImgCodec now covers those use cases. The previously flagged orphaned cp commands and the dangling CUFILE_VERSION variable have both been resolved.

  • CUDA 13.3 (aarch64 & x86_64): CUDA_SUBVERSION bumped to 13.3.1-1, cuFile bumped to 1.18.1.6-1, libnvjpeg/libnvjpeg2k packages and their header-copy steps fully removed.
  • CUDA 12.9 (aarch64 & x86_64): libnvjpeg2k packages and associated cp commands removed; libcufile-dev-12-9 install is retained; CUDA apt repository URL updated from ubuntu2004 to ubuntu2204.

Confidence Score: 5/5

The Dockerfile changes are straightforward dependency removals and a version bump; the previously broken cp commands and dangling variable have been corrected.

All four files cleanly remove nvjpeg/nvjpeg2k packages and their header-copy steps without introducing new logic. The CUDA 13.3 version bumps are self-contained. The one open question — whether ubuntu2204 CUDA packages install cleanly on a ubuntu:20.04 base — is worth confirming but is unlikely to be an oversight given the author notes build tests pass.

The two CUDA 12.9 Dockerfiles (Dockerfile.cuda129.aarch64.deps and Dockerfile.cuda129.x86_64.deps) are worth a second look due to the ubuntu2004 to ubuntu2204 apt repo change against the unchanged ubuntu:20.04 base image.

Important Files Changed

Filename Overview
docker/Dockerfile.cuda129.aarch64.deps Removes nvjpeg2k packages and cp commands (previously flagged as broken), keeps libcufile-dev-12-9 install, and switches CUDA apt repo from ubuntu2004 to ubuntu2204 while the base image remains ubuntu:20.04
docker/Dockerfile.cuda129.x86_64.deps Identical to the aarch64 variant: nvjpeg2k removed, libcufile-dev-12-9 retained, CUDA repo switched from ubuntu2004 to ubuntu2204 on a ubuntu:20.04 base
docker/Dockerfile.cuda133.aarch64.deps Bumps CUDA_SUBVERSION to 13.3.1-1, cuFile to 1.18.1.6-1, and removes libnvjpeg + libnvjpeg2k packages along with their header cp commands; change is clean
docker/Dockerfile.cuda133.x86_64.deps Identical to the aarch64 cuda133 variant: cuFile bumped, nvjpeg/nvjpeg2k removed cleanly

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Base Image] --> B{CUDA Version}
    B -->|12.9| C[Install CUDA 12.9.1 via .run installer]
    B -->|13.3| D[Add ubuntu2204 CUDA apt repo]
    C --> E[Add ubuntu2204 CUDA apt repo]
    E --> F[apt-get install libcufile-dev-12-9=1.14.1.1-1]
    D --> G[apt install cuda-minimal-build + CUDA libs]
    G --> H[libcufile-dev-13-3=1.18.1.6-1]
    F --> I[Download and install nvcomp cuda12]
    H --> J[Download and install nvcomp cuda13]
    J --> K[Install fatbinary wrapper]
    I --> L[Done - CUDA 12.9 image]
    K --> M[Done - CUDA 13.3 image]
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[Base Image] --> B{CUDA Version}
    B -->|12.9| C[Install CUDA 12.9.1 via .run installer]
    B -->|13.3| D[Add ubuntu2204 CUDA apt repo]
    C --> E[Add ubuntu2204 CUDA apt repo]
    E --> F[apt-get install libcufile-dev-12-9=1.14.1.1-1]
    D --> G[apt install cuda-minimal-build + CUDA libs]
    G --> H[libcufile-dev-13-3=1.18.1.6-1]
    F --> I[Download and install nvcomp cuda12]
    H --> J[Download and install nvcomp cuda13]
    J --> K[Install fatbinary wrapper]
    I --> L[Done - CUDA 12.9 image]
    K --> M[Done - CUDA 13.3 image]
Loading

Reviews (6): Last reviewed commit: "Update CUDA to 13.3U1 and related librar..." | Re-trigger Greptile

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56467756]: BUILD FAILED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56467734]: BUILD FAILED

@JanuszL JanuszL changed the title Bumps CUDA to 13.3U1 and dependant libraries for CTK 12 & 13 Update CUDA to 13.3U1 and related libraries for CTK 12 & 13 Jul 1, 2026
@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56479409]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56479419]: BUILD STARTED

Comment thread docker/Dockerfile.cuda129.aarch64.deps Outdated
Comment thread docker/Dockerfile.cuda129.aarch64.deps
@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56480120]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56480121]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56480120]: BUILD FAILED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56484993]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56486770]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56486765]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56486770]: BUILD PASSED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56486765]: BUILD PASSED

Comment thread docker/Dockerfile.cuda129.aarch64.deps
Comment thread docker/Dockerfile.cuda129.aarch64.deps
Comment thread docker/Dockerfile.cuda129.x86_64.deps
- Bump cuFile to 1.18.1.6 for CTK 13
- Remove installation of nvJPEG and nvJPEG2k for CUDA 12 & 13
  (replaced by nvImgCodec; no longer needed as build dependencies)

Signed-off-by: Janusz Lisiecki <jlisiecki@nvidia.com>
@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56714361]: BUILD STARTED

@dali-automaton

Copy link
Copy Markdown
Collaborator

CI MESSAGE: [56714361]: BUILD PASSED

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants