a96a7b9dc6
treewide: use finalAttrs.finalPackage.doCheck ( #543265 )
2026-07-23 11:35:08 +00:00
6bb92e1743
cudaPackages.nvbandwidth: 0.9 -> 0.10.0
...
Diff: https://github.com/NVIDIA/nvbandwidth/compare/v0.9...v0.10.0
Changelog: https://github.com/NVIDIA/nvbandwidth/blob/v0.10.0/CHANGELOG.md
2026-07-21 19:16:35 +00:00
a0d022d0f4
treewide: use finalAttrs.finalPackage.doCheck
2026-07-18 18:23:21 +02:00
ba8e45b395
cudaPackages.nccl-tests: 2.18.5 -> 2.19.1
2026-06-26 05:38:59 +00:00
54690a944c
cuda-modules: Allow SBSA Orin to use CUDA 13+ ( #528562 )
2026-06-24 02:37:24 +00:00
688b86390a
cudaPackages.nccl-tests: 2.18.3 -> 2.18.5 ( #532666 )
2026-06-22 09:46:09 +00:00
3fecd68e9d
cudaPackages_13_3.cccl: init at 13.3.3.3.1
2026-06-21 16:56:43 +00:00
8b249f2032
cudaPackages.nccl-tests: 2.18.3 -> 2.18.5
2026-06-17 13:13:10 +00:00
b26384ff1f
cudaPackages.cudnn-frontend: 1.24.0 -> 1.25.0
2026-06-17 04:38:21 +00:00
44bd912379
cudaPackages.cutlass: 4.5.1 -> 4.5.2
2026-06-16 04:40:20 +00:00
f666e54eb0
cudaPackages.libnvshmem: default withNccl to nccl.meta.available
...
`withNccl` defaults to `true`, unconditionally adding `nccl` to `buildInputs`.
This is problematic on platforms where NCCL is unavailable, such as pre-Thor Jetson (Orin), where `nccl.meta.available` is `false` due to its `badPlatforms` assertion. In such cases, libnvshmem would still pull the unavailable `nccl` into its closure, leading to an eval-time error.
We are changing the default to `nccl.meta.available` to make libnvshmem skip NCCL automatically on platforms where NCCL is unsupported. We preserving the existing behaviour everywhere else (x86_64-linux, Thor Jetson) where `nccl.meta.available` is `true`.
Note that this guard has already been added to other CUDA modules such as `cudaPackages.cudss` in #437723 .
2026-06-15 19:25:15 +02:00
3e1852144b
cudaPackages.gdrcopy: 2.5.2 -> 2.6
...
Diff: https://github.com/NVIDIA/gdrcopy/compare/v2.5.2...v2.6
2026-06-13 12:30:46 +00:00
048adb9b01
cudaPackages.nccl: 2.28.7-1 -> 2.30.7-1
...
Diff: https://github.com/NVIDIA/nccl/compare/v2.28.7-1...v2.30.7-1
Changelog: https://github.com/NVIDIA/nccl/releases/tag/v2.30.7-1
2026-06-11 07:22:15 +00:00
e4d4b7fa70
cuda-modules: Allow SBSA Orin to use CUDA 13+
...
NVIDIA announced an SBSA software stack for the Jetson Orin (capability
8.7) using CUDA 13.2 [1]. Update the redist system selection logic and
assertions to support Orin SBSA.
[1]: https://forums.developer.nvidia.com/t/jetpack-7-2-jetson-linux-r39-2-on-jetson-agx-orin-developer-kit-getting-started-and-feedback-thread/372156
2026-06-06 13:51:30 -07:00
5bb67e892a
cudaPackages.tensorrt-samples: Add support for 10.16.1
2026-06-05 13:33:21 -07:00
9d3d755e8e
cudaPackages.tensorrt: Add TensorRT 10.16.1 manifest
2026-06-05 13:32:52 -07:00
9384fdf955
cudaPackages.cutlass: fix license
2026-06-03 15:00:39 +00:00
c2b2a2e682
cudaPackages.cutlass: 3.9.2 -> 4.5.1
...
Diff: https://github.com/NVIDIA/cutlass/compare/v3.9.2...v4.5.1
2026-06-03 14:31:40 +00:00
ffa10e26ae
cudaPackages.cudnn-frontend: 1.16.0 -> 1.24.0 ( #526689 )
2026-06-02 02:01:59 +00:00
ee06256111
cudaPackages_13_3: init at 13.3.0 ( #525130 )
2026-06-01 18:22:14 +00:00
916a743067
cudaPackages.cudnn-frontend: 1.16.0 -> 1.24.0
2026-06-01 15:18:04 +00:00
4b31bed983
cudaPackages.cuda_cccl: fix cpp syntax error on CUDA 13.2
...
https://github.com/NVIDIA/cccl/pull/8771
Fixes onnxruntime: https://github.com/microsoft/onnxruntime/issues/28023
2026-05-28 21:56:19 +00:00
d5bbc589f2
cudaPackages_13_2: 13.2.0 -> 13.2.1
...
Changelog:
https://docs.nvidia.com/cuda/archive/13.2.1/cuda-toolkit-release-notes/index.html
2026-05-28 07:34:34 +00:00
fe55e1a378
cudaPackages_13_3: init at 13.3.0
...
Changelog:
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html
2026-05-28 07:21:46 +00:00
0bbc8dffae
treewide: set meta.license on hooks
2026-05-25 13:48:20 +02:00
b0015daf6f
cudaPackages.tensorrt: mark insecure
2026-05-22 23:44:39 +03:00
135ac68cdc
cudaPackages.cudnn: 9.13.0 -> 9.22.0
...
u
2026-05-17 20:47:35 +00:00
a2baf7bb69
cudaPackages.libcusolvermp: fix build
2026-05-13 14:30:32 +00:00
4009183fe6
cudaPackages.libcusolvermp: 0.7.0 -> 0.8.0
...
Changelog:
https://docs.nvidia.com/cuda/cusolvermp/release_notes/index.html#cusolvermp-v0-8-0
2026-05-13 08:20:44 +00:00
daa7b35604
cudaPackages.libcublasmp: 0.6.0 -> 0.8.1
...
Changelog:
https://docs.nvidia.com/cuda/cublasmp/release_notes/index.html
2026-05-05 08:25:00 +00:00
ad39f6cb9f
cudaPackages.nvbandwidth: init at 0.9
2026-04-30 15:58:13 +00:00
4ed0f75b07
cudaPackages_13_0: 13.0.2 -> 13.0.3
...
Changelog:
https://docs.nvidia.com/cuda/archive/13.0.3/cuda-toolkit-release-notes/index.html
2026-04-27 18:53:11 +00:00
01d468622e
cudaPackages.cuda_crt: do not patch math headers on >=13.2
2026-04-27 08:57:36 +00:00
0db3511132
cudaPackages.cuda_crt: patch math_functions.h signatures
2026-04-23 23:05:33 +00:00
4bab2aa6af
cudaPackages.libcuobjclient: init at 1.0.0.26
2026-04-21 15:25:07 +00:00
a9128c6c5b
cudaPackages.nccl-tests: 2.18.2 -> 2.18.3
2026-04-17 10:37:57 +00:00
8e55aea6ee
cuda_nvcc: fix host_defines.h __noinline__ conflict under clang
...
host_defines.h redefines __noinline__ as __attribute__((noinline)).
When clang compiles CUDA code against libstdc++ >=12, which uses
__attribute__((__noinline__)), the macro causes the invalid expansion
__attribute__((__attribute__((noinline)))). Clang natively understands
__noinline__ as an attribute, so the macro is unnecessary — skip it
when clang is the compiler.
2026-04-12 22:06:57 +00:00
be983daa60
cudaPackages.libnvshmem: make the build customizable
2026-04-01 07:37:35 +00:00
0ccff482d4
cudaPackages.nccl-tests: 2.17.10 -> 2.18.2 ( #499938 )
2026-03-29 16:47:13 +00:00
238897676d
cudaPackages.libnvshmem: 3.4.5-0 -> 3.6.5-0 ( #503855 )
2026-03-28 21:10:36 +00:00
75b7fb4416
cudaPackages.nsight_compute: include necessary files
2026-03-27 14:20:34 +01:00
6a78d900f4
cudaPackages.nsight_systems: support new directory structure
2026-03-26 22:52:22 +01:00
02e32e37db
cudaPackages.libnvshmem: add GaetanLepage to maintainers
2026-03-26 21:21:44 +00:00
8639b5c1d0
cudaPackages.libnvshmem: 3.4.5-0 -> 3.6.5-0
...
Diff: https://github.com/NVIDIA/nvshmem/compare/v3.4.5-0...v3.6.5-0
Changelog: https://github.com/NVIDIA/nvshmem/releases/tag/v3.6.5-0
2026-03-26 21:21:44 +00:00
49a82e7410
cudaPackages_12_9.cuda_cccl: patch to fix missing _CCCL_PP_SPLICE_WITH_IMPL20 in preprocessor.h
...
PR: https://github.com/NVIDIA/cccl/pull/4972
Issue: https://github.com/NVIDIA/cccl/issues/4967
2026-03-23 17:57:32 +00:00
35e9cb5e37
cudaPackages.gdrcopy: 2.5.1 -> 2.5.2 ( #500610 )
2026-03-23 14:23:56 +00:00
b6a834e180
cuda_nvcc: also patch math_functions.hpp for glibc 2.42 ( #498678 )
2026-03-22 09:08:27 +00:00
9084f40a63
cudaPackages.cuda_cccl: patch to fix torch build on cuda>=13.0
...
Co-authored-by: Connor Baker <ConnorBaker01@gmail.com >
2026-03-18 17:14:10 +00:00
5bac571559
cudaPackages.gdrcopy: 2.5.1 -> 2.5.2
2026-03-17 02:32:25 +00:00
f5b25da559
cudaPackages.nccl-tests: 2.17.10 -> 2.18.2
2026-03-14 22:45:03 +00:00
Yohann Boniface