1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
{
buildRedist,
cudaAtLeast,
cudaOlder,
lib,
fetchpatch,
}:
buildRedist {
redistName = "cuda";
pname = if cudaAtLeast "13.3" then "cccl" else "cuda_cccl";
# Restrict header-only packages to a single output.
# Also, when using multiple outputs (i.e., `out`, `dev`, and `include`), something isn't being patched correctly,
# so libnvshmem fails to build, complaining about being unable to find the thrust include directory. This is likely
# because the `dev` output contains the CMake configuration and is written to assume it will share a parent
# directory with the include directory rather than be in a separate output.
outputs = [ "out" ];
prePatch = lib.optionalString (cudaAtLeast "13.0") ''
nixLog "removing top-level $PWD/include/nv directory"
rm -rfv "$PWD/include/nv"
nixLog "un-nesting top-level $PWD/include/cccl directory"
mv -v "$PWD/include/cccl"/* "$PWD/include/"
nixLog "removing empty $PWD/include/cccl directory"
rmdir -v "$PWD/include/cccl"
'';
patches =
lib.optionals (cudaAtLeast "12.9" && cudaOlder "13.0") [
# Fix missing _CCCL_PP_SPLICE_WITH_IMPL20 in preprocessor.h
# https://github.com/NVIDIA/cccl/issues/4967
# https://github.com/NVIDIA/cccl/pull/4972
(fetchpatch {
name = "fix-missing-_CCCL_PP_SPLICE_WITH_IMPL20";
url = "https://github.com/NVIDIA/cccl/commit/2c2276d8b19d737cb16811ce2eb761030f472e60.patch";
stripLen = 1;
hash = "sha256-hYfMFsd7Y8CwuNGaPYG6uEB+lg1TmWSIIU5ToVMULKY=";
})
]
++ lib.optionals (cudaAtLeast "13.2" && cudaOlder "13.4") [
# Fix onnxruntime compilation error: https://github.com/microsoft/onnxruntime/issues/28023
# Backport: https://github.com/NVIDIA/cccl/pull/8771
(fetchpatch {
name = "fix-invalid-cpp-syntax";
url = "https://github.com/NVIDIA/cccl/commit/8e41eeabe54ab9ae48ad5640cfa7153b0a1071af.patch";
stripLen = 2;
extraPrefix = "include/";
hash = "sha256-eko1GSD2NPSLtbQ3diwgKMbuS6wU9lHDOajLJy5lBwM=";
})
];
# NVIDIA, in their wisdom, expect CCCL to be a directory inside include.
# https://github.com/NVIDIA/cutlass/blob/087c84df83d254b5fb295a7a408f1a1d554085cf/CMakeLists.txt#L773
postInstall = lib.optionalString (cudaAtLeast "13.0") ''
nixLog "creating alias for ''${!outputInclude:?}/include/cccl"
ln -srv "''${!outputInclude:?}/include" "''${!outputInclude:?}/include/cccl"
'';
meta = {
description = "Building blocks that make it easier to write safe and efficient CUDA C++ code";
longDescription = ''
The goal of CCCL is to provide CUDA C++ developers with building blocks that make it easier to write safe and
efficient code.
'';
homepage = "https://github.com/NVIDIA/cccl";
changelog = "https://github.com/NVIDIA/cccl/releases";
};
}
|
