cuda-cccl 0.3.2__cp313-cp313-manylinux_2_24_aarch64.whl → 0.3.4__cp313-cp313-manylinux_2_24_aarch64.whl

cuda/cccl/headers/include/cub/detail/array_utils.cuh

@@ -1,27 +1,5 @@
-/***********************************************************************************************************************
- * SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
- * SPDX-License-Identifier: BSD-3-Clause
- *
- * Redistribution and use in source and binary forms, with or without modification, are permitted provided that the
- * following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
- * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
- * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
- * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- **********************************************************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -48,7 +26,6 @@
 CUB_NAMESPACE_BEGIN
 namespace detail
 {
-
 #ifndef _CCCL_DOXYGEN_INVOKED // Do not document
 
 /***********************************************************************************************************************
@@ -73,6 +50,5 @@ to_array(const Input& input)
 }
 
 #endif // !_CCCL_DOXYGEN_INVOKED
-
 } // namespace detail
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/choose_offset.cuh

@@ -1,29 +1,5 @@
-/******************************************************************************
- * Copyright (c) 2011-2024, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2011-2024, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -51,7 +27,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 /**
  * choose_offset checks NumItemsT, the type of the num_items parameter, and
  * selects the offset type based on it.
@@ -155,7 +130,6 @@ struct common_iterator_value
 };
 template <typename... Iter>
 using common_iterator_value_t = typename common_iterator_value<Iter...>::type;
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/detect_cuda_runtime.cuh

@@ -1,30 +1,6 @@
-/******************************************************************************
- * Copyright (c) 2011, Duane Merrill.  All rights reserved.
- * Copyright (c) 2011-2021, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2011, Duane Merrill. All rights reserved.
+// SPDX-FileCopyrightText: Copyright (c) 2011-2021, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 /**
  * @file

cuda/cccl/headers/include/cub/detail/device_double_buffer.cuh

@@ -32,7 +32,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 /**
  * @brief It's a double-buffer storage wrapper for multi-pass stream
  *        transformations that require more than one storage array for
@@ -90,7 +89,6 @@ public:
     m_alternate_buffer = tmp;
   }
 };
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/device_memory_resource.cuh

@@ -13,15 +13,14 @@
 #  pragma system_header
 #endif // no system header
 
+#include <cuda/__runtime/api_wrapper.h>
 #include <cuda/__stream/stream_ref.h>
-#include <cuda/std/__cuda/api_wrapper.h>
 #include <cuda/std/cstdint>
 
 CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 // TODO(gevtushenko/srinivasyadav18): move cudax `device_memory_resource` to `cuda::__device_memory_resource` and remove
 // this implementation
 struct device_memory_resource
@@ -56,7 +55,6 @@ struct device_memory_resource
     _CCCL_TRY_CUDA_API(::cudaFreeAsync, "deallocate failed", ptr, stream.get());
   }
 };
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/fast_modulo_division.cuh

@@ -1,29 +1,5 @@
-/******************************************************************************
- * Copyright (c) 2011-2024, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2011-2024, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -62,7 +38,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 /***********************************************************************************************************************
  * larger_unsigned_type
  **********************************************************************************************************************/
@@ -243,7 +218,6 @@ private:
   unsigned _shift_right  = 0;
 };
 _CCCL_DIAG_POP
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/integer_utils.cuh

@@ -31,7 +31,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 template <typename Input>
 _CCCL_DEVICE _CCCL_FORCEINLINE auto split_integer(Input input)
 {
@@ -83,6 +82,5 @@ _CCCL_DEVICE _CCCL_FORCEINLINE auto comparable_int_to_floating_point(IntegerType
   auto value1           = static_cast<IntegerType>(value < 0 ? lowest - value : value);
   return cub::detail::unsafe_bitcast<FloatingPointType>(value1);
 }
-
 } // namespace detail
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/launcher/cuda_driver.cuh

@@ -21,7 +21,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 struct CudaDriverLauncher
 {
   dim3 grid;
@@ -136,7 +135,6 @@ struct CudaDriverLauncherFactory
   CUdevice device;
   int cc;
 };
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/launcher/cuda_runtime.cuh

@@ -21,7 +21,6 @@ CUB_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 struct TripleChevronFactory
 {
   CUB_RUNTIME_FUNCTION THRUST_NS_QUALIFIER::cuda_cub::detail::triple_chevron operator()(
@@ -94,7 +93,6 @@ struct TripleChevronFactory
     return cudaDeviceGetAttribute(&max_shared_memory, cudaDevAttrMaxSharedMemoryPerBlock, device);
   }
 };
-
 } // namespace detail
 
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/mdspan_utils.cuh

@@ -24,7 +24,6 @@
 CUB_NAMESPACE_BEGIN
 namespace detail
 {
-
 _CCCL_DIAG_PUSH
 _CCCL_DIAG_SUPPRESS_MSVC(4702) // unreachable code (even if there are no branches!)
 
@@ -109,6 +108,5 @@ template <typename MappingTypeLhs, typename MappingTypeRhs>
   }
   return true;
 }
-
 } // namespace detail
 CUB_NAMESPACE_END

cuda/cccl/headers/include/cub/detail/ptx-json/README.md

@@ -1,13 +1,8 @@
 # Short intro to ptx-json
 
-This directory contains a library for embedding an almost valid JSON string in a PTX instruction stream produced with
-the CUDA C++ compiler. This tool is useful for transmitting structured information from device TUs compiled with NVRTC
-to the code compiling said TUs, and is used in c.parallel for figuring out the correct tuning policies for algorithms.
-
-> [!CAUTION]
-> The PTX instruction stream produced when using this library is NOT valid. If you use this library, make sure to only
-> instantiate its functions when compiling to PTX and no further. You should hide their instantiations behind an opt-in
-> macro for general code.
+This directory contains a library for embedding a valid JSON string in a CUBIN produced with the CUDA C++ compiler. This
+tool is useful for transmitting structured information from device TUs compiled with NVRTC to the code compiling said TUs,
+and is used in c.parallel for figuring out the correct tuning policies for algorithms.
 
 ## Example code
 
@@ -20,8 +15,7 @@ This code is taken from a test of this library, demonstrating all the current ca
                      (ptx_json::key<"c">() = ptx_json::array<1, 2, ptx_json::string("a")>())>();
 ```
 
-When compiled into PTX and extracted using a JSON parser capable of parsing C-style comments (such as `nlohmann/json`
-with an opt-in argument), this produces a JSON equivalent to
+When compiled into a CUBIN and extracted using a JSON parser (such as `nlohmann/json`), this produces a JSON equivalent to
 
 ```json
 {
@@ -46,8 +40,9 @@ This type is used to lift core constant expressions into the type system. All co
 template arguments into `ptx_json::value` objects.
 
 Currently supported types are:
-* `int`
-* `ptx_json::string`
+* integral types;
+* `bool`;
+* `ptx_json::string`.
 
 ### `ptx_json::array`
 

cuda/cccl/headers/include/cub/detail/ptx-json/array.h

@@ -1,29 +1,5 @@
-/******************************************************************************
- * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -42,19 +18,16 @@ struct array<>
 {
   __forceinline__ __device__ static void emit()
   {
-    asm volatile("[]" ::: "memory");
+    asm volatile("[]");
   }
 };
 
 template <not_a_keyed_value auto First, not_a_keyed_value auto... NKVs>
 struct array<First, NKVs...>
 {
-  __forceinline__ __device__ static void emit()
+  __device__ consteval static auto emit()
   {
-    asm volatile("[" ::: "memory");
-    value<First>::emit();
-    ((comma(), value<NKVs>::emit()), ...);
-    asm volatile("]" ::: "memory");
+    return string("[", value<First>::emit(), string(comma(), value<NKVs>::emit())..., "]");
   }
 };
 
@@ -63,6 +36,6 @@ struct is_array : cuda::std::false_type
 {};
 
 template <auto... NKV>
-struct is_array<array<NKV...>> : cuda::std::true_type
+struct is_array<object<NKV...>> : cuda::std::true_type
 {};
 } // namespace ptx_json

cuda/cccl/headers/include/cub/detail/ptx-json/json.h

@@ -1,29 +1,5 @@
-/******************************************************************************
- * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -38,24 +14,25 @@
 
 namespace ptx_json
 {
-template <auto T, typename = value_traits<T>::type>
-struct tagged_json;
+template <auto V, typename = cuda::std::make_index_sequence<V.Length>>
+const char reify[V.Length] = {};
 
-template <int N, string<N> T, cuda::std::size_t... Is>
-struct tagged_json<T, cuda::std::index_sequence<Is...>>
+template <int N, string<N> V, cuda::std::size_t... Is>
+__device__ const char reify<V, cuda::std::index_sequence<Is...>>[] = {V.str[Is]...};
+
+template <auto Tag>
+struct tagged_json
 {
   template <typename V, typename = cuda::std::enable_if_t<is_object<V>::value || is_array<V>::value>>
-  __noinline__ __device__ void operator=(V)
+  __device__ consteval auto& operator=(V v)
   {
-    static constexpr char str[]{T.str[Is]...};
-    asm volatile("cccl.ptx_json.begin(%0)\n\n" ::"C"(str) : "memory");
-    V::emit();
-    asm volatile("\ncccl.ptx_json.end(%0)" ::"C"(str) : "memory");
+    return reify<string(
+      "\ncccl.ptx_json.begin(", value<Tag>::emit(), ")", V::emit(), "cccl.ptx_json.end(", value<Tag>::emit(), ")\n")>;
   }
 };
 
 template <auto T>
-__forceinline__ __device__ tagged_json<T> id()
+__device__ consteval tagged_json<T> id()
 {
   return {};
 }

cuda/cccl/headers/include/cub/detail/ptx-json/object.h

@@ -1,29 +1,5 @@
-/******************************************************************************
- * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
@@ -37,11 +13,9 @@ namespace ptx_json
 template <auto K, typename V>
 struct keyed_value
 {
-  __forceinline__ __device__ static void emit()
+  __device__ consteval static auto emit()
   {
-    value<K>::emit();
-    asm volatile(":" ::: "memory");
-    V::emit();
+    return string(value<K>::emit(), ":", V::emit());
   }
 };
 
@@ -62,21 +36,18 @@ struct object;
 template <>
 struct object<>
 {
-  __forceinline__ __device__ static void emit()
+  __device__ consteval static auto emit()
   {
-    asm volatile("{}" ::: "memory");
+    return string("{}");
   }
 };
 
 template <a_keyed_value auto First, a_keyed_value auto... KVs>
 struct object<First, KVs...>
 {
-  __forceinline__ __device__ static void emit()
+  __device__ consteval static auto emit()
   {
-    asm volatile("{" ::: "memory");
-    First.emit();
-    ((comma(), KVs.emit()), ...);
-    asm volatile("}" ::: "memory");
+    return string("{", First.emit(), string(comma(), KVs.emit())..., "}");
   }
 };
 
@@ -92,7 +63,7 @@ template <string V>
 struct key
 {
   template <typename U>
-  __forceinline__ __device__ constexpr keyed_value<V, U> operator=(U)
+  __device__ consteval keyed_value<V, U> operator=(U u)
   {
     return {};
   }

cuda/cccl/headers/include/cub/detail/ptx-json/string.h

@@ -1,53 +1,79 @@
-/******************************************************************************
- * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in the
- *       documentation and/or other materials provided with the distribution.
- *     * Neither the name of the NVIDIA CORPORATION nor the
- *       names of its contributors may be used to endorse or promote products
- *       derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
- * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
- * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
- * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
- * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- ******************************************************************************/
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: BSD-3
 
 #pragma once
 
+#include <cuda/std/__type_traits/remove_cvref.h>
+
 namespace ptx_json
 {
+#pragma nv_diag_suppress 177
+template <int N>
+struct string;
+
+template <typename T>
+struct string_part_len_impl;
+
+template <int N>
+struct string_part_len_impl<string<N>>
+{
+  static const constexpr auto value = N - 1;
+};
+
+template <int N>
+struct string_part_len_impl<char[N]>
+{
+  static const constexpr auto value = N - 1;
+};
+
+template <typename T>
+constexpr int string_part_len = string_part_len_impl<T>::value;
+
+template <int N>
+__device__ constexpr auto& get_string_part(const string<N>& str)
+{
+  return str.str;
+}
+
+template <int N>
+__device__ constexpr auto& get_string_part(const char (&str)[N])
+{
+  return str;
+}
+
 template <int N>
 struct string
 {
   static const constexpr auto Length = N;
 
-  __device__ constexpr string(const char (&c)[N])
+  template <typename... Ts>
+  __device__ constexpr string(Ts&&... strings)
+  {
+    static_assert(N == (0 + ... + string_part_len<cuda::std::remove_cvref_t<Ts>>) +1);
+    int offset = 0;
+    (init_part(get_string_part(strings), offset), ...);
+    str[offset] = 0;
+  }
+
+  template <int M>
+  __device__ constexpr void init_part(const char (&other)[M], int& offset)
   {
-    for (int i = 0; i < N; ++i)
+    for (int i = 0; i < M - 1; ++i)
     {
-      str[i] = c[i];
+      str[offset + i] = other[i];
     }
-    (void) Length;
+    offset += M - 1;
   }
 
   char str[N];
 };
 
-__forceinline__ __device__ void comma()
+template <typename... Ts>
+string(Ts&&...) -> string<(0 + ... + string_part_len<cuda::std::remove_cvref_t<Ts>>) +1>;
+#pragma nv_diag_default 177
+
+__device__ consteval auto comma()
 {
-  asm volatile("," ::: "memory");
+  return string(",");
 }
 } // namespace ptx_json