accelerated_scan.warp compilation gives error.

[roman8ivanov]

Hi,
I tried scan from accelerated_scan.warp. I observed it requires compilation every time I import function.
However, I constantly receive error related to "error C2872: 'std': ambiguous symbol" (see below).

System params:
Python 3.13.9
Cuda 12.8
accelerated-scan 0.3.1
pytorch 2.9.1
triton 3.5.1
Windows 11

Error:

tmpxft_00006d14_00000000-7_cuda.compute_90.cudafe1.cpp
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1134): error C2872: 'std': ambiguous symbol
C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.41.34120\include\valarray(20): note: could be 'std'
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1134): note: or 'std'
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1134): note: the template instantiation context (the oldest one first) is
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1181): note: see reference to class template instantiation 'torch::dynamo::autograd::IValuePacker<__int64>' being compiled
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1108): note: while compiling class template member function 'c10::TypePtr torch::dynamo::autograd::IValuePacker<__int64>::packed_type(void)'
C:/ProgramData/anaconda3/Lib/site-packages/torch/include\torch/csrc/dynamo/compiled_autograd.h(1181): note: see the first reference to 'torch::dynamo::autograd::IValuePacker<__int64>::packed_type' in 'torch::dynamo::autograd::IValuePacker::packed_type'
ninja: build stopped: subcommand failed.

CalledProcessError Traceback (most recent call last)
File C:\ProgramData\anaconda3\Lib\site-packages\torch\utils\cpp_extension.py:2597, in _run_ninja_build(build_directory, verbose, error_prefix)
2596 stdout_fileno = 1
-> 2597 subprocess.run(
2598 command,
2599 shell=IS_WINDOWS and IS_HIP_EXTENSION,
2600 stdout=stdout_fileno if verbose else subprocess.PIPE,
2601 stderr=subprocess.STDOUT,
2602 cwd=build_directory,
2603 check=True,
2604 env=env)
2605 except subprocess.CalledProcessError as e:
2606 # Python 2 and 3 compatible way of getting the error object.

File C:\ProgramData\anaconda3\Lib\subprocess.py:577, in run(input, capture_output, timeout, check, *popenargs, **kwargs)
576 if check and retcode:
--> 577 raise CalledProcessError(retcode, process.args,
578 output=stdout, stderr=stderr)
579 return CompletedProcess(process.args, retcode, stdout, stderr)

CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 2.

The above exception was the direct cause of the following exception:

RuntimeError Traceback (most recent call last)
Cell In[26], line 1
----> 1 from accelerated_scan.warp import scan

File C:\ProgramData\anaconda3\Lib\site-packages\accelerated_scan\warp.py:13
6 cuda_source = (Path(file).parent / 'warp.cuh').read_text()
8 cpp_source = """
9 at::Tensor warpscan_forward(const at::Tensor &gates, const at::Tensor &tokens, const at::Tensor &out, const bool reverse);
10 void warpscan_backward(const at::Tensor &gates, const at::Tensor &output, const at::Tensor &outGrad, const at::Tensor& gateGradOut, const at::Tensor& valueGradOut);
11 """
---> 13 module = load_inline(
14 name='warpscan',
15 cpp_sources=[cpp_source],
16 cuda_sources=[cuda_source],
17 functions=['warpscan_forward', 'warpscan_backward'],
18 verbose=True,
19 extra_cflags=['-DNOMINMAX'], #I added this
20 extra_cuda_cflags=[
21 "-O3",
22 "-std=c++17",
23 "-DNOMINMAX", #I added this
24 "--ptxas-options=-v",
25 "-lineinfo",
26 "--fmad", "false",
27 "-U__CUDA_NO_HALF_OPERATORS__", "-U__CUDA_NO_HALF_CONVERSIONS__",
28 "-U__CUDA_NO_BFLOAT16_OPERATORS__", "-U__CUDA_NO_BFLOAT16_CONVERSIONS__",
29 ]
30 )
31 warpscan_forward = module.warpscan_forward
32 warpscan_backward = module.warpscan_backward

File C:\ProgramData\anaconda3\Lib\site-packages\torch\utils\cpp_extension.py:2051, in load_inline(name, cpp_sources, cuda_sources, sycl_sources, functions, extra_cflags, extra_cuda_cflags, extra_sycl_cflags, extra_ldflags, extra_include_paths, build_directory, verbose, with_cuda, with_sycl, is_python_module, with_pytorch_error_handling, keep_intermediates, use_pch, no_implicit_headers)
2047 _maybe_write(sycl_source_path, "\n".join(sycl_sources))
2049 sources.append(sycl_source_path)
-> 2051 return _jit_compile(
2052 name,
2053 sources,
2054 extra_cflags,
2055 extra_cuda_cflags,
2056 extra_sycl_cflags,
2057 extra_ldflags,
2058 extra_include_paths,
2059 build_directory,
2060 verbose,
2061 with_cuda,
2062 with_sycl,
2063 is_python_module,
2064 is_standalone=False,
2065 keep_intermediates=keep_intermediates)

File C:\ProgramData\anaconda3\Lib\site-packages\torch\utils\cpp_extension.py:2134, in _jit_compile(name, sources, extra_cflags, extra_cuda_cflags, extra_sycl_cflags, extra_ldflags, extra_include_paths, build_directory, verbose, with_cuda, with_sycl, is_python_module, is_standalone, keep_intermediates)
2130 hipified_sources.add(hipify_result[s_abs].hipified_path if s_abs in hipify_result else s_abs)
2132 sources = list(hipified_sources)
-> 2134 _write_ninja_file_and_build_library(
2135 name=name,
2136 sources=sources,
2137 extra_cflags=extra_cflags or [],
2138 extra_cuda_cflags=extra_cuda_cflags or [],
2139 extra_sycl_cflags=extra_sycl_cflags or [],
2140 extra_ldflags=extra_ldflags or [],
2141 extra_include_paths=extra_include_paths or [],
2142 build_directory=build_directory,
2143 verbose=verbose,
2144 with_cuda=with_cuda,
2145 with_sycl=with_sycl,
2146 is_standalone=is_standalone)
2147 elif verbose:
2148 logger.debug('No modifications detected for re-loaded extension module %s, skipping build step...', name)

File C:\ProgramData\anaconda3\Lib\site-packages\torch\utils\cpp_extension.py:2286, in _write_ninja_file_and_build_library(name, sources, extra_cflags, extra_cuda_cflags, extra_sycl_cflags, extra_ldflags, extra_include_paths, build_directory, verbose, with_cuda, with_sycl, is_standalone)
2284 if verbose:
2285 logger.info('Building extension module %s...', name)
-> 2286 _run_ninja_build(
2287 build_directory,
2288 verbose,
2289 error_prefix=f"Error building extension '{name}'")

File C:\ProgramData\anaconda3\Lib\site-packages\torch\utils\cpp_extension.py:2614, in _run_ninja_build(build_directory, verbose, error_prefix)
2612 if hasattr(error, 'output') and error.output: # type: ignore[union-attr]
2613 message += f": {error.output.decode(*SUBPROCESS_DECODE_ARGS)}" # type: ignore[union-attr]
-> 2614 raise RuntimeError(message) from e

RuntimeError: Error building extension 'warpscan'