fix compile for CUDA 13

[NathanMOlson]

Description

This PR includes the changes that allow PopSift to compile on Ubuntu 24.04 with CUDA 13.0.0.

Features list

• Remove #include <texture_fetch_functions.h>. This include files is no longer included with the latest CUDA Toolkit, and appears to be unused by PopSift.
• Migrate from thrust::identity to cuda::std::identity (required for Thrust version >= 3.0.0).
• Fix macro conflict with nvtxRangePop() (It looks like this fix will be overcome by First-order CUDA follow-up fix: do not use NVTX #162 or Guard calls to nvtx directly #168).

Implementation remarks

Re: 'ntvxRangePushA(a)' and ntvxRangePop(): The definition of these above #include <thrust/copy.h> wreaks havoc on the definition of these functions in nvtx3.hpp include indirectly via thrust/copy.h. I changed their name for this file, but another approach could be to move the #define blow the thrust includes. It seems like it was probably placed above the thrust includes for a reason though. The change to these macros will be overcome by #162 or #168.

In file included from /usr/local/cuda/targets/x86_64-linux/include/nvtx3/nvtx3.hpp:667,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/cub/detail/nvtx.cuh:56,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/cub/device/device_for.cuh:40,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/cuda/detail/parallel_for.h:43,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/cuda/detail/swap_ranges.h:43,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/detail/adl/swap_ranges.h:50,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/swap_ranges.inl:31,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/swap.h:147,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/cuda/detail/iter_swap.h:34,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/detail/adl/iter_swap.h:50,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/reference.h:37,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/memory.h:35,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/allocator/temporary_allocator.h:31,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/temporary_array.h:46,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/cuda/detail/internal/copy_cross_system.h:47,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/cuda/detail/copy.h:79,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/system/detail/adl/copy.h:50,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/copy.inl:29,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/detail/copy.h:72,
                 from /usr/local/cuda-13.0/targets/x86_64-linux/include/cccl/thrust/copy.h:505,
                 from /code/pypopsift/build/_deps/popsift-src/src/popsift/s_filtergrid.cu:21:
/usr/local/cuda/targets/x86_64-linux/include/nvtx3/nvToolsExt.h:1099:45: error: macro "nvtxRangePop" passed 1 arguments, but takes just 0
 1099 | NVTX_DECLSPEC int NVTX_API nvtxRangePop(void);
      |                                             ^
/code/pypopsift/build/_deps/popsift-src/src/popsift/s_filtergrid.cu:16: note: macro "nvtxRangePop" defined here
   16 | #define nvtxRangePop()
      |
In file included from /usr/local/cuda/targets/x86_64-linux/include/nvtx3/nvtxDetail/nvtxImpl.h:381,
                 from /usr/local/cuda/targets/x86_64-linux/include/nvtx3/nvToolsExt.h:1663:
/usr/local/cuda/targets/x86_64-linux/include/nvtx3/nvtxDetail/nvtxImplCore.h:163:45: error: macro "nvtxRangePop" passed 1 arguments, but takes just 0
  163 | NVTX_DECLSPEC int NVTX_API nvtxRangePop(void)
      |                                             ^
/code/pypopsift/build/_deps/popsift-src/src/popsift/s_filtergrid.cu:16: note: macro "nvtxRangePop" defined here
   16 | #define nvtxRangePop()
      |

[Copilot]

Pull Request Overview

This PR fixes compilation issues with CUDA 13.0.0 on Ubuntu 24.04 by addressing three specific compatibility problems in the PopSift codebase.

• Removes obsolete texture_fetch_functions.h include that is no longer available in CUDA 13
• Migrates from deprecated thrust::identity to cuda::std::identity for Thrust >= 3.0.0
• Resolves macro conflicts with NVTX functions by renaming local macro definitions

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/popsift/s_extrema.cu	Removes deprecated texture_fetch_functions.h include
src/popsift/s_filtergrid.cu	Adds Thrust version compatibility and fixes NVTX macro conflicts

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[simogasp]

Thanks for the PR!
I think we will first merge #162 and then this one

[NathanMOlson]

Great! I've removed the parts that conflict with and are overcome by #162.

[simogasp]

Merged! Thanks for your contribution!