[Bugfix]: Compile on CPU backend

[daniellegillai]

Motivation

Fixes #498: compile successfully when rocAL is built for CPU backend.

Technical Details

• tensor.h: Remove #else branch on #if ENABLE_HIP that included non-existent device/device_manager.h
• node_color_cast.cpp, node_lut.cpp, node_slice.cpp: Guard VX_MEMORY_TYPE_HIP with #if ENABLE_HIP
• node_lut.cpp, node_python_function.cpp: Add missing #include <vx_ext_amd.h> header
• master_graph.cpp: Guard _box_encoder_gpu deletion with #if ENABLE_HIP since this object only exists in GPU builds

Test Plan

CTests on both the CPU (-D BACKEND=CPU) and HIP (-D BACKEND=HIP) build. Note that tested on the MIVisionX and RPP branches which fix the CPU builds on those repos as well, rather than TOT (RPP #10295 and MIVisionX #1726).

Test Result

On CPU build, cpu ctests pass:

    Start  1: basic_test_cpu
1/6 Test  #1: basic_test_cpu .....................   Passed    1.07 sec
    Start  5: dataloader_multithread_cpu
2/6 Test  #5: dataloader_multithread_cpu .........   Passed    2.07 sec
    Start  7: performance_tests_cpu
3/6 Test  #7: performance_tests_cpu ..............   Passed    1.03 sec
    Start  9: performance_tests_with_depth_cpu
4/6 Test  #9: performance_tests_with_depth_cpu ...   Passed    1.10 sec
    Start 11: unit_tests_cpu
5/6 Test #11: unit_tests_cpu .....................   Passed    9.25 sec
    Start 18: serialization_test_cpu
6/6 Test #18: serialization_test_cpu .............   Passed    0.93 sec

On HIP build, all ctests still pass:

      Start  1: basic_test_cpu
 1/21 Test  #1: basic_test_cpu .....................   Passed    1.46 sec
      Start  2: basic_test_gpu
 2/21 Test  #2: basic_test_gpu .....................   Passed    0.26 sec
      Start  3: basic_test_gray
 3/21 Test  #3: basic_test_gray ....................   Passed    0.21 sec
      Start  4: basic_test_rgb
 4/21 Test  #4: basic_test_rgb .....................   Passed    0.25 sec
      Start  5: dataloader_multithread_cpu
 5/21 Test  #5: dataloader_multithread_cpu .........   Passed    2.09 sec
      Start  6: dataloader_multithread_gpu
 6/21 Test  #6: dataloader_multithread_gpu .........   Passed    2.44 sec
      Start  7: performance_tests_cpu
 7/21 Test  #7: performance_tests_cpu ..............   Passed    1.04 sec
      Start  8: performance_tests_gpu
 8/21 Test  #8: performance_tests_gpu ..............   Passed    0.27 sec
      Start  9: performance_tests_with_depth_cpu
 9/21 Test  #9: performance_tests_with_depth_cpu ...   Passed    1.13 sec
      Start 10: performance_tests_with_depth_gpu
10/21 Test #10: performance_tests_with_depth_gpu ...   Passed    0.38 sec
      Start 11: unit_tests_cpu
11/21 Test #11: unit_tests_cpu .....................   Passed   11.30 sec
      Start 12: unit_tests_gpu
12/21 Test #12: unit_tests_gpu .....................   Passed    8.26 sec
      Start 13: unit_tests_gray
13/21 Test #13: unit_tests_gray ....................   Passed    3.47 sec
      Start 14: video_tests_ffmpeg
14/21 Test #14: video_tests_ffmpeg .................   Passed   46.46 sec
      Start 15: video_tests_rocdecode
15/21 Test #15: video_tests_rocdecode ..............   Passed   11.16 sec
      Start 16: basic_test_rocjpeg_gray
16/21 Test #16: basic_test_rocjpeg_gray ............   Passed    0.30 sec
      Start 17: basic_test_rocjpeg_rgb
17/21 Test #17: basic_test_rocjpeg_rgb .............   Passed    0.30 sec
      Start 18: serialization_test_cpu
18/21 Test #18: serialization_test_cpu .............   Passed    0.91 sec
      Start 19: serialization_test_gpu
19/21 Test #19: serialization_test_gpu .............   Passed    0.21 sec
      Start 20: checkpoint_save
20/21 Test #20: checkpoint_save ....................   Passed    1.82 sec
      Start 21: checkpoint_restore
21/21 Test #21: checkpoint_restore .................   Passed    2.07 sec

100% tests passed, 0 tests failed out of 21

[Copilot]

Pull request overview

Fixes CPU-backend compilation by removing/guarding HIP-specific code paths and ensuring the correct OpenVX extension headers are included where needed.

Changes:

• Remove a non-existent CPU-side device/device_manager.h include from tensor.h (HIP-only include remains guarded).
• Guard VX_MEMORY_TYPE_HIP usage and GPU-only cleanup logic behind #if ENABLE_HIP.
• Add missing #include <vx_ext_amd.h> to compilation units that require AMD OpenVX extension definitions.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
rocAL/source/pipeline/master_graph.cpp	Guards _box_encoder_gpu deletion behind ENABLE_HIP to avoid CPU build breakage.
rocAL/source/augmentations/node_python_function.cpp	Adds missing vx_ext_amd.h include for AMD OpenVX extension symbols.
rocAL/source/augmentations/geometry_augmentations/node_slice.cpp	Guards VX_MEMORY_TYPE_HIP usage behind ENABLE_HIP.
rocAL/source/augmentations/effects_augmentations/node_lut.cpp	Adds vx_ext_amd.h and guards VX_MEMORY_TYPE_HIP usage behind ENABLE_HIP.
rocAL/source/augmentations/color_augmentations/node_color_cast.cpp	Guards VX_MEMORY_TYPE_HIP usage behind ENABLE_HIP.
rocAL/include/pipeline/tensor.h	Removes invalid non-HIP include; keeps HIP includes under ENABLE_HIP.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.