AVX 10.2 and APX support

[stefanatwork]

Adding support for the AVX 10.2 and APX instruction sets

[Copilot]

Pull request overview

This PR extends Embree’s ISA detection, build configuration, and runtime dispatch to add a new APX (and related AVX10.2) codepath, including updates to CPU feature bitmask width and AVX10.2-specific SIMD helper intrinsics.

Changes:

• Widen CPU feature/ISA bitmasks and plumbing from int to int64_t across state/config/verify paths.
• Add APX target support to CMake (flags, target library, config export) and extend runtime symbol-selection macros to include APX.
• Extend CPU feature detection + SIMD helpers to recognize and use AVX10.2/APX-related capabilities.

Reviewed changes

Copilot reviewed 32 out of 32 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tutorials/verify/verify.h	Switch ISA storage/params to int64_t for wider feature masks.
tutorials/verify/verify.cpp	Use int64_t ISA masks and add APX to verify ISA list.
tutorials/common/tutorial/application.cpp	Document apx in CLI help output.
kernels/common/state.h	Update ISA-related APIs/fields to int64_t.
kernels/common/state.cpp	Update ISA handling and add APX debug verification + AVX10 string parsing.
kernels/common/rtcore.cpp	Extend geometry factory dispatch to include APX-capable symbol selection.
kernels/common/isa.h	Add APX namespace and new ISA selection macro combinations (AVX512+APX, etc.).
kernels/common/device.cpp	Print CPU features with int64_t and include APX in supported target list string.
kernels/CMakeLists.txt	Add embree_apx static library target and integrate APX into file selection.
kernels/bvh/bvh8_factory.h	Widen builder/intersector feature masks to int64_t.
kernels/bvh/bvh8_factory.cpp	Use APX-aware selection macros for BVH8 intersectors.
kernels/bvh/bvh4_factory.h	Widen builder/intersector feature masks to int64_t.
kernels/bvh/bvh4_factory.cpp	Use APX-aware selection macros for BVH4 intersectors.
kernels/bvh/bvh_intersector1_bvh4.cpp	Return int64_t from getISA() for wider ISA values.
common/sys/sysinfo.h	Add APX/AVX10.* feature bits and widen ISA constants/APIs to int64_t.
common/sys/sysinfo.cpp	Implement AVX10/APX CPUID + XCR0 detection and widen feature handling.
common/simd/vuint16_avx512.h	Use AVX10.2 reductions when available; adjust multiply intrinsic.
common/simd/vllong8_avx512.h	Use AVX10.2 reduction intrinsics when available.
common/simd/vfloat16_avx512.h	Use AVX10.2 abs/minmax/reduce helpers when available.
common/simd/vdouble8_avx512.h	Use AVX10.2 abs/minmax/reduce helpers when available.
common/simd/vboolf8_avx512.h	Use AVX10.2 mask intrinsics where applicable.
common/simd/vboolf4_avx512.h	Use AVX10.2 mask intrinsics where applicable.
common/simd/vboolf16_avx512.h	Use AVX10.2 mask intrinsics where applicable.
common/simd/vboold8_avx512.h	Use AVX10.2 mask intrinsics where applicable.
common/simd/vboold4_avx512.h	Use AVX10.2 mask intrinsics where applicable.
common/cmake/msvc.cmake	Add MSVC flags for APX target builds.
common/cmake/gnu.cmake	Add GNU flags for APX target builds.
common/cmake/embree-config.cmake	Export APX target setting and include APX targets for static builds.
common/cmake/dpcpp.cmake	Add DPCPP flags for APX target builds.
common/cmake/clang.cmake	Add Clang flags for APX target builds.
common/cmake/check_isa.cpp	Extend ISA probe to detect/report APX.
CMakeLists.txt	Add APX to max-ISA selection and enable APX build option/defines.

---

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

[svenwoop]

This is not sufficient one also needs to enable APX separately. Also we want 512-bit AVX10.2 not sure what that default actually does, but /arch:AVX10.2/512 should be used here, and /arch:APX

[svenwoop]

Why not just consistently define AVX10_2 if you anyway define some makro? Also you set version here to 102 but compare >= 2 in code.