Symmetry-based acceleration for HF and GW kernels and codebase cleanup

[gauravharsha]

Implements crystal symmetry exploitation in the GPU GW self-energy kernel, reducing the k/q loop workload from full-BZ to IBZ, and refactors the GPU source tree for maintainability.

Key changes:

• cu_symmetry: new class for device-side k-AO and q-P0 unitary transforms; IBZ→full-BZ rotation of G and P0 on GPU
• Low-memory scalar path: race-free IBZ G upload via dedicated stream + pinned staging buffer
• gw_qpt / gw_qkpt separated into own files; cu_routines replaced by cuhf_utils + cugw_utils
• Memory accounting in GW_check_devices_free_space updated for new allocations
• set_up_qkpt_first/second renamed to upload_p0_inputs/upload_sigma_inputs/upload_p0_coulomb

[Copilot]

Pull request overview

This PR introduces GPU-side exploitation of crystal symmetries for the GW self-energy (reducing full-BZ work to IBZ + star expansion) and performs a substantial CUDA codebase refactor by splitting the previous monolithic CUDA “routines” into focused HF/GW utility components plus a reusable cu_symmetry layer.

Changes:

• Add cu_symmetry + cu_symmetry_data to perform IBZ→full-BZ AO (k) and auxiliary (q/P0) unitary transforms on GPU, and update GW to loop over irreducible q-points (inq) rather than full q.
• Refactor GPU CUDA code: replace cu_routines.{h,cu} with cuhf_utils and cugw_utils, and split gw_qkpt into its own compilation unit.
• Add a CUDA-based symmetry transform roundtrip test and wire it into the test build.

Reviewed changes

Copilot reviewed 25 out of 38 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
test/cu_symmetry_test.h	Declares a GPU symmetry transform roundtrip test helper.
test/cu_symmetry_test.cu	Implements a GPU vs CPU reference symmetry transform validation routine.
test/cu_solver_test.cpp	Updates HDF5 key path for ink and adds a new symmetry transform test section.
test/CMakeLists.txt	Enables CUDA language in tests and compiles the new .cu test source.
src/hf_gpu_kernel.cpp	Switches HF exchange to cuhf_utils and updates direct term to use k_symmetry().value_AO/full-BZ looping.
src/gw_gpu_kernel.cpp	Switches to cugw_utils, builds cu_symmetry_data, moves GW outer distribution from ink to inq, and updates memory accounting.
src/green/gpu/hf_gpu_kernel.h	Updates brillouin_zone_utils alias usage.
src/green/gpu/gw_gpu_kernel.h	Updates brillouin_zone_utils alias usage and adjusts X2C copy helper signatures.
src/green/gpu/gpu_kernel.h	Adds q-mesh sizes (_nq, _inq) and switches k-pair counting to k_symmetry().
src/green/gpu/gpu_factory.h	Updates brillouin_zone_utils alias usage.
src/green/gpu/df_integral_t.h	Switches k-pair symmetry bookkeeping from symmetry() to k_symmetry().
src/green/gpu/cuhf_utils.h	New HF CUDA utility class header (replacement for HF parts of cu_routines).
src/green/gpu/cugw_utils.h	New GW CUDA utility class header integrating cu_symmetry and new callbacks.
src/green/gpu/cugw_qpt.h	Extends q-worker interface (constructor signature change + debug dump API).
src/green/gpu/cugw_qkpt.h	New per-(q,k) worker header separated from cugw_qpt.h.
src/green/gpu/cu_symmetry.h	New symmetry data container and GPU symmetry transform interface.
src/green/gpu/cu_routines.h	Removes the previous monolithic CUDA routines header.
src/green/gpu/cu_hf_utils.h	Leaves an empty placeholder file indicating a rename.
src/cuhf_utils.cu	New implementation of HF CUDA utilities (exchange accumulation).
src/cugw_utils.cu	New implementation of GW CUDA utilities (IBZ upload path + symmetry transforms + q-IBZ loop).
src/cugw_qpt.cu	Updates gw_qpt constructor signature and adds the P0 diagonal debug dump implementation.
src/cugw_qkpt.cu	New implementation of gw_qkpt worker split from cugw_qpt.cu.
src/cu_symmetry.cu	New implementation of GPU symmetry transforms and device metadata uploads.
src/cu_routines.cu	Removes the previous monolithic CUDA routines implementation.
src/CMakeLists.txt	Updates CUDA source list and links GREEN::SYMMETRY for the refactor.

---

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

[egull]

I have some minor comment which we can skip for now. There's way too much work in this to really analyze it all without testing how it works in practice. So I propose to merge and then clean up if we see that something doesn't work.
I think none of the math operations on the GPU are perforamnce critical. If they are, we can replace the hand-made multiplications with cublas calls.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 96.96970% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.32%. Comparing base (fae769b) to head (378e813).

Files with missing lines	Patch %	Lines
src/gw_gpu_kernel.cpp	95.76%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #12      +/-   ##
==========================================
+ Coverage   95.19%   95.32%   +0.13%     
==========================================
  Files          12       13       +1     
  Lines         832      877      +45     
==========================================
+ Hits          792      836      +44     
- Misses         40       41       +1     

Flag	Coverage Δ
unittests	95.32% <96.96%> (+0.13%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[gauravharsha]

Merging -- with the tests passing and a cleaner code, I feel we are ready for a 1.0.0a release for GREEN.