define ice diag_ functions for multi_CUDA backend

[pdziekan]

No description provided.

[Copilot]

Pull request overview

This PR extends the multi_CUDA backend diagnostics surface to match existing ice-related diagnostics available in the single-GPU (CUDA) backend, by adding the missing wrapper methods that dispatch via pimpl->mcuda_run(...).

Changes:

• Added multi_CUDA wrappers for ice shape range-selection diagnostics (diag_ice_a_rng, diag_ice_c_rng) and their consecutive-selection variants.
• Added multi_CUDA wrappers for ice shape moment diagnostics (diag_ice_a_mom, diag_ice_c_mom) and ice mixing ratio (diag_ice_mix_ratio).
• Added a multi_CUDA wrapper for diag_precip_rate_ice_mass.

---

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

[Copilot]

In these ice selection helpers, the parameters are named r_mi/r_mx, but they represent ice shape radii ("a" equatorial and "c" polar). Renaming them (e.g., a_mi/a_mx and c_mi/c_mx) would make the intent clearer and match the naming used in the public declarations in particles.hpp and the single-GPU implementation.

Suggested change

	const real_t &r_mi, const real_t &r_mx
	)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_a_rng, r_mi, r_mx);
	}
	template <typename real_t>
	void particles_t<real_t, multi_CUDA>::diag_ice_c_rng(
	const real_t &r_mi, const real_t &r_mx
	)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_c_rng, r_mi, r_mx);
	const real_t &a_mi, const real_t &a_mx
	)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_a_rng, a_mi, a_mx);
	}
	template <typename real_t>
	void particles_t<real_t, multi_CUDA>::diag_ice_c_rng(
	const real_t &c_mi, const real_t &c_mx
	)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_c_rng, c_mi, c_mx);

[Copilot]

The moment diagnostics are implemented in c-then-a order here, while the rest of the API and the single-GPU implementation use a-then-c ordering. Reordering these two wrappers to keep diag_ice_a_mom adjacent and before diag_ice_c_mom would make the file easier to scan and keep the multi-GPU backend consistent with other backends.

Suggested change

	void particles_t<real_t, multi_CUDA>::diag_ice_c_mom(const int &k)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_c_mom, k);
	}
	template <typename real_t>
	void particles_t<real_t, multi_CUDA>::diag_ice_a_mom(const int &k)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_a_mom, k);
	void particles_t<real_t, multi_CUDA>::diag_ice_a_mom(const int &k)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_a_mom, k);
	}
	template <typename real_t>
	void particles_t<real_t, multi_CUDA>::diag_ice_c_mom(const int &k)
	{
	pimpl->mcuda_run(&particles_t<real_t, CUDA>::diag_ice_c_mom, k);

[Copilot]

New multi-GPU wrappers are added for ice-related diagnostics (e.g., diag_ice_mix_ratio, diag_precip_rate_ice_mass, and the new diag_ice_[ac]_rng[_cons] / diag_ice_[ac]_mom). There don't appear to be any tests exercising these diagnostics on the multi_CUDA backend; consider adding a GPU-optional test (skip when multi_CUDA is unavailable) that enables ice_switch and smoke-tests these calls and their basic output invariants (e.g., non-negative, non-NaN).