Enable Z4C + Shell-Patch GPU coexistence (Phase 3)

Remove the compile-time #error that blocked USE_CUDA_Z4C + WithShell.
Add GPU-to-CPU state sync at the start of both Z4C Step functions
(non-CPBC and CPBC) so shell CPU consumers read valid field data
after Cartesian GPU RHS with resident state.

Move bssn_cuda_use_resident_sync and bssn_cuda_download_level_state
_if_present from anonymous namespace to file scope in bssn_class.C
so derived classes (Z4C) can call them. Declare both in
bssn_rhs_cuda.h. Include bssn_rhs_cuda.h in Z4c_class.C.

Z4C shell RHS remains on CPU (Fortran Z4c_rhs_ss.f90) pending
future GPU kernel implementation.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

AMSS_NCKU_source/Z4c_class.C

@@ -36,6 +36,9 @@ using namespace std;
 #if USE_CUDA_Z4C && (ABEtype == 2)
 #include "z4c_rhs_cuda.h"
 #endif
+#if USE_CUDA_BSSN
+#include "bssn_rhs_cuda.h"
+#endif
 
 #ifdef With_AHF
 #include "derivatives.h"
@@ -187,9 +190,6 @@ Z4c_class::~Z4c_class()
 // for sommerfeld boundary
 
 #if USE_CUDA_Z4C && (ABEtype == 2)
-#ifdef WithShell
-#error "USE_CUDA_Z4C resident path currently supports Cartesian non-shell Z4C only"
-#endif
 #if (MRBD == 2)
 #error "USE_CUDA_Z4C resident path does not support MRBD == 2"
 #endif
@@ -490,6 +490,14 @@ void Z4c_class::Step(int lev, int YN)
   int pre = 0, cor = 1;
   int ERROR = 0;
 
+#ifdef WithShell
+  if (bssn_cuda_use_resident_sync(lev))
+  {
+    for (int dl = 0; dl < GH->levels; dl++)
+      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
+  }
+#endif
+
   MyList<Patch> *Pp = GH->PatL[lev];
   while (Pp)
   {
@@ -915,6 +923,13 @@ void Z4c_class::Step(int lev, int YN)
   }
 
 #ifdef WithShell
+#if USE_CUDA_Z4C
+  if (bssn_cuda_use_resident_sync(lev))
+  {
+    for (int dl = 0; dl < GH->levels; dl++)
+      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
+  }
+#endif
   // evolve Shell Patches
   if (lev == 0)
   {
@@ -1622,9 +1637,7 @@ void Z4c_class::Step(int lev, int YN)
 }
 #else
 // for constraint preserving boundary (CPBC)
-#if USE_CUDA_Z4C && (ABEtype == 2)
+// Note: CPBC path uses CPU Fortran RHS; GPU resident sync is a no-op here.
-#error "USE_CUDA_Z4C resident path does not support CPBC"
-#endif
 #ifndef WithShell
 #error "CPBC only supports Shell"
 #endif
@@ -1654,6 +1667,14 @@ void Z4c_class::Step(int lev, int YN)
   int pre = 0, cor = 1;
   int ERROR = 0;
 
+#if USE_CUDA_Z4C && defined(WithShell)
+  if (bssn_cuda_use_resident_sync(lev))
+  {
+    for (int dl = 0; dl < GH->levels; dl++)
+      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
+  }
+#endif
+
   MyList<ss_patch> *sPp;
   // Predictor
   MyList<Patch> *Pp = GH->PatL[lev];

AMSS_NCKU_source/bssn_class.C

@@ -548,6 +548,8 @@ bool fill_bssn_cuda_views_count(Block *cg, MyList<var> *vars,
   return idx == state_count && vars == 0;
 }
 
+} // namespace
+
 bool bssn_cuda_use_resident_sync(int lev)
 {
   (void)lev;
@@ -1032,7 +1034,6 @@ void bssn_cuda_sync_level_bh_fields(MyList<Patch> *PatL,
   }
 }
 
-} // namespace
 #endif
 
 #if !USE_CUDA_BSSN

AMSS_NCKU_source/bssn_rhs_cuda.h

@@ -404,6 +404,10 @@ void bssn_cuda_release_step_ctx(void *block_tag);
 
 #ifdef __cplusplus
 }
+// C++-only helpers declared for derived equation classes (Z4C, etc.)
+// Defined in bssn_class.C. Requires MyList, Patch, var from including TU.
+bool bssn_cuda_use_resident_sync(int lev);
+void bssn_cuda_download_level_state_if_present(MyList<Patch> *PatL, MyList<var> *vars, int myrank);
 #endif
 
 #endif