fix ADM Constrant Violation Analysis

AMSS_NCKU_source/bssn_rhs_cuda.cu

@@ -4997,6 +4997,17 @@ static void download_state_outputs(double **state_host_out, size_t all)
     }
 }
 
+static void download_constraint_outputs(double **constraint_host_out, size_t all)
+{
+    const size_t bytes = all * sizeof(double);
+    CUDA_CHECK(cudaMemcpy(g_buf.h_stage, g_buf.slot[S_ham_Res],
+                          (size_t)D2H_CONSTRAINT_SLOT_COUNT * bytes,
+                          cudaMemcpyDeviceToHost));
+    for (int i = 0; i < D2H_CONSTRAINT_SLOT_COUNT; ++i) {
+        std::memcpy(constraint_host_out[i], g_buf.h_stage + (size_t)i * all, bytes);
+    }
+}
+
 __global__ void kern_pack_state_region_batch(const double * __restrict__ src_mem,
                                              double * __restrict__ dst,
                                              int nx, int ny,
@@ -5818,6 +5829,17 @@ int bssn_cuda_download_resident_state(void *block_tag,
     return 0;
 }
 
+extern "C"
+int bssn_cuda_download_constraint_outputs(int *ex,
+                                          double **constraint_host_out)
+{
+    init_gpu_dispatch();
+    CUDA_CHECK(cudaSetDevice(g_dispatch.my_device));
+    const size_t all = (size_t)ex[0] * ex[1] * ex[2];
+    download_constraint_outputs(constraint_host_out, all);
+    return 0;
+}
+
 extern "C"
 int bssn_cuda_pack_state_region_to_host_buffer(void *block_tag,
                                                int state_index,