From 28aa8dfe00a2e0daa69ab43c8bd97ed37e5aa0b9 Mon Sep 17 00:00:00 2001
From: ristvan <ristvan@ss01-gh01.cm.cluster>
Date: Wed, 20 Dec 2023 14:09:11 -0800
Subject: [PATCH] int/size_t mismatches, nvhpc with omp offload

---
 makefiles/compilers/c/nvhpc.mk            | 2 +-
 op2/src/cuda/op_cuda_rt_support.cpp       | 4 ++--
 op2/src/openmp4/op_openmp4_decl.cpp       | 4 ++--
 op2/src/openmp4/op_openmp4_rt_support.cpp | 9 +++++----
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/makefiles/compilers/c/nvhpc.mk b/makefiles/compilers/c/nvhpc.mk
index 0245eda5c..68a7af3fc 100644
--- a/makefiles/compilers/c/nvhpc.mk
+++ b/makefiles/compilers/c/nvhpc.mk
@@ -20,5 +20,5 @@ CONFIG_CPP_HAS_OMP ?= true
 GPU_FFLAG := -gpu=fastmath,ptxinfo,lineinfo
 $(foreach arch,$(CUDA_GEN),$(eval GPU_FFLAG := $(GPU_FFLAG),cc$(arch)))
 
-CONFIG_OMP_OFFLOAD_CPPFLAGS ?= -mp=gpu $(GPU_FFLAG)
+CONFIG_OMP_OFFLOAD_CXXFLAGS ?= -mp=gpu $(GPU_FFLAG)
 CONFIG_CPP_HAS_OMP_OFFLOAD ?= true
diff --git a/op2/src/cuda/op_cuda_rt_support.cpp b/op2/src/cuda/op_cuda_rt_support.cpp
index 52414e54d..673d353c1 100644
--- a/op2/src/cuda/op_cuda_rt_support.cpp
+++ b/op2/src/cuda/op_cuda_rt_support.cpp
@@ -94,7 +94,7 @@ void __cutilCheckMsg(const char *errorMessage, const char *file,
 // routines to move arrays to/from GPU device
 //
 
-void op_mvHostToDevice(void **map, int size) {
+void op_mvHostToDevice(void **map, size_t size) {
   if (!OP_hybrid_gpu || size == 0)
     return;
   void *tmp;
@@ -105,7 +105,7 @@ void op_mvHostToDevice(void **map, int size) {
   *map = tmp;
 }
 
-void op_cpHostToDevice(void **data_d, void **data_h, int size) {
+void op_cpHostToDevice(void **data_d, void **data_h, size_t size) {
   if (!OP_hybrid_gpu)
     return;
   if (*data_d != NULL) cutilSafeCall(cudaFree(*data_d));
diff --git a/op2/src/openmp4/op_openmp4_decl.cpp b/op2/src/openmp4/op_openmp4_decl.cpp
index da8048c12..d8ef6980f 100644
--- a/op2/src/openmp4/op_openmp4_decl.cpp
+++ b/op2/src/openmp4/op_openmp4_decl.cpp
@@ -9,9 +9,9 @@
 
 void cutilDeviceInit(int argc, char **argv);
 
-void op_mvHostToDevice(void **map, int size);
+void op_mvHostToDevice(void **map, size_t size);
 
-void op_cpHostToDevice(void **data_d, void **data_h, int size);
+void op_cpHostToDevice(void **data_d, void **data_h, size_t size);
 
 void op_cuda_exit();
 
diff --git a/op2/src/openmp4/op_openmp4_rt_support.cpp b/op2/src/openmp4/op_openmp4_rt_support.cpp
index 078484915..32a59f7ad 100644
--- a/op2/src/openmp4/op_openmp4_rt_support.cpp
+++ b/op2/src/openmp4/op_openmp4_rt_support.cpp
@@ -19,7 +19,7 @@
 // routines to move arrays to/from GPU device
 //
 
-void op_mvHostToDevice(void **map, int size) {
+void op_mvHostToDevice(void **map, size_t size) {
   if (!OP_hybrid_gpu)
     return;
   char *temp = (char*)*map;
@@ -28,7 +28,7 @@ void op_mvHostToDevice(void **map, int size) {
   //TODO test
 }
 
-void op_cpHostToDevice(void **data_d, void **data_h, int size) {
+void op_cpHostToDevice(void **data_d, void **data_h, size_t size) {
   if (!OP_hybrid_gpu)
     return;
   *data_d = (char*)op_malloc(size);
@@ -121,8 +121,9 @@ void op_cuda_exit() {
     return;
   op_dat_entry *item;
   TAILQ_FOREACH(item, &OP_dat_list, entries) {
-    #pragma omp target exit data map(from: (item->dat)->data_d)
-    free((item->dat)->data_d);
+    char *data_d = (item->dat)->data_d;
+    #pragma omp target exit data map(from: data_d)
+    free(data_d);
   }
   /*
   for (int ip = 0; ip < OP_plan_index; ip++) {