fix(tree): avoid sign comparison issues

stack-info: PR: #575, branch: aws-nslick/stack/22

src/nccl_ofi_rdma.c

@@ -4384,7 +4384,7 @@ static int send_progress(nccl_net_ofi_rdma_req_t *req)
 
 			ret = post_rdma_eager_send(req, comm_rail, xfer_info);
 		} else {
-			for (int rail_it = send_data->xferred_rail_id;
+			for (size_t rail_it = send_data->xferred_rail_id;
 			     rail_it < schedule->num_xfer_infos; rail_it++) {
 				/* Get xfer information from the schedule */
 				nccl_net_ofi_xfer_info_t *xfer_info = &xfers[rail_it];
@@ -4693,7 +4693,7 @@ static int send(nccl_net_ofi_send_comm_t *send_comm, void *data, int size, int t
 
 	/* Determine if this should be sent eagerly. */
 	eager = false;
-	if ((!have_ctrl && size <= eager_max_size) ||
+	if ((!have_ctrl && (size_t)size <= eager_max_size) ||
 		 (size == 0)) {
 		eager = true;
 	}
@@ -5100,7 +5100,7 @@ static inline int create_send_comm(nccl_net_ofi_conn_handle_t *handle,
 
 
  error:
-	if (ret_s_comm && ~0 != ret_s_comm->local_comm_id) {
+	if (ret_s_comm && COMM_ID_MASK != ret_s_comm->local_comm_id) {
 		if (0 != nccl_ofi_idpool_free_id(device->comm_idpool, ret_s_comm->local_comm_id)) {
 			NCCL_OFI_WARN("Error freeing communicator ID %" PRIu32, ret_s_comm->local_comm_id);
 		}

src/nccl_ofi_scheduler.c

@@ -58,7 +58,7 @@ void nccl_net_ofi_set_multiplexing_schedule(size_t size, int num_rails,
  */
 static inline int set_round_robin_schedule(nccl_net_ofi_threshold_scheduler_t *scheduler,
 					   size_t size,
-					   int num_rails,
+					   size_t num_rails,
 					   nccl_net_ofi_schedule_t *schedule)
 {
 	int rail_id;

src/nccl_ofi_sendrecv.c

@@ -2567,7 +2567,7 @@ int nccl_net_ofi_sendrecv_init(const char *provider_filter,
 			       nccl_net_ofi_plugin_t **plugin_p)
 {
 	int ret = 0;
-	int dev_id = 0;
+	size_t dev_id = 0;
 	struct fi_info *provider_list = NULL, *info;
 	unsigned int num_providers;
 	nccl_net_ofi_plugin_t *plugin = NULL;

src/platform-aws.c

@@ -550,7 +550,7 @@ int platform_init(const char **provider_filter)
 
 		ret = snprintf(topology_path, sizeof(topology_path), "%s/%s",
 			       XML_DIR, platform_data->topology);
-		if (ret < 0 || ret >= sizeof(topology_path)) {
+		if (ret < 0 || (size_t)ret >= sizeof(topology_path)) {
 			NCCL_OFI_WARN("Error occurred while forming the complete topology XML file path. RC: %d, Buffer Size: %d, XML dir: %s, Topology file: %s",
 				      ret, PATH_MAX, XML_DIR, platform_data->topology);
 			ret = -ENOMEM;

src/tuner/nccl_ofi_regions.c

@@ -140,7 +140,7 @@ int is_inside_region(nccl_ofi_tuner_point_t point, nccl_ofi_tuner_region_t *regi
 {
 	assert(region->num_vertices > 1);
 
-	int i, k;
+	size_t i, k;
 	nccl_ofi_tuner_point_t *pv;
 	double min_x, max_x, min_y, max_y;
 	const double eps = 1e-10;

src/tuner/nccl_ofi_tuner.c

@@ -347,7 +347,7 @@ ncclResult_t nccl_ofi_tuner_get_coll_info(void *context,
 	nccl_ofi_tuner_point_t p = {.x = (double)nBytes, .y = (double)nccl_ofi_tuner_ctx->dims.num_ranks};
 
 	/* Check all regions */
-	for (int i = 0; i < nccl_ofi_tuner_ctx->num_regions && in_out < 0; i++) {
+	for (size_t i = 0; i < nccl_ofi_tuner_ctx->num_regions && in_out < 0; i++) {
 		algorithm = nccl_ofi_tuner_ctx->regions[i].algorithm;
 		protocol = nccl_ofi_tuner_ctx->regions[i].protocol;
 		if (table[algorithm][protocol] == NCCL_ALGO_PROTO_IGNORE || algorithm >= numAlgo ||
@@ -412,7 +412,7 @@ ncclResult_t nccl_ofi_tuner_get_coll_info_v2(
 	nccl_ofi_tuner_point_t p = {.x = (double)nBytes, .y = (double)nccl_ofi_tuner_ctx->dims.num_ranks};
 
 	/* Check all regions */
-	for (int i = 0; i < nccl_ofi_tuner_ctx->num_regions && in_out < 0; i++) {
+	for (size_t i = 0; i < nccl_ofi_tuner_ctx->num_regions && in_out < 0; i++) {
 		if (nccl_ofi_tuner_ctx->regions[i].algorithm == NCCL_ALGO_NVLS_TREE && nvlsSupport == 0) {
 			continue;
 		}