RPC: add -sm tensor (llama/26610)

* rpc: allow -sm tensor

* fix flush for apple rdma

* move graph_uids to rpc_dispatcher

* cont : fix conflict

* cont: stop spinning dispatcher thread

* remove meta backend change

* add TODO to simplify logic

* rpc: bump major version

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

ggml/include/ggml-rpc.h

@@ -6,7 +6,7 @@
 extern "C" {
 #endif
 
-#define RPC_PROTO_MAJOR_VERSION    7
+#define RPC_PROTO_MAJOR_VERSION    8
 #define RPC_PROTO_MINOR_VERSION    0
 #define RPC_PROTO_PATCH_VERSION    0
 

ggml/src/ggml-backend-meta.cpp

@@ -869,7 +869,12 @@ static struct ggml_backend_meta_split_state ggml_backend_meta_get_split_state(
         ggml_backend_meta_split_state split_state;
         switch (tensor->op) {
             case GGML_OP_NONE: {
-                split_state = {GGML_BACKEND_SPLIT_AXIS_MIRRORED, {0}, {1}, 1};
+                if (tensor->view_src != nullptr) {
+                    // full-tensor view created with ggml_view_tensor, transparent for the split state
+                    split_state = ggml_backend_meta_get_split_state(stc, tensor->view_src, assume_sync);
+                } else {
+                    split_state = {GGML_BACKEND_SPLIT_AXIS_MIRRORED, {0}, {1}, 1};
+                }
             } break;
             case GGML_OP_DUP: {
                 split_state = handle_generic(src_ss, /*scalar_only =*/ true);