llama: llama_kv_cache_state was removed, only the result of llama_graph_compute is returned

Author: Xarbirus

src/llama.cpp

@@ -2815,42 +2815,6 @@ struct llama_kv_cache {
     }
 };
 
-class llama_kv_cache_state {
-    struct llama_kv_cache_state_short {
-        uint32_t head = 0;
-        uint32_t size = 0;
-        uint32_t used = 0;
-        uint32_t n    = 0;
-
-        std::vector<llama_kv_cell> cells;
-    } old_state;
-
-    bool saved = false;
-
-public:
-    void save_state(const llama_kv_cache& cache) {
-        old_state.head  = cache.head;
-        old_state.size  = cache.size;
-        old_state.used  = cache.used;
-        old_state.n     = cache.n;
-        old_state.cells = cache.cells;
-
-        saved = true;
-    }
-
-    void restore(llama_kv_cache& cache) {
-        if (saved) {
-            cache.head  = old_state.head;
-            cache.size  = old_state.size;
-            cache.used  = old_state.used;
-            cache.n     = old_state.n;
-            cache.cells = std::move(old_state.cells);
-
-            saved = false;
-        }
-    }
-};
-
 struct llama_control_vector {
     std::vector<struct ggml_tensor *> tensors; // per layer
     std::vector<struct ggml_context *> ctxs;
@@ -17220,7 +17184,6 @@ static int llama_decode_internal(
     lctx.n_queued_tokens += n_tokens_all;
 
     auto & kv_self = lctx.kv_self;
-    llama_kv_cache_state kv_cache_state_holder;
 
     const int64_t n_embd  = hparams.n_embd;
     const int64_t n_vocab = hparams.n_vocab;
@@ -17298,7 +17261,6 @@ static int llama_decode_internal(
         // non-causal masks do not use the KV cache
         if (hparams.causal_attn) {
             llama_kv_cache_update(&lctx);
-            kv_cache_state_holder.save_state(kv_self);
 
             // if we have enough unused cells before the current head ->
             //   better to start searching from the beginning of the cache, hoping to fill it
@@ -17356,17 +17318,16 @@ static int llama_decode_internal(
         llama_set_inputs(lctx, ubatch);
 
         const auto compute_status = llama_graph_compute(lctx, gf, n_threads, threadpool);
-        if (compute_status != GGML_STATUS_SUCCESS) {
-            kv_cache_state_holder.restore(kv_self);
-            switch (compute_status) {
-                case GGML_STATUS_ABORTED:
-                    return 2;
-                case GGML_STATUS_ALLOC_FAILED:
-                    return -2;
-                case GGML_STATUS_FAILED:
-                default:
-                    return -3;
-            }
+        switch (compute_status) {
+            case GGML_STATUS_SUCCESS:
+                break;
+            case GGML_STATUS_ABORTED:
+                return 2;
+            case GGML_STATUS_ALLOC_FAILED:
+                return -2;
+            case GGML_STATUS_FAILED:
+            default:
+                return -3;
         }
 
         // update the kv ring buffer