Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

openvino: driver setup, CI split, thread safety, and NPU optimizations (#21944)

Browse Source 
* Thread safety per request only

* Fix ROPE yarn case

* Fix sticky stateful config

* Use i4/i8 directly for symmetric quant

* Use weightless caching

* Add WeightlessCacheAttribute to reduce NPU memory usage

* Gelu tanh support (#125)

* Imrope support (#126)

* fix(openvino): explicit ov::Tensor frees in ggml_backend_openvino_free

* add GPU,NPU support in OV Dockerfile

* add build-openvino.yml ci

* Fix sticky stateful config

* add concurrency to ov-gpu ci runs. Move OV CI to build-openvino.yml

* fix thread-safety of shared runtime context

* rope type abstraction for frontend translations

* fix editorconfig

---------

Co-authored-by: Mustafa Cavus <mustafa.cavus@intel.com>
Co-authored-by: Dan Hoffman <dhoff749@gmail.com>
Co-authored-by: Ravi Panchumarthy <ravi.panchumarthy@intel.com>
This commit is contained in:
Zijun Yu
2026-04-21 18:58:34 +03:00
committed by GitHub
co-authored by Mustafa Cavus Dan Hoffman Ravi Panchumarthy
parent 606fa42f5d
commit 52f1096f21
21 changed files with 820 additions and 541 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-openvino/utils.h
+23 -3
View File
@@ -3,12 +3,15 @@
#include "ggml-impl.h"
#include <algorithm>
#include <atomic>
#include <cstddef>
#include <memory>
#include <mutex>
#include <openvino/runtime/core.hpp>
#include <openvino/runtime/infer_request.hpp>
#include <string>
#include <unordered_map>
#include <utility>
#include <vector>
struct graph_key {
@@ -40,11 +43,17 @@ struct graph_key_hash {
}
};
struct decoder_runtime_ctx {
decoder_runtime_ctx(std::shared_ptr<std::mutex> mutex) : mutex(std::move(mutex)) {}
std::shared_ptr<std::mutex> mutex;
std::shared_ptr<GgmlOvDecoder> ptr;
};
struct ov_runtime_context {
std::mutex ov_compute_mutex;
mutable std::mutex ctx_mutex;
std::string device;
bool stateful;
std::unordered_map<graph_key, std::shared_ptr<GgmlOvDecoder>, graph_key_hash> decoder_cache;
std::unordered_map<graph_key, std::shared_ptr<decoder_runtime_ctx>, graph_key_hash> decoder_cache;
std::unordered_map<graph_key, std::shared_ptr<ov::InferRequest>, graph_key_hash> infer_request_cache;
std::unordered_map<graph_key, std::shared_ptr<ov::InferRequest>, graph_key_hash> infer_request_cache_prefill;
std::unordered_map<graph_key, std::vector<std::string>, graph_key_hash> ov_input_names_cache;
@@ -53,11 +62,22 @@ struct ov_runtime_context {
// Simultanous stateful inference request support to be added.
size_t stateful_kv_size;
std::map<std::string, std::string> kv_state_input_name_map;
std::atomic<int> backend_count;
ov_runtime_context() :
device("CPU"),
stateful(false),
stateful_kv_size(0) {}
stateful_kv_size(0),
backend_count(0) {}
void clear_caches() {
std::lock_guard<std::mutex> lock(ctx_mutex);
decoder_cache.clear();
infer_request_cache.clear();
infer_request_cache_prefill.clear();
ov_input_names_cache.clear();
ov_output_names_cache.clear();
}
};
enum ggml_status ov_graph_compute(struct ggml_cgraph * cgraph, ggml_backend_t backend);