…g_src to recorde the src ggml tensor for OpenVINO dynamic shape infer

…utput_info

Clean node information in OpenVINO backend

…issue

Enable zero-size copy for OpenVINO backend view

enable qwen35

Fix after rebase

remove logging

…t reason: the backend test initializes unary op inputs over a wide range, [-150, 150]. For FP32, exp(x) overflows around x ~= 88.7, so this test can randomly generate values right in or beyond the overflow region

In stateful mode the NEOX RoPE branch fed rank-3 data ([S, n_heads,
head_size]) into the Multiply against the rank-4 cos/sin tables
([1, S, 1, n_dims/2]). That mixed-rank broadcast is miscomputed by the
OpenVINO GPU plugin, corrupting the rotated Q/K and producing garbage
output (e.g. Phi-3-mini). Lift the data to rank-4 before the split/
Multiply so the operands are equal-rank, matching what the TYPE_NORMAL
branch already does. CPU and stateless paths are unaffected.

Phi-3-mini-Q4_K_M, wiki.test perplexity, GPU stateful:
  before: PPL = 27120.43
  after:  PPL = 6.2263   (CPU reference: 6.2251)

…ov name in ov bk; 3) fix issue in arch test & op test with latest code update

Update OpenVINO backend: remove unique name, add new OV name, fix tests

…s-in-llama

OpenVINO Backenb: remove changes in llama.cpp

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Doc change (use x64 Native Tools Command Prompt for VS)

…stances

get_tensor_ov_name() disambiguates COMPUTE/kvcache tensors that share a ggml name
by appending a per-graph suffix. It used ggml_hash_find(visited_hash_set, tensor),
whose value is a hash-set slot derived from the tensor POINTER and therefore differs
between cgraph instances.

The decoder / infer-request cache (ov_runtime_context) is process-global and keyed by
graph_key (node names + count). When several contexts or several same-architecture
models run (e.g. test-thread-safety, which builds 3 models x 4 contexts), they share
one cache under the same graph_key. On a cache hit, the OV param names captured at
compile time (e.g. "cache_k_l0#975") are looked up against the reused decoder's
m_inputs map, which was rebuilt from a *different* cgraph instance with different hash
slots -> std::map::at throws "map::at" and llama_decode fails.

Fix: derive the suffix from the tensor's topological index in cgraph->nodes[] /
cgraph->leafs[] instead of the pointer-hash slot. That index is identical for
corresponding tensors across graphs that share the same structure, so cached OV names
resolve correctly on reuse, while still uniquely disambiguating same-named tensors
within a graph.

Fixes test-thread-safety (test 33) on the OpenVINO backend:
"GGML OpenVINO backend std::exception: map::at" / "One or more threads failed".
Bisected to 49852ce (the commit that introduced get_tensor_ov_name): passes at its
parent, fails from that commit onward. Verified: test-thread-safety now passes, and
test-backend-ops (OPENVINO) stays 2622/2622.