Project: llamacpp

Language	c++
OSS-Fuzz project	link
Project repository	link
Build status: Fuzzers	succeeding: Build log
Build status: Code coverage	succeeding: Build log
Build status: Fuzz Introspector	succeeding: Build log
Fuzzer count	6
Lines of code	75888
Lines covered	7026
Code coverage	9.26%
Code coverage report	Report link
Static reachability	38.67%
Fuzz Introspector report	Report link (from 2025-12-28)

Fuzzer	Code coverage (lines)	Latest Report	Comments
fuzz_apply_template (plot)	`1.09% (avg: 1.13%, max: 1.15%)`	2025-12-28
fuzz_grammar (plot)	`0.56% (avg: 0.56%, max: 0.59%)`	2025-12-28	Coverage error: out-of-memory (used: 2112Mb; limit: 2048Mb)
fuzz_inference (plot)	`2.16% (avg: 2.27%, max: 2.36%)`	2025-12-28	Coverage error: out-of-memory (used: 2237Mb; limit: 2048Mb)
fuzz_json_to_grammar (plot)	`5.73% (avg: 5.57%, max: 5.74%)`	2025-12-28
fuzz_load_model (plot)	`0.6% (avg: 0.62%, max: 0.63%)`	2025-12-28
fuzz_structured (plot)	`0.66% (avg: 0.69%, max: 0.7%)`	2025-12-28

Historical Progression

Per Fuzzer Progression

This section shows graphs for the coverage results per fuzz target over the past 30 days. Included are the coverage percentage, total number of lines, covered number of lines, and the number of coverage inputs.

Functions of interest to fuzz

This section outlines functions that may be of interest to fuzz. They are based on ranking functions that have a lot of complexity but currently exhibit low code coverage. The complexity is calculated based on the function itself as well as the functions called by the given function, i.e. the tree of code that the function triggers.

This is only a minor amount of introspection information available for this project. Please consult the Fuzz Introspector report for more information, e.g. the introspection table of all functions in the target project available here.

Function name	Function source file	Accumulated cyclomatic complexity	Code coverage
common_init_from_params(common_params&)	/src/llama.cpp/common/common.cpp	19207	0.0%
common_init_result::common_init_result(common_params&)	/src/llama.cpp/common/common.cpp	17731	0.0%
llama_params_fit	/src/llama.cpp/src/llama.cpp	16567	0.0%
llama_params_fit_impl(charconst,llama_model_params,llama_context_params,float,llama_model_tensor_buft_override*,unsignedlong,unsignedint,ggml_log_level)	/src/llama.cpp/src/llama.cpp	16549	0.0%
llama_params_fit_impl(charconst,llama_model_params,llama_context_params,float,llama_model_tensor_buft_override,unsignedlong,unsignedint,ggml_log_level)::$_2::operator()(charconst,std::__1::vector<llama_params_fit_impl(charconst,llama_model_params,llama_context_params,float,llama_model_tensor_buft_override,unsignedlong,unsignedint,ggml_log_level)::ngl_t,std::__1::allocator<llama_params_fit_impl(charconst,llama_model_params,llama_context_params,float,llama_model_tensor_buft_override,unsignedlong,unsignedint,ggml_log_level)::ngl_t>>const&,std::__1::vector<ggml_backend_buffer_type,std::__1::allocator<ggml_backend_buffer_type>>const&)const	/src/llama.cpp/src/llama.cpp	16111	0.0%
llama_get_device_memory_data(charconst,llama_model_paramsconst,llama_context_paramsconst,std::__1::vector<ggml_backend_device,std::__1::allocator<ggml_backend_device*>>&,unsignedint&,unsignedint&,unsignedint&,ggml_log_level)	/src/llama.cpp/src/llama.cpp	16002	0.0%
llama_model_load_from_splits	/src/llama.cpp/src/llama.cpp	13593	0.0%
llama_model::load_tensors(llama_model_loader&)	/src/llama.cpp/src/llama-model.cpp	8853	0.0%
ggml_graph_compute_with_ctx	/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c	7474	0.0%
ggml_backend_cpu_graph_compute(ggml_backend,ggml_cgraph)	/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	7469	0.0%