High level conclusions

Fuzzer fuzz_inference is blocked:

The runtime code coverage of fuzz_inference covers 25.15% of its statically reachable code. This means there is some place that blocks the fuzzer to continue exploring more code at run time.

Fuzzer fuzz_structured is blocked:

The runtime code coverage of fuzz_structured covers 6.153% of its statically reachable code. This means there is some place that blocks the fuzzer to continue exploring more code at run time.

Fuzzer fuzz_load_model is blocked:

The runtime code coverage of fuzz_load_model covers 6.160% of its statically reachable code. This means there is some place that blocks the fuzzer to continue exploring more code at run time.

Fuzzer fuzz_grammar is blocked:

The runtime code coverage of fuzz_grammar covers 15.27% of its statically reachable code. This means there is some place that blocks the fuzzer to continue exploring more code at run time.

Fuzzers reach 38.67% of cyclomatic complexity.

Fuzzers reach 24.10% of all functions.

Reachability and coverage overview

Functions statically reachable by fuzzers

1158 / 4804

Cyclomatic complexity statically reachable by fuzzers

19891 / 51437

Runtime code coverage of functions

930 / 4804

The following table shows data about each function in the project. The functions included in this table correspond to all functions that exist in the executables of the fuzzers. As such, there may be functions that are from third-party libraries.

For further technical details on the meaning of columns in the below table, please see the Glossary .

Func name	Functions filename	Args	Function call depth	Reached by Fuzzers	Runtime reached by Fuzzers	Combined reached by Fuzzers	Fuzzers runtime hit	Func lines hit %	I Count	BB Count	Cyclomatic complexity	Functions reached	Reached by functions	Accumulated cyclomatic complexity	Undiscovered complexity

Fuzzer: fuzz_apply_template

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	2	2.29%
gold	[1:9]	4	4.59%
yellow	[10:29]	12	13.7%
greenyellow	[30:49]	10	11.4%
lawngreen	50+	59	67.8%
All colors	87	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
1	55	llm_chat_detect_template(std::__1::basic_string , std::__1::allocator > const&)	call site: 00055
1	75	llm_chat_apply_template(llm_chat_template, std::__1::vector > const&, std::__1::basic_string , std::__1::allocator >&, bool)	call site: 00075

Runtime coverage analysis

Covered functions

7

Functions that are reachable but not covered

5

Reachable functions

17

Percentage of reachable functions covered

70.59%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_apply_template.cpp	1
/src/llama.cpp/src/llama.cpp	1
/src/llama.cpp/src/llama-chat.cpp	5

Fuzzer: fuzz_grammar

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	402	76.8%
gold	[1:9]	42	8.03%
yellow	[10:29]	4	0.76%
greenyellow	[30:49]	12	2.29%
lawngreen	50+	63	12.0%
All colors	523	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
378	72	parse_token(llama_vocab const*, char const*)	call site: 00072	ggml_abort
5	38	parse_char(char const*)	call site: 00038	__cxa_allocate_exception
5	50	llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string , std::__1::allocator > const&, std::__1::vector >&, bool)	call site: 00050	__cxa_allocate_exception
4	33	parse_char(char const*)	call site: 00033	__cxa_allocate_exception
4	468	llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string , std::__1::allocator > const&, std::__1::vector >&, bool)	call site: 00468	__cxa_allocate_exception
4	480	llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string , std::__1::allocator > const&, std::__1::vector >&, bool)	call site: 00480	__cxa_allocate_exception
1	26	llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string , std::__1::allocator > const&, std::__1::vector >&, bool)	call site: 00026
1	498	llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string , std::__1::allocator > const&, std::__1::vector >&, bool)	call site: 00498

Runtime coverage analysis

Covered functions

19

Functions that are reachable but not covered

233

Reachable functions

275

Percentage of reachable functions covered

15.27%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_grammar.cpp	1
/src/llama.cpp/src/llama-grammar.h	2
/src/llama.cpp/src/llama-grammar.cpp	17
/src/llama.cpp/src/llama-vocab.cpp	65
/src/llama.cpp/ggml/src/ggml.c	1
/src/llama.cpp/src/llama-impl.cpp	3
/src/llama.cpp/src/unicode.cpp	32
/usr/local/bin/../include/c++/v1/stdexcept	1
/src/llama.cpp/src/unicode.h	3

Fuzzer: fuzz_load_model

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	8012	98.9%
gold	[1:9]	63	0.77%
yellow	[10:29]	4	0.04%
greenyellow	[30:49]	6	0.07%
lawngreen	50+	13	0.16%
All colors	8098	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
7974	104	llama_model_load_from_file_impl(std::__1::basic_string , std::__1::allocator > const&, std::__1::vector , std::__1::allocator >, std::__1::allocator , std::__1::allocator > > >&, llama_model_params)	call site: 00104	gguf_init_from_file
12	6	ggml_init	call site: 00006	ggml_abort
10	93	ggml_backend_dev_type	call site: 00093	ggml_backend_dev_backend_reg
4	36	LLVMFuzzerTestOneInput	call site: 00036	llama_model_load_from_file
2	8080	llama_model::~llama_model()	call site: 08080	llama_free_model
1	19	ggml_init	call site: 00019	ggml_log_internal
1	21	ggml_aligned_malloc	call site: 00021	ggml_log_internal
1	24	ggml_init	call site: 00024	ggml_free
1	49	ggml_cpu_init	call site: 00049	clock_gettime
1	55	ggml_compute_fp16_to_fp32	call site: 00055	fp32_from_bits
1	59	ggml_cpu_init	call site: 00059	fp32_to_bits
1	77	get_reg()	call site: 00077

Runtime coverage analysis

Covered functions

55

Functions that are reachable but not covered

914

Reachable functions

974

Percentage of reachable functions covered

6.16%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_load_model.cpp	2
/src/llama.cpp/src/llama.cpp	10
/src/llama.cpp/ggml/src/ggml.c	69
/src/llama.cpp/ggml/src/ggml-threading.cpp	2
/src/llama.cpp/ggml/src/ggml-backend-reg.cpp	13
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	1
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c	1
/src/llama.cpp/ggml/src/./ggml-impl.h	4
/src/llama.cpp/ggml/src/ggml-cpu/vec.h	2
/src/llama.cpp/ggml/src/ggml-backend.cpp	48
/src/llama.cpp/src/llama-impl.cpp	10
/src/llama.cpp/src/llama-model.cpp	28
/src/llama.cpp/src/llama-vocab.cpp	87
/src/llama.cpp/src/llama-model-loader.cpp	92
/src/llama.cpp/src/llama-arch.cpp	9
/src/llama.cpp/ggml/src/gguf.cpp	79
/src/llama.cpp/src/llama-mmap.cpp	24
/src/llama.cpp/src/llama-model-loader.h	2
/src/llama.cpp/src/llama-hparams.cpp	11
/src/llama.cpp/src/unicode.cpp	34
/usr/local/bin/../include/c++/v1/stdexcept	1
/src/llama.cpp/src/unicode.h	3
/src/llama.cpp/src/llama-arch.h	5
/src/llama.cpp/ggml/src/ggml-impl.h	1
/src/llama.cpp/ggml/src/ggml-alloc.c	7
/src/llama.cpp/ggml/src/ggml-quants.c	10

Fuzzer: fuzz_structured

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	8012	98.8%
gold	[1:9]	53	0.65%
yellow	[10:29]	6	0.07%
greenyellow	[30:49]	16	0.19%
lawngreen	50+	16	0.19%
All colors	8103	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
7974	108	llama_model_load_from_file_impl(std::__1::basic_string , std::__1::allocator > const&, std::__1::vector , std::__1::allocator >, std::__1::allocator , std::__1::allocator > > >&, llama_model_params)	call site: 00108	gguf_init_from_file
12	6	ggml_init	call site: 00006	ggml_abort
10	97	ggml_backend_dev_type	call site: 00097	ggml_backend_dev_backend_reg
4	40	LLVMFuzzerTestOneInput	call site: 00040	llama_model_load_from_file
2	8084	llama_model::~llama_model()	call site: 08084	llama_free_model
1	19	ggml_init	call site: 00019	ggml_log_internal
1	21	ggml_aligned_malloc	call site: 00021	ggml_log_internal
1	24	ggml_init	call site: 00024	ggml_free
1	53	ggml_cpu_init	call site: 00053	clock_gettime
1	59	ggml_compute_fp16_to_fp32	call site: 00059	fp32_from_bits
1	63	ggml_cpu_init	call site: 00063	fp32_to_bits
1	81	get_reg()	call site: 00081

Runtime coverage analysis

Covered functions

55

Functions that are reachable but not covered

915

Reachable functions

975

Percentage of reachable functions covered

6.15%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_structured.cpp	2
/src/llama.cpp/src/llama.cpp	10
/src/llama.cpp/ggml/src/ggml.c	69
/src/llama.cpp/ggml/src/ggml-threading.cpp	2
/src/llama.cpp/ggml/src/ggml-backend-reg.cpp	13
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	1
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c	1
/src/llama.cpp/ggml/src/./ggml-impl.h	4
/src/llama.cpp/ggml/src/ggml-cpu/vec.h	2
/src/llama.cpp/ggml/src/ggml-backend.cpp	48
/src/llama.cpp/src/llama-impl.cpp	10
/src/llama.cpp/src/llama-model.cpp	28
/src/llama.cpp/src/llama-vocab.cpp	87
/src/llama.cpp/src/llama-model-loader.cpp	92
/src/llama.cpp/src/llama-arch.cpp	9
/src/llama.cpp/ggml/src/gguf.cpp	79
/src/llama.cpp/src/llama-mmap.cpp	24
/src/llama.cpp/src/llama-model-loader.h	2
/src/llama.cpp/src/llama-hparams.cpp	11
/src/llama.cpp/src/unicode.cpp	34
/usr/local/bin/../include/c++/v1/stdexcept	1
/src/llama.cpp/src/unicode.h	3
/src/llama.cpp/src/llama-arch.h	5
/src/llama.cpp/ggml/src/ggml-impl.h	1
/src/llama.cpp/ggml/src/ggml-alloc.c	7
/src/llama.cpp/ggml/src/ggml-quants.c	10

Fuzzer: fuzz_inference

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	8418	95.0%
gold	[1:9]	101	1.14%
yellow	[10:29]	9	0.10%
greenyellow	[30:49]	3	0.03%
lawngreen	50+	325	3.66%
All colors	8856	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
6213	1869	llama_model_load(std::__1::basic_string , std::__1::allocator > const&, std::__1::vector , std::__1::allocator >, std::__1::allocator , std::__1::allocator > > >&, llama_model&, llama_model_params&)	call site: 01869	llama_supports_gpu_offload
1092	776	llama_model::load_hparams(llama_model_loader&)	call site: 00776	llama_model_rope_type
741	8099	llama_model::~llama_model()	call site: 08099	llama_new_context_with_model
78	442	GGUFMeta::GKV ::get_kv(gguf_context const*, int)	call site: 00442	gguf_init_from_file
30	677	GGUFMeta::GKV ::get_kv(gguf_context const*, int)	call site: 00677	gguf_find_key
16	27	LLVMFuzzerTestOneInput	call site: 00027
15	661	llama_model::load_hparams(llama_model_loader&)	call site: 00661	gguf_find_key
14	352	_ZN8GGUFMeta3GKVINSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEEE12try_overrideIS7_EENS1_9enable_ifIXsr3std7is_sameIT_S7_EE5valueEbE4typeERS7_PK23llama_model_kv_override	call site: 00352	__cxa_allocate_exception
12	521	GGUFMeta::GKV ::get_kv(gguf_context const*, int)	call site: 00521	gguf_get_val_i32
11	112	ggml_backend_dev_type	call site: 00112	ggml_backend_dev_backend_reg
11	753	llama_model::load_hparams(llama_model_loader&)	call site: 00753	ggml_abort
10	710	llama_model::load_hparams(llama_model_loader&)	call site: 00710	_ZN8GGUFMeta3GKVIbE12try_overrideIbEENSt3__19enable_ifIXsr3std7is_sameIT_bEE5valueEbE4typeERS5_PK23llama_model_kv_override

Runtime coverage analysis

Covered functions

211

Functions that are reachable but not covered

949

Reachable functions

1268

Percentage of reachable functions covered

25.16%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_inference.cpp	1
/src/llama.cpp/src/llama.cpp	11
/src/llama.cpp/ggml/src/ggml.c	98
/src/llama.cpp/ggml/src/ggml-threading.cpp	2
/src/llama.cpp/common/common.h	13
/src/llama.cpp/common/common.cpp	2
/src/llama.cpp/src/llama-model.cpp	36
/src/llama.cpp/ggml/src/ggml-backend-reg.cpp	14
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	1
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c	1
/src/llama.cpp/ggml/src/./ggml-impl.h	4
/src/llama.cpp/ggml/src/ggml-cpu/vec.h	2
/src/llama.cpp/ggml/src/ggml-backend.cpp	76
/src/llama.cpp/src/llama-impl.cpp	10
/src/llama.cpp/src/llama-vocab.cpp	87
/src/llama.cpp/src/llama-model-loader.cpp	92
/src/llama.cpp/src/llama-arch.cpp	10
/src/llama.cpp/ggml/src/gguf.cpp	79
/src/llama.cpp/src/llama-mmap.cpp	24
/src/llama.cpp/src/llama-model-loader.h	2
/src/llama.cpp/src/llama-hparams.cpp	18
/src/llama.cpp/src/unicode.cpp	34
/usr/local/bin/../include/c++/v1/stdexcept	1
/src/llama.cpp/src/unicode.h	3
/src/llama.cpp/src/llama-arch.h	5
/src/llama.cpp/ggml/src/ggml-impl.h	14
/src/llama.cpp/ggml/src/ggml-alloc.c	34
/src/llama.cpp/ggml/src/ggml-quants.c	10
/src/llama.cpp/src/llama-context.cpp	16
/src/llama.cpp/src/llama-adapter.h	2
/src/llama.cpp/src/llama-graph.h	6
/src/llama.cpp/src/llama-memory-recurrent.cpp	4
/src/llama.cpp/src/llama-memory.h	2
/src/llama.cpp/src/llama-memory-hybrid.cpp	1
/src/llama.cpp/src/llama-kv-cache.cpp	4
/src/llama.cpp/src/llama-kv-cache.h	3
/src/llama.cpp/src/llama-kv-cells.h	3
/src/llama.cpp/src/llama-kv-cache-iswa.cpp	1
/src/llama.cpp/src/llama-graph.cpp	8
/src/llama.cpp/src/llama-hparams.h	1
/src/llama.cpp/src/llama-batch.h	5
/src/llama.cpp/src/llama-batch.cpp	5
/src/llama.cpp/ggml/src/ggml-opt.cpp	2

Fuzzer: fuzz_json_to_grammar

Call tree

The calltree shows the control flow of the fuzzer. This is overlaid with coverage information to display how much of the potential code a fuzzer can reach is in fact covered at runtime. In the following there is a link to a detailed calltree visualisation as well as a bitmap showing a high-level view of the calltree. For further information about these topics please see the glossary for full calltree and calltree overview

Call tree overview bitmap:

The distribution of callsites in terms of coloring is

Color	Runtime hitcount	Callsite count	Percentage
red	0	368	43.2%
gold	[1:9]	3	0.35%
yellow	[10:29]	8	0.94%
greenyellow	[30:49]	3	0.35%
lawngreen	50+	469	55.1%
All colors	851	100

Fuzz blockers

The following nodes represent call sites where fuzz blockers occur.

Amount of callsites blocked	Calltree index	Parent function	Callsite	Largest blocked function
93	472	void nlohmann::json_abi_v3_12_0::detail::external_constructor<(nlohmann::json_abi_v3_12_0::detail::value_t)6>::construct , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> >(nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>&, nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::number_unsigned_t)	call site: 00472	_ZN8nlohmann16json_abi_v3_12_06detail11parse_error6createIDnTnNSt3__19enable_ifIXsr21is_basic_json_contextIT_EE5valueEiE4typeELi0EEES2_iRKNS1_10position_tERKNS4_12basic_stringIcNS4_11char_traitsIcEENS4_9allocatorIcEEEES6_
39	242	nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::end()	call site: 00242	_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvE5eraseINS0_6detail9iter_implISE_EETnNS2_9enable_ifIXoosr3std7is_sameIT_SI_EE5valuesr3std7is_sameISK_NSH_IKSE_EEEE5valueEiE4typeELi0EEESK_SK_
21	370	_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvEC2IRSA_SA_TnNS2_9enable_ifIXaantsr6detail13is_basic_jsonIT0_EE5valuesr6detail18is_compatible_typeISE_SI_EE5valueEiE4typeELi0EEEOT_	call site: 00370	_ZN8nlohmann16json_abi_v3_12_06detail11parse_error6createIDnTnNSt3__19enable_ifIXsr21is_basic_json_contextIT_EE5valueEiE4typeELi0EEES2_iRKNS1_10position_tERKNS4_12basic_stringIcNS4_11char_traitsIcEENS4_9allocatorIcEEEES6_
16	392	std::__1::basic_string , std::__1::allocator > nlohmann::json_abi_v3_12_0::detail::concat , std::__1::allocator >, char const (&) [23], std::__1::basic_string , std::__1::allocator > >(char const (&) [23], std::__1::basic_string , std::__1::allocator >&&)	call site: 00392	_ZN8nlohmann16json_abi_v3_12_06detail12out_of_range6createIPNS0_10basic_jsonINSt3__13mapENS5_6vectorENS5_12basic_stringIcNS5_11char_traitsIcEENS5_9allocatorIcEEEEblmdSB_NS0_14adl_serializerENS7_IhNSB_IhEEEEvEETnNS5_9enable_ifIXsr21is_basic_json_contextIT_EE5valueEiE4typeELi0EEES2_iRKSD_SK_
15	218	nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::json_value::json_value(std::__1::basic_string , std::__1::allocator > const&)	call site: 00218
15	345	nlohmann::json_abi_v3_12_0::detail::parse_error::parse_error(int, unsigned long, char const*)	call site: 00345	_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvEC2IRSA_SA_TnNS2_9enable_ifIXaantsr6detail13is_basic_jsonIT0_EE5valuesr6detail18is_compatible_typeISE_SI_EE5valueEiE4typeELi0EEEOT_
12	801	nlohmann::json_abi_v3_12_0::byte_container_with_subtype > > const& nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::get_ref_impl > > const&, nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> const>(nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> const&)	call site: 00801	_ZN8nlohmann16json_abi_v3_12_014adl_serializerINS0_27byte_container_with_subtypeINSt3__16vectorIhNS3_9allocatorIhEEEEEEvE7to_jsonINS0_10basic_jsonINS0_11ordered_mapES4_NS3_12basic_stringIcNS3_11char_traitsIcEENS5_IcEEEEblmdS5_S1_S7_vEERKS8_EEDTcmclL_ZNS0_7to_jsonEEfp_clsr3stdE7forwardIT0_Efp0_EEcvv_EERT_OSL_
10	183	nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::operator=(nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>)	call site: 00183	__cxa_allocate_exception
10	411	_ZN8nlohmann16json_abi_v3_12_06detail12out_of_range6createIDnTnNSt3__19enable_ifIXsr21is_basic_json_contextIT_EE5valueEiE4typeELi0EEES2_iRKNS4_12basic_stringIcNS4_11char_traitsIcEENS4_9allocatorIcEEEES6_	call site: 00411	_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvEC2IRddTnNS2_9enable_ifIXaantsr6detail13is_basic_jsonIT0_EE5valuesr6detail18is_compatible_typeISE_SI_EE5valueEiE4typeELi0EEEOT_
8	172	nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::data::~data()	call site: 00172
8	204	nlohmann::json_abi_v3_12_0::detail::out_of_range::out_of_range(int, char const*)	call site: 00204	__cxa_throw
8	440	void nlohmann::json_abi_v3_12_0::detail::external_constructor<(nlohmann::json_abi_v3_12_0::detail::value_t)4>::construct , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> >(nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>&, nlohmann::json_abi_v3_12_0::basic_json , std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::boolean_t)	call site: 00440	_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvEC2IRllTnNS2_9enable_ifIXaantsr6detail13is_basic_jsonIT0_EE5valuesr6detail18is_compatible_typeISE_SI_EE5valueEiE4typeELi0EEEOT_

Runtime coverage analysis

Covered functions

511

Functions that are reachable but not covered

160

Reachable functions

675

Percentage of reachable functions covered

76.3%

NB: The sum of covered functions and functions that are reachable but not covered need not be equal to Reachable functions . This is because the reachability analysis is an approximation and thus at runtime some functions may be covered that are not included in the reachability analysis. This is a limitation of our static analysis capabilities.

Function name	source code lines	source lines hit	percentage hit

Files reached

filename	functions hit
/src/llama.cpp/fuzzers/fuzz_json_to_grammar.cpp	1
/src/llama.cpp/vendor/nlohmann/json.hpp	375
/usr/local/bin/../include/c++/v1/__exception/exception.h	2
/src/llama.cpp/common/json-schema-to-grammar.cpp	6
/src/llama.cpp/common/common.cpp	1
/usr/local/bin/../include/c++/v1/stdexcept	1
/src/llama.cpp/common/json-schema-to-grammar.h	1

Optimal target analysis

Remaining optimal interesting functions

The following table shows a list of functions that are optimal targets. Optimal targets are identified by finding the functions that in combination, yield a high code coverage.

Func name	Functions filename	Arg count	Args	Function depth	hitcount	instr count	bb count	cyclomatic complexity	Reachable functions	Incoming references	total cyclomatic complexity	Unreached complexity
ggml_backend_cpu_graph_compute(ggml_backend*,ggml_cgraph*)	/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	2	['N/A', 'N/A']	15	0	59	9	4	1720	0	7469	7327
common_init_from_params(common_params&)	/src/llama.cpp/common/common.cpp	2	['N/A', 'N/A']	19	0	700	171	151	1589	0	19207	3169
common_schema_converter::visit(nlohmann::json_abi_v3_12_0::basic_json ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>const&,std::__1::basic_string ,std::__1::allocator >const&)	/src/llama.cpp/common/json-schema-to-grammar.cpp	4	['N/A', 'N/A', 'N/A', 'N/A']	18	0	2479	632	505	507	5	2821	2515
ggml_quantize_chunk	/src/llama.cpp/ggml/src/ggml.c	7	['int', 'N/A', 'N/A', 'size_t', 'size_t', 'size_t', 'N/A']	4	0	274	41	2	99	0	2265	2220
llm_build_qwen3next::llm_build_qwen3next(llama_modelconst&,llm_graph_paramsconst&)	/src/llama.cpp/src/models/qwen3next.cpp	3	['N/A', 'N/A', 'N/A']	12	0	266	62	33	305	0	839	538
llama_opt_epoch	/src/llama.cpp/src/llama-context.cpp	7	['N/A', 'N/A', 'N/A', 'N/A', 'size_t', 'N/A', 'N/A']	14	0	12	3	2	394	0	2292	493
llama_model_save_to_file	/src/llama.cpp/src/llama.cpp	2	['N/A', 'N/A']	8	0	53	12	9	212	0	541	364

Implementing fuzzers that target the above functions will improve reachability such that it becomes:

Functions statically reachable by fuzzers

2834 / 4804

Cyclomatic complexity statically reachable by fuzzers

35956 / 51437

All functions overview

If you implement fuzzers for these functions, the status of all functions in the project will be:

Func name	Functions filename	Args	Function call depth	Reached by Fuzzers	Runtime reached by Fuzzers	Combined reached by Fuzzers	Fuzzers runtime hit	Func lines hit %	I Count	BB Count	Cyclomatic complexity	Functions reached	Reached by functions	Accumulated cyclomatic complexity	Undiscovered complexity

This sections provides heuristics that can be used as input to a fuzz engine when running a given fuzz target. The current focus is on providing input that is usable by libFuzzer.

/src/llama.cpp/fuzzers/fuzz_apply_template.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['llm_chat_detect_template(std::__1::basic_string, std::__1::allocator > const&)', 'llm_chat_apply_template(llm_chat_template, std::__1::vector > const&, std::__1::basic_string, std::__1::allocator >&, bool)']

/src/llama.cpp/fuzzers/fuzz_grammar.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['parse_token(llama_vocab const*, char const*)', 'parse_char(char const*)', 'llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string, std::__1::allocator > const&, std::__1::vector >&, bool)', 'parse_char(char const*)', 'llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string, std::__1::allocator > const&, std::__1::vector >&, bool)', 'llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string, std::__1::allocator > const&, std::__1::vector >&, bool)', 'llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string, std::__1::allocator > const&, std::__1::vector >&, bool)', 'llama_grammar_parser::parse_sequence(char const*, std::__1::basic_string, std::__1::allocator > const&, std::__1::vector >&, bool)']

/src/llama.cpp/fuzzers/fuzz_load_model.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['llama_model_load_from_file_impl(std::__1::basic_string, std::__1::allocator > const&, std::__1::vector, std::__1::allocator >, std::__1::allocator, std::__1::allocator > > >&, llama_model_params)', 'ggml_init', 'ggml_backend_dev_type', 'LLVMFuzzerTestOneInput', 'llama_model::~llama_model()', 'ggml_aligned_malloc', 'ggml_cpu_init', 'ggml_compute_fp16_to_fp32']

/src/llama.cpp/fuzzers/fuzz_structured.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['llama_model_load_from_file_impl(std::__1::basic_string, std::__1::allocator > const&, std::__1::vector, std::__1::allocator >, std::__1::allocator, std::__1::allocator > > >&, llama_model_params)', 'ggml_init', 'ggml_backend_dev_type', 'LLVMFuzzerTestOneInput', 'llama_model::~llama_model()', 'ggml_aligned_malloc', 'ggml_cpu_init', 'ggml_compute_fp16_to_fp32']

/src/llama.cpp/fuzzers/fuzz_inference.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['llama_model_load(std::__1::basic_string, std::__1::allocator > const&, std::__1::vector, std::__1::allocator >, std::__1::allocator, std::__1::allocator > > >&, llama_model&, llama_model_params&)', 'llama_model::load_hparams(llama_model_loader&)', 'llama_model::~llama_model()', 'GGUFMeta::GKV::get_kv(gguf_context const*, int)', 'GGUFMeta::GKV::get_kv(gguf_context const*, int)', 'LLVMFuzzerTestOneInput', 'llama_model::load_hparams(llama_model_loader&)', '_ZN8GGUFMeta3GKVINSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEEE12try_overrideIS7_EENS1_9enable_ifIXsr3std7is_sameIT_S7_EE5valueEbE4typeERS7_PK23llama_model_kv_override', 'GGUFMeta::GKV::get_kv(gguf_context const*, int)', 'ggml_backend_dev_type']

/src/llama.cpp/fuzzers/fuzz_json_to_grammar.cpp

Dictionary

Use this with the libFuzzer -dict=DICT.file flag

Fuzzer function priority

Use one of these functions as input to libfuzzer with flag: -focus_function name

-focus_function=['void nlohmann::json_abi_v3_12_0::detail::external_constructor<(nlohmann::json_abi_v3_12_0::detail::value_t)6>::construct, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> >(nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>&, nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::number_unsigned_t)', 'nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::end()', '_ZN8nlohmann16json_abi_v3_12_010basic_jsonINSt3__13mapENS2_6vectorENS2_12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEblmdS8_NS0_14adl_serializerENS4_IhNS8_IhEEEEvEC2IRSA_SA_TnNS2_9enable_ifIXaantsr6detail13is_basic_jsonIT0_EE5valuesr6detail18is_compatible_typeISE_SI_EE5valueEiE4typeELi0EEEOT_', 'std::__1::basic_string, std::__1::allocator > nlohmann::json_abi_v3_12_0::detail::concat, std::__1::allocator >, char const (&) [23], std::__1::basic_string, std::__1::allocator > >(char const (&) [23], std::__1::basic_string, std::__1::allocator >&&)', 'nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::json_value::json_value(std::__1::basic_string, std::__1::allocator > const&)', 'nlohmann::json_abi_v3_12_0::detail::parse_error::parse_error(int, unsigned long, char const*)', 'nlohmann::json_abi_v3_12_0::byte_container_with_subtype > > const& nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::get_ref_impl > > const&, nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> const>(nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void> const&)', 'nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::operator=(nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>)', '_ZN8nlohmann16json_abi_v3_12_06detail12out_of_range6createIDnTnNSt3__19enable_ifIXsr21is_basic_json_contextIT_EE5valueEiE4typeELi0EEES2_iRKNS4_12basic_stringIcNS4_11char_traitsIcEENS4_9allocatorIcEEEES6_', 'nlohmann::json_abi_v3_12_0::basic_json, std::__1::allocator >, bool, long, unsigned long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector >, void>::data::~data()']

This section shows analysis of runtime coverage data.

For futher technical details on how this section is generated, please see the Glossary .

Complex functions with low coverage

Func name	Function total lines	Lines covered at runtime	percentage covered	Reached by fuzzers
parse_token(llama_vocabconst*,charconst*)	38	19	50.0%	['fuzz_grammar']
llama_model_loader::llama_model_loader(std::__1::basic_string ,std::__1::allocator >const&,std::__1::vector ,std::__1::allocator >,std::__1::allocator ,std::__1::allocator >>>&,bool,bool,bool,llama_model_kv_overrideconst*,llama_model_tensor_buft_overrideconst*)	186	88	47.31%	['fuzz_structured', 'fuzz_load_model', 'fuzz_inference']
llama_model_ftype_name(llama_ftype)	41	9	21.95%	['fuzz_structured', 'fuzz_load_model', 'fuzz_inference']
llama_model::load_hparams(llama_model_loader&)	1641	72	4.387%	['fuzz_structured', 'fuzz_load_model', 'fuzz_inference']
nlohmann::json_abi_v3_12_0::basic_json ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>::json_value::json_value(nlohmann::json_abi_v3_12_0::detail::value_t)	60	17	28.33%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::detail::parser ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>,nlohmann::json_abi_v3_12_0::detail::iterator_input_adapter >::parse(bool,nlohmann::json_abi_v3_12_0::basic_json ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>&)	40	16	40.0%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::basic_json ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>::json_value::json_value(nlohmann::json_abi_v3_12_0::detail::value_t)	60	9	15.0%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::detail::iter_impl ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>>::set_begin()	33	14	42.42%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::detail::iter_impl ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>>::operator*()const	33	14	42.42%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::detail::serializer ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>>::dump(nlohmann::json_abi_v3_12_0::basic_json ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>const&,bool,bool,unsignedint,unsignedint)	215	99	46.04%	['fuzz_json_to_grammar']
nlohmann::json_abi_v3_12_0::detail::serializer ,std::__1::allocator >,bool,long,unsignedlong,double,std::__1::allocator,nlohmann::json_abi_v3_12_0::adl_serializer,std::__1::vector >,void>>::dump_escaped(std::__1::basic_string ,std::__1::allocator >const&,bool)	188	94	50.0%	['fuzz_json_to_grammar']

This section shows which files and directories are considered in this report. The main reason for showing this is fuzz introspector may include more code in the reasoning than is desired. This section helps identify if too many files/directories are included, e.g. third party code, which may be irrelevant for the threat model. In the event too much is included, fuzz introspector supports a configuration file that can exclude data from the report. See the following link for more information on how to create a config file: link

Files in report

Source file	Reached by	Covered by
	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/vec.h	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/bailingmoe2.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-backend.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/mistral3.cpp	[]	[]
/src/llama.cpp/src/unicode-data.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/repack.cpp	[]	[]
/src/llama.cpp/src/models/arcee.cpp	[]	[]
/src/llama.cpp/fuzzers/fuzz_structured.cpp	['fuzz_structured']	['fuzz_structured']
/src/llama.cpp/src/models/smollm3.cpp	[]	[]
/src/llama.cpp/src/llama.cpp	['fuzz_apply_template', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_apply_template', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/ggml/src/ggml-cpu/ops.cpp	[]	[]
/src/llama.cpp/src/llama-context.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/models/bert.cpp	[]	[]
/src/llama.cpp/src/models/ernie4-5.cpp	[]	[]
/src/llama.cpp/common/json-schema-to-grammar.h	['fuzz_json_to_grammar']	[]
/src/llama.cpp/src/models/wavtokenizer-dec.cpp	[]	[]
/src/llama.cpp/src/models/grovemoe.cpp	[]	[]
/src/llama.cpp/src/models/stablelm.cpp	[]	[]
/usr/local/bin/../include/c++/v1/__exception/exception_ptr.h	[]	[]
/src/llama.cpp/src/models/grok.cpp	[]	[]
/src/llama.cpp/src/models/plamo2.cpp	[]	[]
/usr/local/bin/../include/c++/v1/string	[]	[]
/src/llama.cpp/src/models/xverse.cpp	[]	[]
/src/llama.cpp/src/models/codeshell.cpp	[]	[]
/src/llama.cpp/src/llama-batch.h	['fuzz_inference']	[]
/src/llama.cpp/src/models/mimo2-iswa.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml.c	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/common/common.cpp	['fuzz_inference', 'fuzz_json_to_grammar']	['fuzz_inference', 'fuzz_json_to_grammar']
/src/llama.cpp/src/models/openai-moe-iswa.cpp	[]	[]
/src/llama.cpp/src/llama-model-loader.h	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_inference']
/src/llama.cpp/src/models/llama.cpp	[]	[]
/src/llama.cpp/common/../vendor/nlohmann/json_fwd.hpp	[]	[]
/src/llama.cpp/ggml/src/ggml-opt.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/models/qwen3.cpp	[]	[]
/src/llama.cpp/src/llama-model.h	[]	[]
/src/llama.cpp/src/models/hunyuan-moe.cpp	[]	[]
/src/llama.cpp/src/llama-grammar.cpp	['fuzz_grammar']	['fuzz_grammar']
/src/llama.cpp/src/models/gemma3.cpp	[]	[]
/src/llama.cpp/src/models/olmoe.cpp	[]	[]
/src/llama.cpp/src/models/qwen3moe.cpp	[]	[]
/src/llama.cpp/src/models/rwkv6-base.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/common.h	[]	[]
/src/llama.cpp/src/unicode.h	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	[]
/src/llama.cpp/src/models/rwkv6qwen2.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-alloc.c	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	[]
/src/llama.cpp/src/models/cohere2-iswa.cpp	[]	[]
/src/llama.cpp/src/llama-io.h	[]	[]
/src/llama.cpp/src/models/gemma-embedding.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/traits.cpp	[]	[]
/src/llama.cpp/src/llama-memory-recurrent.h	[]	[]
/src/llama.cpp/src/models/qwen3vl-moe.cpp	[]	[]
/src/llama.cpp/src/models/rwkv6.cpp	[]	[]
/src/llama.cpp/src/models/deepseek2.cpp	[]	[]
/src/llama.cpp/src/models/falcon-h1.cpp	[]	[]
/src/llama.cpp/src/models/jais.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.cpp	[]	[]
/src/llama.cpp/src/models/t5-dec.cpp	[]	[]
/src/llama.cpp/src/models/bloom.cpp	[]	[]
/src/llama.cpp/src/models/qwen3vl.cpp	[]	[]
/src/llama.cpp/src/models/jamba.cpp	[]	[]
/src/llama.cpp/src/llama-adapter.h	['fuzz_inference']	[]
/src/llama.cpp/src/models/gptneox.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/unary-ops.cpp	[]	[]
/src/llama.cpp/src/models/internlm2.cpp	[]	[]
/src/llama.cpp/src/models/lfm2.cpp	[]	[]
/src/llama.cpp/src/models/cogvlm.cpp	[]	[]
/src/llama.cpp/src/llama-memory-recurrent.cpp	['fuzz_inference']	[]
/src/llama.cpp/ggml/src/ggml-impl.h	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/dbrx.cpp	[]	[]
/src/llama.cpp/src/unicode.cpp	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_grammar', 'fuzz_inference']
/src/llama.cpp/ggml/src/../include/ggml-cpp.h	[]	[]
/src/llama.cpp/src/models/llada-moe.cpp	[]	[]
/src/llama.cpp/ggml/src/gguf.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_inference']
/src/llama.cpp/ggml/src/ggml-cpu/traits.h	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h	[]	[]
/src/llama.cpp/src/models/qwen2vl.cpp	[]	[]
/src/llama.cpp/common/../vendor/nlohmann/json.hpp	[]	['fuzz_json_to_grammar']
/src/llama.cpp/src/models/dots1.cpp	[]	[]
/src/llama.cpp/src/models/modern-bert.cpp	[]	[]
/src/llama.cpp/src/models/smallthinker.cpp	[]	[]
/src/llama.cpp/src/llama-grammar.h	['fuzz_grammar']	['fuzz_grammar']
/src/llama.cpp/src/models/olmo2.cpp	[]	[]
/src/llama.cpp/src/models/plm.cpp	[]	[]
/src/llama.cpp/src/llama-kv-cache.h	['fuzz_inference']	['fuzz_inference']
/src/llama.cpp/src/llama-graph.cpp	['fuzz_inference']	[]
/src/llama.cpp/common/json-schema-to-grammar.cpp	['fuzz_json_to_grammar']	['fuzz_json_to_grammar']
/src/llama.cpp/src/llama-mmap.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/nemotron-h.cpp	[]	[]
/src/llama.cpp/src/models/rwkv7-base.cpp	[]	[]
/src/llama.cpp/src/llama-adapter.cpp	[]	[]
/src/llama.cpp/vendor/nlohmann/json_fwd.hpp	[]	[]
/src/llama.cpp/src/models/models.h	[]	[]
/src/llama.cpp/src/models/glm4.cpp	[]	[]
/src/llama.cpp/src/llama-impl.cpp	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/phi3.cpp	[]	[]
/src/llama.cpp/ggml/src/./ggml-impl.h	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/gemma2-iswa.cpp	[]	[]
/src/llama.cpp/src/llama-vocab.h	[]	[]
/src/llama.cpp/src/llama-memory.cpp	[]	[]
/src/llama.cpp/src/models/deci.cpp	[]	[]
/src/llama.cpp/src/models/t5-enc.cpp	[]	[]
/src/llama.cpp/fuzzers/fuzz_load_model.cpp	['fuzz_load_model']	['fuzz_load_model']
/src/llama.cpp/src/llama-model-loader.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_inference']
/src/llama.cpp/src/llama-memory.h	['fuzz_inference']	[]
/src/llama.cpp/src/models/qwen2moe.cpp	[]	[]
/src/llama.cpp/common/sampling.h	[]	[]
/src/llama.cpp/src/models/baichuan.cpp	[]	[]
/src/llama.cpp/src/models/minicpm3.cpp	[]	[]
/src/llama.cpp/fuzzers/fuzz_grammar.cpp	['fuzz_grammar']	['fuzz_grammar']
/src/llama.cpp/src/models/arctic.cpp	[]	[]
/src/llama.cpp/common/log.cpp	[]	[]
/src/llama.cpp/src/models/arwkv7.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/arch/x86/quants.c	[]	[]
/src/llama.cpp/src/llama-sampling.cpp	[]	[]
/src/llama.cpp/src/llama-kv-cache-iswa.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/../include/llama-cpp.h	[]	[]
/src/llama.cpp/src/llama-arch.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_inference']
/src/llama.cpp/src/llama-model.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/phi2.cpp	[]	[]
/usr/local/bin/../include/c++/v1/istream	[]	[]
/src/llama.cpp/src/models/llama-iswa.cpp	[]	[]
/src/llama.cpp/src/models/command-r.cpp	[]	[]
/src/llama.cpp/src/models/qwen2.cpp	[]	[]
/src/llama.cpp/src/llama-context.h	[]	[]
/src/llama.cpp/src/llama-hparams.h	['fuzz_inference']	[]
/src/llama.cpp/src/llama-graph.h	['fuzz_inference']	[]
/src/llama.cpp/fuzzers/fuzz_apply_template.cpp	['fuzz_apply_template']	['fuzz_apply_template']
/src/llama.cpp/vendor/nlohmann/json.hpp	['fuzz_json_to_grammar']	['fuzz_json_to_grammar']
/src/llama.cpp/src/models/bailingmoe.cpp	[]	[]
/src/llama.cpp/src/models/seed-oss.cpp	[]	[]
/src/llama.cpp/src/llama-sampling.h	[]	[]
/src/llama.cpp/src/llama-kv-cache.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/llama-memory-hybrid.h	[]	[]
/src/llama.cpp/src/models/olmo.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/llama-vocab.cpp	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/ggml/src/ggml-threading.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/ggml/src/ggml-quants.c	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/common/common.h	['fuzz_inference']	[]
/src/llama.cpp/src/models/starcoder.cpp	[]	[]
/src/llama.cpp/src/models/gemma.cpp	[]	[]
/src/llama.cpp/src/models/starcoder2.cpp	[]	[]
/src/llama.cpp/src/models/chameleon.cpp	[]	[]
/usr/local/bin/../include/c++/v1/stdexcept	['fuzz_grammar', 'fuzz_load_model', 'fuzz_structured', 'fuzz_inference', 'fuzz_json_to_grammar']	[]
/src/llama.cpp/src/llama-chat.cpp	['fuzz_apply_template']	['fuzz_apply_template']
/src/llama.cpp/src/llama-kv-cells.h	['fuzz_inference']	['fuzz_inference']
/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/gpt2.cpp	[]	[]
/src/llama.cpp/src/models/granite.cpp	[]	[]
/src/llama.cpp/src/models/rwkv7.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/binary-ops.cpp	[]	[]
/src/llama.cpp/src/models/dream.cpp	[]	[]
/src/llama.cpp/fuzzers/fuzz_inference.cpp	['fuzz_inference']	['fuzz_inference']
/src/llama.cpp/src/models/apertus.cpp	[]	[]
/src/llama.cpp/src/models/orion.cpp	[]	[]
/src/llama.cpp/src/llama-kv-cache-iswa.h	[]	[]
/src/llama.cpp/src/llama-hparams.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	[]
/src/llama.cpp/src/models/glm4-moe.cpp	[]	[]
/src/llama.cpp/src/models/neo-bert.cpp	[]	[]
/src/llama.cpp/src/models/nemotron.cpp	[]	[]
/src/llama.cpp/src/models/mamba.cpp	[]	[]
/src/llama.cpp/src/models/granite-hybrid.cpp	[]	[]
/src/llama.cpp/src/llama-memory-hybrid.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/models/pangu-embedded.cpp	[]	[]
/src/llama.cpp/src/models/hunyuan-dense.cpp	[]	[]
/src/llama.cpp/src/llama-model-saver.cpp	[]	[]
/src/llama.cpp/src/models/exaone.cpp	[]	[]
/src/llama.cpp/src/models/bitnet.cpp	[]	[]
/src/llama.cpp/src/models/mpt.cpp	[]	[]
/src/llama.cpp/src/models/../llama-graph.h	[]	[]
/src/llama.cpp/src/llama-arch.h	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	[]
/src/llama.cpp/src/models/graph-context-mamba.cpp	[]	[]
/src/llama.cpp/src/models/qwen.cpp	[]	[]
/src/llama.cpp/src/models/exaone4.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/vec.cpp	[]	[]
/src/llama.cpp/src/models/llada.cpp	[]	[]
/src/llama.cpp/src/models/minimax-m2.cpp	[]	[]
/src/llama.cpp/src/models/chatglm.cpp	[]	[]
/src/llama.cpp/src/models/refact.cpp	[]	[]
/src/llama.cpp/src/models/openelm.cpp	[]	[]
/src/llama.cpp/src/llama-io.cpp	[]	[]
/usr/local/bin/../include/c++/v1/__exception/exception.h	['fuzz_json_to_grammar']	[]
/src/llama.cpp/src/llama-batch.cpp	['fuzz_inference']	[]
/src/llama.cpp/src/models/rnd1.cpp	[]	[]
/src/llama.cpp/src/models/plamo.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/quants.c	[]	[]
/src/llama.cpp/common/sampling.cpp	[]	[]
/src/llama.cpp/src/models/ernie4-5-moe.cpp	[]	[]
/src/llama.cpp/fuzzers/fuzz_json_to_grammar.cpp	['fuzz_json_to_grammar']	['fuzz_json_to_grammar']
/src/llama.cpp/src/models/afmoe.cpp	[]	[]
/src/llama.cpp/src/models/deepseek.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-backend-reg.cpp	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']	['fuzz_load_model', 'fuzz_structured', 'fuzz_inference']
/src/llama.cpp/src/models/gemma3n-iswa.cpp	[]	[]
/src/llama.cpp/src/models/qwen3next.cpp	[]	[]
/src/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp	[]	[]
/src/llama.cpp/src/models/falcon.cpp	[]	[]

Directories in report

Directory
/src/llama.cpp/common/../vendor/nlohmann/
/src/llama.cpp/vendor/nlohmann/
/src/llama.cpp/ggml/src/
/src/llama.cpp/src/../include/
/src/llama.cpp/ggml/src/ggml-cpu/llamafile/
/src/llama.cpp/fuzzers/
/usr/local/bin/../include/c++/v1/
/src/llama.cpp/src/models/
/src/llama.cpp/ggml/src/./
/src/llama.cpp/src/models/../
/src/llama.cpp/ggml/src/../include/
/src/llama.cpp/ggml/src/ggml-cpu/arch/x86/
/usr/local/bin/../include/c++/v1/__exception/
/src/llama.cpp/ggml/src/ggml-cpu/
/src//src/
/src/llama.cpp/src/
/src//src/models/
/src/llama.cpp/common/