compiled llama.cpp on a Jetson AGX Orin, but when running Qwen3-VL, the inference speed is extremely slow.

[hezichuanqi]

I compiled llama.cpp on a Jetson AGX Orin, but when running Qwen3-VL, the inference speed is extremely slow.
Below is the complete inference log. Does anyone know what the problem is?Are there any points to pay attention to when compiling on AGX orin? Help me,thank u!
--------------------------------------------------------LOG---------------------------------------------------
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 1 CUDA devices:
Device 0: Orin, compute capability 8.7, VMM: yes
build: 7193 (d82b7a7) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for aarch64-linux-gnu
llama_model_load_from_file_impl: using device CUDA0 (Orin) (0000:00:00.0) - 22310 MiB free
llama_model_loader: loaded meta data with 32 key-value pairs and 398 tensors from ./Qwen3VL-4B-Instruct-Q4_K_M.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = qwen3vl
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Qwen3Vl 4b Instruct
llama_model_loader: - kv 3: general.finetune str = instruct
llama_model_loader: - kv 4: general.basename str = qwen3vl
llama_model_loader: - kv 5: general.size_label str = 4B
llama_model_loader: - kv 6: general.license str = apache-2.0
llama_model_loader: - kv 7: general.tags arr[str,1] = ["image-text-to-text"]
llama_model_loader: - kv 8: qwen3vl.block_count u32 = 36
llama_model_loader: - kv 9: qwen3vl.context_length u32 = 262144
llama_model_loader: - kv 10: qwen3vl.embedding_length u32 = 2560
llama_model_loader: - kv 11: qwen3vl.feed_forward_length u32 = 9728
llama_model_loader: - kv 12: qwen3vl.attention.head_count u32 = 32
llama_model_loader: - kv 13: qwen3vl.attention.head_count_kv u32 = 8
llama_model_loader: - kv 14: qwen3vl.rope.freq_base f32 = 5000000.000000
llama_model_loader: - kv 15: qwen3vl.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 16: qwen3vl.attention.key_length u32 = 128
llama_model_loader: - kv 17: qwen3vl.attention.value_length u32 = 128
llama_model_loader: - kv 18: qwen3vl.rope.dimension_sections arr[i32,4] = [24, 20, 20, 0]
llama_model_loader: - kv 19: qwen3vl.n_deepstack_layers u32 = 3
llama_model_loader: - kv 20: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 21: tokenizer.ggml.pre str = qwen2
llama_model_loader: - kv 22: tokenizer.ggml.tokens arr[str,151936] = ["!", """, "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 23: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 24: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv 25: tokenizer.ggml.eos_token_id u32 = 151645
llama_model_loader: - kv 26: tokenizer.ggml.padding_token_id u32 = 151643
llama_model_loader: - kv 27: tokenizer.ggml.bos_token_id u32 = 151643
llama_model_loader: - kv 28: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 29: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>...
llama_model_loader: - kv 30: general.quantization_version u32 = 2
llama_model_loader: - kv 31: general.file_type u32 = 15
llama_model_loader: - type f32: 145 tensors
llama_model_loader: - type q4_K: 216 tensors
llama_model_loader: - type q6_K: 37 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q4_K - Medium
print_info: file size = 2.32 GiB (4.95 BPW)
load: printing all EOG tokens:
load: - 151643 ('<|endoftext|>')
load: - 151645 ('<|im_end|>')
load: - 151662 ('<|fim_pad|>')
load: - 151663 ('<|repo_name|>')
load: - 151664 ('<|file_sep|>')
load: special tokens cache size = 26
load: token to piece cache size = 0.9311 MB
print_info: arch = qwen3vl
print_info: vocab_only = 0
print_info: n_ctx_train = 262144
print_info: n_embd = 2560
print_info: n_embd_inp = 10240
print_info: n_layer = 36
print_info: n_head = 32
print_info: n_head_kv = 8
print_info: n_rot = 128
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 128
print_info: n_embd_head_v = 128
print_info: n_gqa = 4
print_info: n_embd_k_gqa = 1024
print_info: n_embd_v_gqa = 1024
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-06
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 0.0e+00
print_info: n_ff = 9728
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: n_expert_groups = 0
print_info: n_group_used = 0
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 40
print_info: rope scaling = linear
print_info: freq_base_train = 5000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 262144
print_info: rope_finetuned = unknown
print_info: mrope sections = [24, 20, 20, 0]
print_info: model type = 4B
print_info: model params = 4.02 B
print_info: general.name = Qwen3Vl 4b Instruct
print_info: vocab type = BPE
print_info: n_vocab = 151936
print_info: n_merges = 151387
print_info: BOS token = 151643 '<|endoftext|>'
print_info: EOS token = 151645 '<|im_end|>'
print_info: EOT token = 151645 '<|im_end|>'
print_info: PAD token = 151643 '<|endoftext|>'
print_info: LF token = 198 'Ċ'
print_info: FIM PRE token = 151659 '<|fim_prefix|>'
print_info: FIM SUF token = 151661 '<|fim_suffix|>'
print_info: FIM MID token = 151660 '<|fim_middle|>'
print_info: FIM PAD token = 151662 '<|fim_pad|>'
print_info: FIM REP token = 151663 '<|repo_name|>'
print_info: FIM SEP token = 151664 '<|file_sep|>'
print_info: EOG token = 151643 '<|endoftext|>'
print_info: EOG token = 151645 '<|im_end|>'
print_info: EOG token = 151662 '<|fim_pad|>'
print_info: EOG token = 151663 '<|repo_name|>'
print_info: EOG token = 151664 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 36 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 37/37 layers to GPU
load_tensors: CPU_Mapped model buffer size = 304.28 MiB
load_tensors: CUDA0 model buffer size = 2375.91 MiB
................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 1
llama_context: n_ctx = 4096
llama_context: n_ctx_seq = 4096
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = auto
llama_context: kv_unified = false
llama_context: freq_base = 5000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_seq (4096) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
llama_context: CUDA_Host output buffer size = 0.58 MiB
llama_kv_cache: CUDA0 KV buffer size = 576.00 MiB
llama_kv_cache: size = 576.00 MiB ( 4096 cells, 36 layers, 1/1 seqs), K (f16): 288.00 MiB, V (f16): 288.00 MiB
llama_context: Flash Attention was auto, set to enabled
llama_context: CUDA0 compute buffer size = 301.75 MiB
llama_context: CUDA_Host compute buffer size = 13.02 MiB
llama_context: graph nodes = 1267
llama_context: graph splits = 2
common_init_from_params: added <|endoftext|> logit bias = -inf
common_init_from_params: added <|im_end|> logit bias = -inf
common_init_from_params: added <|fim_pad|> logit bias = -inf
common_init_from_params: added <|repo_name|> logit bias = -inf
common_init_from_params: added <|file_sep|> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 4096
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
mtmd_cli_context: chat template example:
<|im_start|>system
You are a helpful assistant<|im_end|>
<|im_start|>user
Hello<|im_end|>
<|im_start|>assistant
Hi there<|im_end|>
<|im_start|>user
How are you?<|im_end|>
<|im_start|>assistant

clip_model_loader: model name: Qwen3Vl 4b Instruct
clip_model_loader: description:
clip_model_loader: GGUF version: 3
clip_model_loader: alignment: 32
clip_model_loader: n_tensors: 316
clip_model_loader: n_kv: 25

clip_model_loader: has vision encoder
clip_ctx: CLIP using CUDA0 backend
load_hparams: Qwen-VL models require at minimum 1024 image tokens to function correctly on grounding tasks
load_hparams: if you encounter problems with accuracy, try adding --image-min-tokens 1024
load_hparams: more info: #16842

load_hparams: projector: qwen3vl_merger
load_hparams: n_embd: 1024
load_hparams: n_head: 16
load_hparams: n_ff: 4096
load_hparams: n_layer: 24
load_hparams: ffn_op: gelu
load_hparams: projection_dim: 2560

--- vision hparams ---
load_hparams: image_size: 768
load_hparams: patch_size: 16
load_hparams: has_llava_proj: 0
load_hparams: minicpmv_version: 0
load_hparams: n_merge: 2
load_hparams: n_wa_pattern: 0
load_hparams: image_min_pixels: 8192
load_hparams: image_max_pixels: 4194304

load_hparams: model size: 797.43 MiB
load_hparams: metadata size: 0.11 MiB
alloc_compute_meta: warmup with image size = 1472 x 1472
alloc_compute_meta: CUDA0 compute buffer size = 322.49 MiB
alloc_compute_meta: CPU compute buffer size = 24.93 MiB
alloc_compute_meta: graph splits = 1, nodes = 766
warmup: flash attention is enabled
main: loading model: ./Qwen3VL-4B-Instruct-Q4_K_M.gguf
encoding image slice...
image slice encoded in 4079 ms
decoding image batch 1/1, n_tokens_batch = 2040
image decoded (batch 1/1) in 1181 ms

llama_perf_context_print: load time = 1323.52 ms
llama_perf_context_print: prompt eval time = 5736.27 ms / 2060 tokens ( 2.78 ms per token, 359.12 tokens per second)
llama_perf_context_print: eval time = 2690.64 ms / 88 runs ( 30.58 ms per token, 32.71 tokens per second)
llama_perf_context_print: total time = 9112.27 ms / 2148 tokens