gistfile1.txt

(bitnet-cpp) willy@linux2025:~/Desktop/linux2025/term-project/BitNet$ uftrace record ./build/bin/llama-cli   -m models/BitNet-b1.58-2B-4T/ggml-model-i2_s.gguf   -p "why sky blue" -n 50 --temp 0
(bitnet-cpp) willy@linux2025:~/Desktop/linux2025/term-project/BitNet$ uftrace report


Total time   Self time       Calls  Function
  ==========  ==========  ==========  ====================
    1.057  m  118.955 us          10  ggml_graph_compute_secondary_thread
    1.056  m   85.570 ms         306  ggml_graph_compute_thread
    1.050  m  134.310 ms      341496  ggml_compute_forward
    1.049  m  453.917 ms       82926  ggml_compute_forward_mul_mat
    1.047  m    1.953 ms         730  std::condition_variable::wait
    1.006  m    6.459  s      715668  ggml_compute_forward_mul_mat_one_chunk
   59.627  s    8.976 us           2  llama_load_model_from_file
   53.581  s    9.249 ms       53676  std::__invoke_impl
   53.532  s   53.532  s         475  linux:schedule
   53.522  s  488.271 us           1  main
   53.511  s    0.394 us           1  std::thread::_State_impl::_M_run
   53.511  s    0.243 us           1  std::thread::_Invoker::operator()
   53.511  s    0.233 us           1  std::thread::_Invoker::_M_invoke
   53.511  s    0.222 us           1  std::__invoke
   53.511  s  356.595 us           1  common_log::resume::$_0::operator()
   41.933  s   51.806 us         102  llama_decode
   40.352  s   37.444  s    37528064  ggml_vec_dot_i2_i8_s
   38.564  s   20.057  s      424062  ggml_barrier
   30.567  s   10.666 us           1  common_init_from_params
   29.813  s    3.619 us           1  llama_model_load
   25.760  s  655.968 ms           1  llm_load_vocab
   23.430  s   23.362  s   219431233  ggml_thread_cpu_relax
   20.966  s  565.998 us          51  llama_decode_internal
   20.692  s  188.142 us         265  ggml_graph_compute_check_for_work
   20.691  s   12.213  s         265  ggml_graph_compute_poll_for_work
   19.392  s  134.185 us          51  llama_graph_compute
   19.392  s   14.396 us          51  ggml_backend_sched_graph_compute_async
   19.392  s   43.438 us          51  ggml_backend_sched_compute_splits
   19.392  s   17.850 us          51  ggml_backend_graph_compute_async
   19.392  s   53.413 us          51  ggml_backend_cpu_graph_compute
   19.343  s   36.156 us          51  ggml_graph_compute
   18.876  s   18.866  s    11763456  ggml_vec_dot_f16
   10.305  s  242.974 ms      280147  std::map::emplace
    7.129  s   27.913 ms      283667  std::map::lower_bound
    7.101  s   70.411 ms      283667  std::_Rb_tree::lower_bound
    7.087  s    4.670  s       11538  quantize_row_i8_s
    6.977  s    1.007  s      287758  std::_Rb_tree::_M_lower_bound
    6.603  s   20.279 ms      128256  llama_token_to_piece::cxx11
    6.583  s   13.725 ms      130209  llama_token_to_piece
    6.569  s   68.857 ms      130155  llama_token_to_piece_impl
    6.351  s  338.351 ms      129591  llama_decode_text
    6.254  s    2.722  s    17604382  std::operator<
    4.376  s   76.622 ms      871921  unicode_utf8_to_byte
    4.298  s   77.186 ms      872268  std::unordered_map::at
    4.221  s  132.559 ms      872268  std::__detail::_Map_base::at
    4.174  s  565.350 ms     5865830  std::less::operator()
    4.174  s  508.574 ms      409750  unicode_cpts_from_utf8
    3.874  s  380.694 ms      873171  std::_Hashtable::find
    3.770  s    1.000 ms           1  llama_model_loader::llama_model_loader
    3.540  s  431.597 ms          44  gguf_kv_to_str::cxx11