Skip to content
TheChatArchiveThe Chat Archive
Qwenqwen3-30b-a3b· Alibaba (Qwen)Hugging Face LMArena Human Preference 140K

I analyzed a simple call with fprof: fprof:trace(start), dnftyvariable:any(),…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with grok-3-mini-beta’s answer.
  1. User

    I analyzed a simple call with fprof:

    fprof:trace(start), dnf_ty_variable:any(), fprof:trace(stop), fprof:profile(), fprof:analyse(),

    This is the result. What are the bottlenecks?

    %% Analysis results: { analysis_options, [{callers, true}, {sort, acc}, {totals, false}, {details, true}]}.

    % CNT ACC OWN
    [{ totals, 311, 21.179, 7.253}]. %%%

    % CNT ACC OWN
    [{ "<0.405.0>", 311,undefined, 7.253}]. %%

    {[{{debug_tests,slow_test,0}, 0, 21.118, 0.008},
    {undefined, 0, 0.030, 0.021},
    {{fprof,call,1}, 1, 0.013, 0.010}],
    { {fprof,just_call,2}, 1, 21.161, 0.039}, % [{{error_handler,undefined_function,3}, 1, 21.110, 0.015},
    {{erlang,demonitor,1}, 1, 0.009, 0.009},
    {{erlang,monitor,2}, 1, 0.003, 0.003},
    {suspend, 1, 0.000, 0.000}]}.

    {[{undefined, 0, 21.149, 0.003}],
    { {debug_tests,slow_test,0}, 0, 21.149, 0.003}, % [{{fprof,just_call,2}, 0, 21.118, 0.008},
    {{fprof,trace,1}, 1, 0.028, 0.004}]}.

    {[{{fprof,just_call,2}, 1, 21.110, 0.015},
    {{ty_tuples,any,0}, 1, 0.000, 0.012},
    {{ty_rec,'-any/0-fun-0-',2}, 6, 0.000, 0.065},
    {{ty_functions,any,0}, 1, 0.000, 0.011},
    {{dnf_ty_variable,any,0}, 1, 0.000, 0.012},
    {{dnf_ty_list,any,0}, 1, 0.000, 0.013}],
    { {error_handler,undefined_function,3}, 11, 21.110, 0.128}, % [{{error_handler,ensure_loaded,1}, 11, 20.666, 0.086},
    {{dnf_ty_variable,any,0}, 1, 18.435, 0.007},
    {{ty_rec,any,0}, 1, 16.288, 0.004},
    {{ty_tuples,any,0}, 1, 2.033, 0.006},
    {{ty_functions,any,0}, 1, 1.991, 0.007},
    {{dnf_ty_list,any,0}, 1, 1.517, 0.006},
    {{erlang,function_exported,3}, 11, 0.032, 0.032},
    {{dnf_ty_tuple,any,0}, 1, 0.008, 0.005},
    {{dnf_ty_function,any,0}, 1, 0.008, 0.005},
    {{ty_bool,any,0}, 1, 0.003, 0.003},
    {{dnf_ty_predefined,any,0}, 1, 0.003, 0.003},
    {{dnf_ty_interval,any,0}, 1, 0.003, 0.003},
    {{dnf_ty_atom,any,0}, 1, 0.003, 0.003}]}.

    {[{{error_handler,undefined_function,3}, 11, 20.666, 0.086}],
    { {error_handler,ensure_loaded,1}, 11, 20.666, 0.086}, % [{{code,ensure_loaded,1}, 11, 20.544, 0.197},
    {{erlang,whereis,1}, 11, 0.036, 0.036}]}.

    {[{{error_handler,ensure_loaded,1}, 11, 20.544, 0.197}],
    { {code,ensure_loaded,1}, 11, 20.544, 0.197}, % [{{code,call,1}, 22, 14.595, 0.081},
    {{code,ensure_prepare_loading,3}, 11, 5.575, 0.043},
    {{code,get_mode,0}, 11, 0.145, 0.044},
    {{erlang,module_loaded,1}, 11, 0.032, 0.032}]}.

    {[{{error_handler,undefined_function,3}, 1, 18.435, 0.007}],
    { {dnf_ty_variable,any,0}, 1, 18.435, 0.007}, % [{{error_handler,undefined_function,3}, 1, 0.000, 0.012}]}.

    {[{{error_handler,undefined_function,3}, 1, 16.288, 0.004}],
    { {ty_rec,any,0}, 1, 16.288, 0.004}, % [{{ty_rec,map,2}, 1, 16.284, 0.010}]}.

    {[{{ty_rec,any,0}, 1, 16.284, 0.010}],
    { {ty_rec,map,2}, 1, 16.284, 0.010}, % [{{lists,foldl,3}, 1, 16.225, 0.008},
    {{lists,zip,2}, 1, 0.036, 0.004},
    {{lists,seq,2}, 1, 0.013, 0.003}]}.

    {[{{ty_rec,map,2}, 1, 16.225, 0.008}],
    { {lists,foldl,3}, 1, 16.225, 0.008}, % [{{lists,foldl_1,3}, 1, 14.660, 0.008},
    {{ty_rec,'-map/2-fun-0-',4}, 1, 1.557, 0.011}]}.

    {[{{lists,foldl_1,3}, 5, 14.619, 0.046},
    {{lists,foldl,3}, 1, 1.557, 0.011}],
    { {ty_rec,'-map/2-fun-0-',4}, 6, 16.176, 0.057}, % [{{ty_rec,'-any/0-fun-0-',2}, 6, 16.089, 0.031},
    {{erlang,setelement,3}, 6, 0.030, 0.030}]}.

    {[{{ty_rec,'-map/2-fun-0-',4}, 6, 16.089, 0.031}],
    { {ty_rec,'-any/0-fun-0-',2}, 6, 16.089, 0.031}, % [{{error_handler,undefined_function,3}, 6, 0.000, 0.065}]}.

    {[{{lists,foldl,3}, 1, 14.660, 0.008},
    {{lists,foldl_1,3}, 5, 0.000, 0.033}],
    { {lists,foldl_1,3}, 6, 14.660, 0.041}, % [{{ty_rec,'-map/2-fun-0-',4}, 5, 14.619, 0.046},
    {{lists,foldl_1,3}, 5, 0.000, 0.033}]}.

    {[{{code,ensure_loaded,1}, 22, 14.595, 0.081}],
    { {code,call,1}, 22, 14.595, 0.081}, % [{{code_server,call,1}, 22, 14.514, 0.397}]}.

    {[{{code,call,1}, 22, 14.514, 0.397}],
    { {code_server,call,1}, 22, 14.514, 0.397}, % [{suspend, 22, 13.926, 0.000},
    {{erlang,monitor,2}, 22, 0.096, 0.096},
    {{erlang,demonitor,2}, 22, 0.095, 0.095}]}.

    {[{{code_server,call,1}, 22, 13.926, 0.000},
    {{fprof,just_call,2}, 1, 0.000, 0.000}],
    { suspend, 23, 13.926, 0.000}, % [ ]}.

    {[{{code,ensure_loaded,1}, 11, 5.575, 0.043}],
    { {code,ensure_prepare_loading,3}, 11, 5.575, 0.043}, % [{{erlang,prepare_loading,2}, 11, 5.532, 0.048}]}.

    {[{{code,ensure_prepare_loading,3}, 11, 5.532, 0.048}],
    { {erlang,prepare_loading,2}, 11, 5.532, 0.048}, % [{{erlang,prepare_loading_1,2}, 11, 5.484, 0.091}]}.

    {[{{erlang,prepare_loading,2}, 11, 5.484, 0.091}],
    { {erlang,prepare_loading_1,2}, 11, 5.484, 0.091}, % [{{erts_internal,prepare_loading,2}, 11, 5.393, 5.393}]}.

    {[{{erlang,prepare_loading_1,2}, 11, 5.393, 5.393}],
    { {erts_internal,prepare_loading,2}, 11, 5.393, 5.393}, % [ ]}.

    {[{{error_handler,undefined_function,3}, 1, 2.033, 0.006}],
    { {ty_tuples,any,0}, 1, 2.033, 0.006}, % [{{error_handler,undefined_function,3}, 1, 0.000, 0.012}]}.

    {[{{error_handler,undefined_function,3}, 1, 1.991, 0.007}],
    { {ty_functions,any,0}, 1, 1.991, 0.007}, % [{{error_handler,undefined_function,3}, 1, 0.000, 0.011}]}.

    {[{{error_handler,undefined_function,3}, 1, 1.517, 0.006}],
    { {dnf_ty_list,any,0}, 1, 1.517, 0.006}, % [{{error_handler,undefined_function,3}, 1, 0.000, 0.013}]}.

    {[{{code,ensure_loaded,1}, 11, 0.145, 0.044}],
    { {code,get_mode,0}, 11, 0.145, 0.044}, % [{{code_server,get_mode,0}, 11, 0.101, 0.065}]}.

    {[{{code,get_mode,0}, 11, 0.101, 0.065}],
    { {code_server,get_mode,0}, 11, 0.101, 0.065}, % [{{persistent_term,get,1}, 11, 0.036, 0.036}]}.

    {[{{code_server,call,1}, 22, 0.096, 0.096},
    {{fprof,just_call,2}, 1, 0.003, 0.003}],
    { {erlang,monitor,2}, 23, 0.099, 0.099}, % [ ]}.

    {[{{code_server,call,1}, 22, 0.095, 0.095}],
    { {erlang,demonitor,2}, 22, 0.095, 0.095}, % [ ]}.

    {[{{error_handler,ensure_loaded,1}, 11, 0.036, 0.036},
    {{fprof,call,1}, 1, 0.004, 0.004}],
    { {erlang,whereis,1}, 12, 0.040, 0.040}, % [ ]}.

    {[{{code_server,get_mode,0}, 11, 0.036, 0.036}],
    { {persistent_term,get,1}, 11, 0.036, 0.036}, % [ ]}.

    {[{{ty_rec,map,2}, 1, 0.036, 0.004}],
    { {lists,zip,2}, 1, 0.036, 0.004}, % [{{lists,zip,3}, 1, 0.032, 0.009}]}.

    {[{{lists,zip,2}, 1, 0.032, 0.009},
    {{lists,zip,3}, 6, 0.000, 0.023}],
    { {lists,zip,3}, 7, 0.032, 0.032}, % [{{lists,zip,3}, 6, 0.000, 0.023}]}.

    {[{{code,ensure_loaded,1}, 11, 0.032, 0.032}],
    { {erlang,module_loaded,1}, 11, 0.032, 0.032}, % [ ]}.

    {[{{error_handler,undefined_function,3}, 11, 0.032, 0.032}],
    { {erlang,function_exported,3}, 11, 0.032, 0.032}, % [ ]}.

    {[{{ty_rec,'-map/2-fun-0-',4}, 6, 0.030, 0.030}],
    { {erlang,setelement,3}, 6, 0.030, 0.030}, % [ ]}.

    {[{{debug_tests,slow_test,0}, 1, 0.028, 0.004}],
    { {fprof,trace,1}, 1, 0.028, 0.004}, % [{{fprof,call,1}, 1, 0.024, 0.007}]}.

    {[{{fprof,trace,1}, 1, 0.024, 0.007}],
    { {fprof,call,1}, 1, 0.024, 0.007}, % [{{fprof,just_call,2}, 1, 0.013, 0.010},
    {{erlang,whereis,1}, 1, 0.004, 0.004}]}.

    {[{{ty_rec,map,2}, 1, 0.013, 0.003}],
    { {lists,seq,2}, 1, 0.013, 0.003}, % [{{lists,seq_loop,3}, 1, 0.010, 0.004}]}.

    {[{{lists,seq,2}, 1, 0.010, 0.004},
    {{lists,seq_loop,3}, 2, 0.000, 0.006}],
    { {lists,seq_loop,3}, 3, 0.010, 0.010}, % [{{lists,seq_loop,3}, 2, 0.000, 0.006}]}.

    {[{{error_handler,undefined_function,3}, 1, 0.003, 0.003},
    {{dnf_ty_tuple,any,0}, 1, 0.003, 0.003},
    {{dnf_ty_function,any,0}, 1, 0.003, 0.003}],
    { {ty_bool,any,0}, 3, 0.009, 0.009}, % [ ]}.

    {[{{fprof,just_call,2}, 1, 0.009, 0.009}],
    { {erlang,demonitor,1}, 1, 0.009, 0.009}, % [ ]}.

    {[{{error_handler,undefined_function,3}, 1, 0.008, 0.005}],
    { {dnf_ty_tuple,any,0}, 1, 0.008, 0.005}, % [{{ty_bool,any,0}, 1, 0.003, 0.003}]}.

    {[{{error_handler,undefined_function,3}, 1, 0.008, 0.005}],
    { {dnf_ty_function,any,0}, 1, 0.008, 0.005}, % [{{ty_bool,any,0}, 1, 0.003, 0.003}]}.

    {[{{error_handler,undefined_function,3}, 1, 0.003, 0.003}],
    { {dnf_ty_predefined,any,0}, 1, 0.003, 0.003}, % [ ]}.

    {[{{error_handler,undefined_function,3}, 1, 0.003, 0.003}],
    { {dnf_ty_interval,any,0}, 1, 0.003, 0.003}, % [ ]}.

    {[{{error_handler,undefined_function,3}, 1, 0.003, 0.003}],
    { {dnf_ty_atom,any,0}, 1, 0.003, 0.003}, % [ ]}.

    {[ ], { undefined, 0, 0.000, 0.000}, % [{{debug_tests,slow_test,0}, 0, 21.149, 0.003},
    {{fprof,just_call,2}, 0, 0.030, 0.021}]}.

  2. Qwen Assistant

    Based on the fprof analysis, the primary bottlenecks are code loading and error handling overhead caused by the call to dnf_ty_variable:any(). Here's a breakdown of the key issues:


    1. Error Handling for Undefined Functions

    • Key Functions Involved:
      • {error_handler, undefined_function, 3} (11 calls, ACC = 21.110)
      • {error_handler, ensure_loaded, 1} (11 calls, ACC = 20.666)
    • What's Happening:
      • The system repeatedly checks if the module containing dnf_ty_variable:any() is loaded.
      • If the module is not found, it triggers error handling (e.g., undefined_function checks), which is expensive.
      • This suggests the module might not be properly compiled or available in the code path.

    2. Code Loading Overhead

    • Key Functions Involved:
      • {code, ensure_loaded, 1} (11 calls, ACC = 20.544)
      • {code, ensure_prepare_loading, 3} (11 calls, ACC = 5.575)
      • {erts_internal, prepare_loading, 2} (11 calls, ACC = 5.393)
    • What's Happening:
      • The system attempts to load the module multiple times, which involves checking for dependencies, preparing the module, and handling errors.
      • This is likely due to the module not being preloaded or being missing from the code path.

    3. List and Data Structure Processing

    • Key Functions Involved:
      • {lists,foldl,3} (ACC = 14.660)
      • {ty_rec,map,2} (ACC = 16.284)
      • {ty_rec,'-map/2-fun-0-',4} (ACC = 16.089)
    • What's Happening:
      • These functions are part of the internal logic of dnf_ty_variable:any() (or its dependencies).
      • While their individual ACC values are lower than the error/loading overhead, they contribute to the total runtime.

    4. Process Monitoring and Synchronization

    • Key Functions Involved:
      • {erlang, monitor, 2} (22 calls, ACC = 0.096)
      • {erlang, demonitor, 2} (22 calls, ACC = 0.095)
    • What's Happening:
      • The system uses process monitoring (e.g., monitor/2) for code loading or error handling, which adds overhead.

    Root Cause

    The primary bottleneck is the system attempting to load a module that is either missing or not properly compiled, triggering repeated error checks and code-loading operations. The actual dnf_ty_variable:any() function itself appears to be lightweight (ACC = 0.000), but the surrounding infrastructure (error handling, code loading) is the real performance drag.


    Recommendations

    1. Ensure the Module is Preloaded:
      • Compile and load the dnf_ty_variable module explicitly before calling any().
      • Use code:ensure_loaded/1 or erlang:load_module/2 if necessary.
    2. Check for Module Availability:
      • Verify that the module is in the Erlang code path (CODEPATH).
    3. Optimize Error Handling:
      • If the function is expected to fail, handle errors explicitly rather than relying on the system's default error handling.
    4. Profile the Module:
      • Run fprof on the dnf_ty_variable module directly to isolate its internal logic (e.g., lists:folding or ty_rec:map).

    Summary of Bottlenecks

    CategoryFunctionCallsACCNotes
    Error Handlingerror_handler:undefined_function/31121.110Repeated checks for missing module
    Code Loadingcode:ensure_loaded/11120.544Module loading overhead
    Internal Logiclists:foldl/3114.660Data structure processing
    Process Monitoringerlang:monitor/2220.096Synchronization overhead

    The error handling and code loading are the dominant performance issues here. Addressing these will likely resolve the bottlenecks.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.