How to build historgram of methods by time spent inside with Mono? - mono

Question

I have tried the following:
mono --profile=log myprog.exe
to collect profiler data. Then to interpret those I invoke:
> mprof-report output.mlpd
Mono log profiler data
Profiler version: 2.0
Data version: 14
Arguments: log
Architecture: x86-64
Operating system: linux
Mean timer overhead: 51 nanoseconds
Program startup: Fri Jul 20 00:11:12 2018
Program ID: 19840
Server listening on: 59374
JIT summary
Compiled methods: 8349
Generated code size: 2621631
JIT helpers: 0
JIT helpers code size: 0
GC summary
GC resizes: 0
Max heap size: 0
Object moves: 0
Metadata summary
Loaded images: 16
Loaded assemblies: 16
Exception summary
Throws: 0
Thread summary
Thread: 0x7fb49c50a700, name: ""
Thread: 0x7fb49d27b700, name: "Threadpool worker"
Thread: 0x7fb49d07a700, name: "Threadpool worker"
Thread: 0x7fb49ce79700, name: "Threadpool worker"
Thread: 0x7fb49cc78700, name: "Threadpool worker"
Thread: 0x7fb49d6b9700, name: ""
Thread: 0x7fb4bbff1700, name: "Finalizer"
Thread: 0x7fb4bfe3f740, name: "Main"
Domain summary
Domain: (nil), friendly name: "myprog.exe"
Domain: 0x1d037f0, friendly name: "(null)"
Context summary
Context: (nil), domain: (nil)
However, there's no information concerning which methods were called often and took long to complete, which was the only one thing I expected from profiling.
How do I use Mono profiling to gather and output information about method calls' total run time? Like hprof with cpu=times will generate.

Answer 1

The Mono docs are "slightly" wrong as the methods calls are not tracked by default. This option creates huge profile log output and massively slows down "total" execution time and when combined with other options like alloc, effect the execution time of the methods and thus any timings that are being collected.
Personally I would recommend using calls profiling by itself adjusting the calldepthto a level that matters to your profiling. i.e. do you need to profile into the framework calls or not? Also a smaller call depth also greatly decreases the size of the log produced.
Example:
mono --profile=log:calls,calldepth=10 Console_Ling.exe
Produces:
Method call summary
Total(ms) Self(ms) Calls Method name
53358 0 1 (wrapper runtime-invoke) <Module>:runtime_invoke_void_object (object,intptr,intptr,intptr)
53358 2 1 Console_Ling.MainClass:Main (string[])
53340 2 1 Console_Ling.MainClass:Stuff ()
53337 0 3 System.Linq.Enumerable:ToList<int> (System.Collections.Generic.IEnumerable`1<int>)
53194 13347 1 System.Linq.Enumerable/WhereListIterator`1<int>:ToList ()
33110 13181 20000000 Console_Ling.MainClass/<>c__DisplayClass0_0:<Stuff>b__0 (int)
19928 13243 20000000 System.Collections.Generic.List`1<int>:Contains (int)
6685 6685 20000000 System.Collections.Generic.GenericEqualityComparer`1<int>:Equals (int,int)
~~~~
Re: http://www.mono-project.com/docs/debug+profile/profile/profiler/#profiler-option-documentation