Dear Nigel and all,
I just used 'perftools' to profile my parser, and the result is:
Total: 1570 samples
855 54.5% 54.5% 855 54.5% garbage_collector
338 21.5% 76.0% 338 21.5% Hash#[]=
125 8.0% 83.9% 234 14.9%
Parslet::Atoms::Named#produce_return_value
36 2.3% 86.2% 36 2.3% Hash#initialize_copy
36 2.3% 88.5% 89 5.7%
Parslet::Atoms::CanFlatten#flatten_sequence
21 1.3% 89.9% 715 45.5% Enumerable#each_with_index
20 1.3% 91.1% 715 45.5% Parslet::Atoms::Context#try_with_cache
10 0.6% 91.8% 10 0.6% Array#initialize
10 0.6% 92.4% 30 1.9% Parslet::Atoms::Context#lookup
9 0.6% 93.0% 715 45.5% Parslet::Atoms::Base#apply
9 0.6% 93.6% 10 0.6%
Parslet::Atoms::CanFlatten#flatten_repetition
9 0.6% 94.1% 715 45.5% Parslet::Atoms::Sequence#try
7 0.4% 94.6% 7 0.4% Array#&
7 0.4% 95.0% 695 44.3% Array#map
7 0.4% 95.5% 10 0.6% Parslet::Source#pos
5 0.3% 95.8% 5 0.3% Parslet::Atoms::Base#succ
5 0.3% 96.1% 715 45.5% Parslet::Atoms::Named#apply
5 0.3% 96.4% 13 0.8% Parslet::Atoms::Str#try
4 0.3% 96.7% 109 6.9% Parslet::Atoms::CanFlatten#flatten
4 0.3% 96.9% 325 20.7% Parslet::Atoms::Context#set
4 0.3% 97.2% 5 0.3% Parslet::Source#consume
3 0.2% 97.4% 3 0.2% Array#[]=
3 0.2% 97.6% 3 0.2% Array#join
3 0.2% 97.8% 53 3.4% Enumerable#inject
3 0.2% 98.0% 3 0.2% Kernel#hash
3 0.2% 98.2% 3 0.2% StringScanner#pos
2 0.1% 98.3% 2 0.1% Hash#keys
2 0.1% 98.4% 49 3.1% Parslet::Atoms::CanFlatten#merge_fold
2 0.1% 98.5% 2 0.1% Parslet::Atoms::Context#err_at
2 0.1% 98.7% 51 3.2% Parslet::Atoms::Mark#try
2 0.1% 98.8% 8 0.5% Parslet::Atoms::Re#try
2 0.1% 98.9% 565 36.0% Parslet::Atoms::Repetition#try
2 0.1% 99.0% 4 0.3% Parslet::Source#pos=
2 0.1% 99.2% 2 0.1% StringScanner#match?
1 0.1% 99.2% 1 0.1% Array#[]
1 0.1% 99.3% 37 2.4% Hash#merge
1 0.1% 99.4% 1 0.1% Kernel#is_a?
1 0.1% 99.4% 566 36.1% Kernel#loop
1 0.1% 99.5% 691 44.0% Parslet::Atoms::Alternative#try
1 0.1% 99.6% 1 0.1% Parslet::Atoms::Base#cached?
1 0.1% 99.6% 1 0.1% Parslet::Atoms::Context#err
1 0.1% 99.7% 715 45.5% Parslet::Atoms::Entity#try
1 0.1% 99.7% 2 0.1% Parslet::Atoms::Named#to_s_inner
1 0.1% 99.8% 1 0.1% Parslet::Slice#+
1 0.1% 99.9% 1 0.1% StringScanner#pos=
1 0.1% 99.9% 1 0.1% StringScanner#rest_size
1 0.1% 100.0% 1 0.1% StringScanner#scan
0 0.0% 100.0% 1 0.1% Array#inspect
0 0.0% 100.0% 4 0.3% BasicObject#instance_eval
0 0.0% 100.0% 2 0.1% FortranParser#ranged
0 0.0% 100.0% 36 2.3% Kernel#initialize_dup
0 0.0% 100.0% 1 0.1% Parslet::Atoms::Alternative#initialize
0 0.0% 100.0% 4 0.3% Parslet::Atoms::Base#inspect
0 0.0% 100.0% 1 0.1% Parslet::Atoms::Base#parse
0 0.0% 100.0% 1 0.1% Parslet::Atoms::Base#setup_and_apply
0 0.0% 100.0% 4 0.3% Parslet::Atoms::Base#to_s
0 0.0% 100.0% 54 3.4% Parslet::Atoms::CanFlatten#foldl
0 0.0% 100.0% 9 0.6%
Parslet::Atoms::CanFlatten#warn_about_duplicate_keys
0 0.0% 100.0% 2 0.1% Parslet::Atoms::DSL#>>
0 0.0% 100.0% 1 0.1% Parslet::Atoms::DSL#|
0 0.0% 100.0% 4 0.3% Parslet::Atoms::Entity#parslet
0 0.0% 100.0% 37 2.4% Parslet::Atoms::Lookahead#try
0 0.0% 100.0% 1 0.1% Parslet::Atoms::Repetition#to_s_inner
0 0.0% 100.0% 1 0.1% Parslet::Atoms::Sequence#>>
0 0.0% 100.0% 3 0.2% Parslet::Atoms::Sequence#initialize
0 0.0% 100.0% 3 0.2% Parslet::Atoms::Sequence#to_s_inner
0 0.0% 100.0% 1 0.1% Parslet::Parser#try
0 0.0% 100.0% 1 0.1% Parslet::Source#chars_left
0 0.0% 100.0% 2 0.1% Parslet::Source#matches?
0 0.0% 100.0% 17 1.1% Proc#yield
Any idea?
Cheers
Li
在 2013-7-1，上午11:53，Nigel Thorne <[email protected]> 写道：
> just google it
>
> http://stackoverflow.com/questions/4092641/profiling-ruby-code
>
> ---
> "No man is an island... except Philip"
>
>
> On Mon, Jul 1, 2013 at 1:44 PM, Li Dong <[email protected]> wrote:
> Oh! Could you tell me what is the best tool can I use to draw a table of time
> cost? Thanks!
>
> Cheers
> Li
>
> 在 2013-7-1，上午11:27，Nigel Thorne <[email protected]> 写道：
>
>> no sorry. I meant.. "Have you run your code through a profiler to find the
>> bottlenecks?"
>>
>> ie. What percentage of your time is being spent on each method?
>>
>> Cheers
>> Nigel
>>
>> ---
>> "No man is an island... except Philip"
>>
>>
>> On Mon, Jul 1, 2013 at 12:59 PM, Li Dong <[email protected]> wrote:
>> Hi Nigel,
>>
>> Yes, the timing result is already outputted by the 'rspec' command. Here is
>> in my environment:
>>
>> [Notice]: Test fortran code parsing
>> ---> Parsing uses 17.100005 seconds.
>> ---> Converting uses 1.838105 seconds.
>>
>> Best,
>> Li
>>
>> 在 2013-7-1，上午10:46，Nigel Thorne <[email protected]> 写道：
>>
>>> Please can you include your profiler results.
>>>
>>> Cheers
>>> Nigel
>>>
>>> ---
>>> "No man is an island... except Philip"
>>>
>>>
>>> On Mon, Jul 1, 2013 at 12:51 AM, Li Dong <[email protected]> wrote:
>>> Dear all,
>>>
>>> After several hard working days, I have implemented a Fortran parser.
>>> Although
>>> it is still incomplete, the most thing that I am worrying about is the
>>> performance. I have tried it on a Fortran code with 2000+ lines, and it took
>>> around 20 seconds on my MacBook Pro for parsing. This can not be practical.
>>> So
>>> I should optimize the parser, but I have no idea where to start, and which
>>> parts should be heavily optimized.
>>>
>>> The parser is located in htt ps://gist.github.com/dongli/5791976, and the
>>> example
>>> can be run as:
>>>
>>> rspec rspec_fortran_parser.rb
>>>
>>> Any idea is appreciated!
>>> Best regards,
>>>
>>> Li
>>>
>>>
>>
>>
>
>
