Nanosecond-level profiling in Pharo?
Locked
Nanosecond-level profiling in Pharo?
 
|
Hi!
I am currently working on a new profiler for Pharo. The idea is to graphically renders a profiling using Mondrian. I hope to release it soon. When discussing with an expert in the field (Walter Binder), the question about time grain was raised. I am partly based on MessageTally, which has a millisecond grain. Anyone is aware of a finer grain? At the level of nanosecond? Does it makes sense since there is no JIT (yet)? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
2010/1/7 Alexandre Bergel <[hidden email]>:
> Hi! > > I am currently working on a new profiler for Pharo. The idea is to > graphically renders a profiling using Mondrian. I hope to release it > soon. > When discussing with an expert in the field (Walter Binder), the > question about time grain was raised. I am partly based on > MessageTally, which has a millisecond grain. Anyone is aware of a > finer grain? At the level of nanosecond? Does it makes sense since > there is no JIT (yet)? > AFAIK, there's no primitives in VM, which provide timing with nanosecond precision. > Cheers, > Alexandre > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- Best regards, Igor Stasenko AKA sig. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
Ok, good to know.
Thanks Igor, Alexandre On 7 Jan 2010, at 19:29, Igor Stasenko wrote: > 2010/1/7 Alexandre Bergel <[hidden email]>: >> Hi! >> >> I am currently working on a new profiler for Pharo. The idea is to >> graphically renders a profiling using Mondrian. I hope to release it >> soon. >> When discussing with an expert in the field (Walter Binder), the >> question about time grain was raised. I am partly based on >> MessageTally, which has a millisecond grain. Anyone is aware of a >> finer grain? At the level of nanosecond? Does it makes sense since >> there is no JIT (yet)? >> > > AFAIK, there's no primitives in VM, which provide timing with > nanosecond precision. > > >> Cheers, >> Alexandre >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > > > -- > Best regards, > Igor Stasenko AKA sig. > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
|
|
In reply to this post by Alexandre Bergel
Some times I wanted smaller grain for running tests... but most of the
time millis are good enough :) On 2010-01-07 18:52:16 -0300, Alexandre Bergel <[hidden email]> said: > Hi! > > I am currently working on a new profiler for Pharo. The idea is to > graphically renders a profiling using Mondrian. I hope to release it > soon. > When discussing with an expert in the field (Walter Binder), the > question about time grain was raised. I am partly based on > MessageTally, which has a millisecond grain. Anyone is aware of a > finer grain? At the level of nanosecond? Does it makes sense since > there is no JIT (yet)? > > Cheers, > Alexandre _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Igor Stasenko
On Fri, Jan 08, 2010 at 12:29:18AM +0200, Igor Stasenko wrote:
> 2010/1/7 Alexandre Bergel <[hidden email]>: > > Hi! > > > > I am currently working on a new profiler for Pharo. The idea is to > > graphically renders a profiling using Mondrian. I hope to release it > > soon. > > When discussing with an expert in the field (Walter Binder), the > > question about time grain was raised. I am partly based on > > MessageTally, which has a millisecond grain. Anyone is aware of a > > finer grain? At the level of nanosecond? Does it makes sense since > > there is no JIT (yet)? > > > > AFAIK, there's no primitives in VM, which provide timing with > nanosecond precision. Also, the accuracy of the #millisecondClockValue available to the image is different on different platforms. I haven't checked lately, but I think the Windows VM is good pretty much to the nearest millisecond, while Linux is more variable. Not sure about OS X or other unixy platforms. So even if you are dealing with times that are measured to millisecond precision, the accuracy may be less depending on the underlying platform and VM. Dave _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
|
|
In reply to this post by EstebanLM
I think Eliot wanted those in my vm work I think qwaq has them
time to add them I think should it return a Long integer or a double On 1/7/10, Esteban Lorenzano <[hidden email]> wrote: > Some times I wanted smaller grain for running tests... but most of the > time millis are good enough :) > > On 2010-01-07 18:52:16 -0300, Alexandre Bergel > <[hidden email]> said: > >> Hi! >> >> I am currently working on a new profiler for Pharo. The idea is to >> graphically renders a profiling using Mondrian. I hope to release it >> soon. >> When discussing with an expert in the field (Walter Binder), the >> question about time grain was raised. I am partly based on >> MessageTally, which has a millisecond grain. Anyone is aware of a >> finer grain? At the level of nanosecond? Does it makes sense since >> there is no JIT (yet)? >> >> Cheers, >> Alexandre > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- =========================================================================== John M. McIntosh <[hidden email]> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com =========================================================================== _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
Thanks for all your answers.
cheers, Alexandre On 8 Jan 2010, at 01:22, John McIntosh wrote: > I think Eliot wanted those in my vm work I think qwaq has them > time to add them I think > > should it return a Long integer or a double > > On 1/7/10, Esteban Lorenzano <[hidden email]> wrote: >> Some times I wanted smaller grain for running tests... but most of >> the >> time millis are good enough :) >> >> On 2010-01-07 18:52:16 -0300, Alexandre Bergel >> <[hidden email]> said: >> >>> Hi! >>> >>> I am currently working on a new profiler for Pharo. The idea is to >>> graphically renders a profiling using Mondrian. I hope to release it >>> soon. >>> When discussing with an expert in the field (Walter Binder), the >>> question about time grain was raised. I am partly based on >>> MessageTally, which has a millisecond grain. Anyone is aware of a >>> finer grain? At the level of nanosecond? Does it makes sense since >>> there is no JIT (yet)? >>> >>> Cheers, >>> Alexandre >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > > -- > = > = > = > = > = > ====================================================================== > John M. McIntosh <[hidden email]> > Corporate Smalltalk Consulting Ltd. http:// > www.smalltalkconsulting.com > = > = > = > = > = > ====================================================================== > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
|
|
In reply to this post by Alexandre Bergel
I don't want to rain in your folks party, but a nanosecond is a millionth of a millisecond. Also, it is at the timing of silicon memory access on present microcomputers.
Isn't enough you increase the accuracy of the timing by one thousand looking for a solution to get *microsecond* resolution? HTH, -- Cesar Rabak Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > escreveu: Ok, good to know. Thanks Igor, Alexandre On 7 Jan 2010, at 19:29, Igor Stasenko wrote: > 2010/1/7 Alexandre Bergel : >> Hi! >> >> I am currently working on a new profiler for Pharo. The idea is to >> graphically renders a profiling using Mondrian. I hope to release it >> soon. >> When discussing with an expert in the field (Walter Binder), the >> question about time grain was raised. I am partly based on >> MessageTally, which has a millisecond grain. Anyone is aware of a >> finer grain? At the level of nanosecond? Does it makes sense since >> there is no JIT (yet)? >> > > AFAIK, there's no primitives in VM, which provide timing with > nanosecond precision. > > >> Cheers, >> Alexandre >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > > > -- > Best regards, > Igor Stasenko AKA sig. > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
This is the level of state-of-the-Art Java profilers operate with.
Cheers, Alexandre On 8 Jan 2010, at 14:48, [hidden email] wrote: > I don't want to rain in your folks party, but a nanosecond is a > millionth of a millisecond. Also, it is at the timing of silicon > memory access on present microcomputers. > > Isn't enough you increase the accuracy of the timing by one thousand > looking for a solution to get *microsecond* resolution? > > HTH, > > -- > Cesar Rabak > > > > Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > > escreveu: > > > Ok, good to know. > > Thanks Igor, > Alexandre > > > On 7 Jan 2010, at 19:29, Igor Stasenko wrote: > >> 2010/1/7 Alexandre Bergel : >>> Hi! >>> >>> I am currently working on a new profiler for Pharo. The idea is to >>> graphically renders a profiling using Mondrian. I hope to release it >>> soon. >>> When discussing with an expert in the field (Walter Binder), the >>> question about time grain was raised. I am partly based on >>> MessageTally, which has a millisecond grain. Anyone is aware of a >>> finer grain? At the level of nanosecond? Does it makes sense since >>> there is no JIT (yet)? >>> >> >> AFAIK, there's no primitives in VM, which provide timing with >> nanosecond precision. >> >> >>> Cheers, >>> Alexandre >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> >> >> -- >> Best regards, >> Igor Stasenko AKA sig. >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
|
|
Do you have information if the "level" is resolution only or effective accuracy?
I got curious on this! -- Cesar Rabak Em 08/01/2010 15:56, Alexandre Bergel < [hidden email] > escreveu: This is the level of state-of-the-Art Java profilers operate with. Cheers, Alexandre On 8 Jan 2010, at 14:48, [hidden email] wrote: > I don't want to rain in your folks party, but a nanosecond is a > millionth of a millisecond. Also, it is at the timing of silicon > memory access on present microcomputers. > > Isn't enough you increase the accuracy of the timing by one thousand > looking for a solution to get *microsecond* resolution? > > HTH, > > -- > Cesar Rabak > > > > Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > > escreveu: > > > Ok, good to know. > > Thanks Igor, > Alexandre > > > On 7 Jan 2010, at 19:29, Igor Stasenko wrote: > >> 2010/1/7 Alexandre Bergel : >>> Hi! >>> >>> I am currently working on a new profiler for Pharo. The idea is to >>> graphically renders a profiling using Mondrian. I hope to release it >>> soon. >>> When discussing with an expert in the field (Walter Binder), the >>> question about time grain was raised. I am partly based on >>> MessageTally, which has a millisecond grain. Anyone is aware of a >>> finer grain? At the level of nanosecond? Does it makes sense since >>> there is no JIT (yet)? >>> >> >> AFAIK, there's no primitives in VM, which provide timing with >> nanosecond precision. >> >> >>> Cheers, >>> Alexandre >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> >> >> -- >> Best regards, >> Igor Stasenko AKA sig. >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
> Do you have information if the "level" is resolution only or
> effective accuracy? No idea. You can contact the guy who mentioned this to me: http://www.inf.usi.ch/faculty/binder/ Cheers, Alexandre > > -- > Cesar Rabak > > > Em 08/01/2010 15:56, Alexandre Bergel < [hidden email] > > escreveu: > > > This is the level of state-of-the-Art Java profilers operate with. > > Cheers, > Alexandre > > On 8 Jan 2010, at 14:48, [hidden email] wrote: > >> I don't want to rain in your folks party, but a nanosecond is a >> millionth of a millisecond. Also, it is at the timing of silicon >> memory access on present microcomputers. >> >> Isn't enough you increase the accuracy of the timing by one thousand >> looking for a solution to get *microsecond* resolution? >> >> HTH, >> >> -- >> Cesar Rabak >> >> >> >> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >> escreveu: >> >> >> Ok, good to know. >> >> Thanks Igor, >> Alexandre >> >> >> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >> >>> 2010/1/7 Alexandre Bergel : >>>> Hi! >>>> >>>> I am currently working on a new profiler for Pharo. The idea is to >>>> graphically renders a profiling using Mondrian. I hope to release >>>> it >>>> soon. >>>> When discussing with an expert in the field (Walter Binder), the >>>> question about time grain was raised. I am partly based on >>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>> finer grain? At the level of nanosecond? Does it makes sense since >>>> there is no JIT (yet)? >>>> >>> >>> AFAIK, there's no primitives in VM, which provide timing with >>> nanosecond precision. >>> >>> >>>> Cheers, >>>> Alexandre >>>> -- >>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>> Alexandre Bergel http://www.bergel.eu >>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>> >>>> >>>> >>>> >>>> >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>> >>> >>> >>> -- >>> Best regards, >>> Igor Stasenko AKA sig. >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by johnmci
this is a reply to John's point and to Cesar's ,essage
On Thu, Jan 7, 2010 at 8:22 PM, John McIntosh <[hidden email]> wrote: I think Eliot wanted those in my vm work I think qwaq has them Yes. We haven't moved to using it at the image level yet but the VM uses 64-bit microseconds from 1901 internally. The current value of the clock however is updated by the VM's heartbeat which runs anywhere from 500Hz to 1KHz. So while the clock has microsecond precision it only has millisecond resolution.
I made a similar change to VisualWorks, which uses 64-bit microseconds from 1901. The advantage is that sub-second time values are actually wall clock values, as opposed to a 32-bit millisecond value that will wrap after 49 days that is unsynchronised with wall time.
BTW, 64 bit microseconds from 1901 is ~580,000 years. So with e.g. 61-bit tagged integers in a 64-bit VM with maximum positive value of 2^60 - 1, the clock would take ~35,000 years to overflow tagged integers.
An unsigned long integer.
_______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Alexandre Bergel
really?
When I read what elliot said also with the precision I have some doubts. > This is the level of state-of-the-Art Java profilers operate with. > > Cheers, > Alexandre > > On 8 Jan 2010, at 14:48, [hidden email] wrote: > >> I don't want to rain in your folks party, but a nanosecond is a >> millionth of a millisecond. Also, it is at the timing of silicon >> memory access on present microcomputers. >> >> Isn't enough you increase the accuracy of the timing by one thousand >> looking for a solution to get *microsecond* resolution? >> >> HTH, >> >> -- >> Cesar Rabak >> >> >> >> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >> escreveu: >> >> >> Ok, good to know. >> >> Thanks Igor, >> Alexandre >> >> >> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >> >>> 2010/1/7 Alexandre Bergel : >>>> Hi! >>>> >>>> I am currently working on a new profiler for Pharo. The idea is to >>>> graphically renders a profiling using Mondrian. I hope to release it >>>> soon. >>>> When discussing with an expert in the field (Walter Binder), the >>>> question about time grain was raised. I am partly based on >>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>> finer grain? At the level of nanosecond? Does it makes sense since >>>> there is no JIT (yet)? >>>> >>> >>> AFAIK, there's no primitives in VM, which provide timing with >>> nanosecond precision. >>> >>> >>>> Cheers, >>>> Alexandre >>>> -- >>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>> Alexandre Bergel http://www.bergel.eu >>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>> >>>> >>>> >>>> >>>> >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>> >>> >>> >>> -- >>> Best regards, >>> Igor Stasenko AKA sig. >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Alexandre Bergel
ask him. I do not see how people can ask him like that while you discussed with him
On Jan 8, 2010, at 7:38 PM, Alexandre Bergel wrote: >> Do you have information if the "level" is resolution only or >> effective accuracy? > > No idea. You can contact the guy who mentioned this to me: http://www.inf.usi.ch/faculty/binder/ > > Cheers, > Alexandre > >> >> -- >> Cesar Rabak >> >> >> Em 08/01/2010 15:56, Alexandre Bergel < [hidden email] > >> escreveu: >> >> >> This is the level of state-of-the-Art Java profilers operate with. >> >> Cheers, >> Alexandre >> >> On 8 Jan 2010, at 14:48, [hidden email] wrote: >> >>> I don't want to rain in your folks party, but a nanosecond is a >>> millionth of a millisecond. Also, it is at the timing of silicon >>> memory access on present microcomputers. >>> >>> Isn't enough you increase the accuracy of the timing by one thousand >>> looking for a solution to get *microsecond* resolution? >>> >>> HTH, >>> >>> -- >>> Cesar Rabak >>> >>> >>> >>> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >>> escreveu: >>> >>> >>> Ok, good to know. >>> >>> Thanks Igor, >>> Alexandre >>> >>> >>> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >>> >>>> 2010/1/7 Alexandre Bergel : >>>>> Hi! >>>>> >>>>> I am currently working on a new profiler for Pharo. The idea is to >>>>> graphically renders a profiling using Mondrian. I hope to release >>>>> it >>>>> soon. >>>>> When discussing with an expert in the field (Walter Binder), the >>>>> question about time grain was raised. I am partly based on >>>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>>> finer grain? At the level of nanosecond? Does it makes sense since >>>>> there is no JIT (yet)? >>>>> >>>> >>>> AFAIK, there's no primitives in VM, which provide timing with >>>> nanosecond precision. >>>> >>>> >>>>> Cheers, >>>>> Alexandre >>>>> -- >>>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>>> Alexandre Bergel http://www.bergel.eu >>>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> Pharo-project mailing list >>>>> [hidden email] >>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>>> >>>> >>>> >>>> >>>> -- >>>> Best regards, >>>> Igor Stasenko AKA sig. >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > -- > _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: > Alexandre Bergel http://www.bergel.eu > ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. > > > > > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Stéphane Ducasse
Intel CPUs have the ability to read the clock count through the RDTSC
instruction, with OS access through QueryPerformanceCounter() on windows, clock_gettime() under linux and mach_absolute_time() on MacOS. Using this, I think it should be practical to have nanosecond precision for profiling under some conditions. See also http://en.wikipedia.org/wiki/Time_Stamp_Counter Cheers, Martin On 09/01/2010, at 10:45 PM, Stéphane Ducasse wrote: > really? > When I read what elliot said also with the precision I have some > doubts. > >> This is the level of state-of-the-Art Java profilers operate with. >> >> Cheers, >> Alexandre >> >> On 8 Jan 2010, at 14:48, [hidden email] wrote: >> >>> I don't want to rain in your folks party, but a nanosecond is a >>> millionth of a millisecond. Also, it is at the timing of silicon >>> memory access on present microcomputers. >>> >>> Isn't enough you increase the accuracy of the timing by one thousand >>> looking for a solution to get *microsecond* resolution? >>> >>> HTH, >>> >>> -- >>> Cesar Rabak >>> >>> >>> >>> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >>> escreveu: >>> >>> >>> Ok, good to know. >>> >>> Thanks Igor, >>> Alexandre >>> >>> >>> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >>> >>>> 2010/1/7 Alexandre Bergel : >>>>> Hi! >>>>> >>>>> I am currently working on a new profiler for Pharo. The idea is to >>>>> graphically renders a profiling using Mondrian. I hope to >>>>> release it >>>>> soon. >>>>> When discussing with an expert in the field (Walter Binder), the >>>>> question about time grain was raised. I am partly based on >>>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>>> finer grain? At the level of nanosecond? Does it makes sense since >>>>> there is no JIT (yet)? >>>>> >>>> >>>> AFAIK, there's no primitives in VM, which provide timing with >>>> nanosecond precision. >>>> >>>> >>>>> Cheers, >>>>> Alexandre >>>>> -- >>>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>>> Alexandre Bergel http://www.bergel.eu >>>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> Pharo-project mailing list >>>>> [hidden email] >>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- >>>>> project >>>>> >>>> >>>> >>>> >>>> -- >>>> Best regards, >>>> Igor Stasenko AKA sig. >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Stéphane Ducasse
I remember to have seen once at a keynote that some people use
statistic to reach that level of granularity. Something like, if executing n instructions took 1 second, and in 1 second you have so many CPU cycle, then you can deduce the average number of CPU cycle to execute a fraction of the whole computation. Cheers, Alexandre On 9 Jan 2010, at 09:15, Stéphane Ducasse wrote: > really? > When I read what elliot said also with the precision I have some > doubts. > >> This is the level of state-of-the-Art Java profilers operate with. >> >> Cheers, >> Alexandre >> >> On 8 Jan 2010, at 14:48, [hidden email] wrote: >> >>> I don't want to rain in your folks party, but a nanosecond is a >>> millionth of a millisecond. Also, it is at the timing of silicon >>> memory access on present microcomputers. >>> >>> Isn't enough you increase the accuracy of the timing by one thousand >>> looking for a solution to get *microsecond* resolution? >>> >>> HTH, >>> >>> -- >>> Cesar Rabak >>> >>> >>> >>> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >>> escreveu: >>> >>> >>> Ok, good to know. >>> >>> Thanks Igor, >>> Alexandre >>> >>> >>> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >>> >>>> 2010/1/7 Alexandre Bergel : >>>>> Hi! >>>>> >>>>> I am currently working on a new profiler for Pharo. The idea is to >>>>> graphically renders a profiling using Mondrian. I hope to >>>>> release it >>>>> soon. >>>>> When discussing with an expert in the field (Walter Binder), the >>>>> question about time grain was raised. I am partly based on >>>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>>> finer grain? At the level of nanosecond? Does it makes sense since >>>>> there is no JIT (yet)? >>>>> >>>> >>>> AFAIK, there's no primitives in VM, which provide timing with >>>> nanosecond precision. >>>> >>>> >>>>> Cheers, >>>>> Alexandre >>>>> -- >>>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>>> Alexandre Bergel http://www.bergel.eu >>>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> Pharo-project mailing list >>>>> [hidden email] >>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- >>>>> project >>>>> >>>> >>>> >>>> >>>> -- >>>> Best regards, >>>> Igor Stasenko AKA sig. >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Re: Nanosecond-level profiling in Pharo?
|
|
In reply to this post by Stéphane Ducasse
This is troubling indeed.
I will ask. Alexandre On 9 Jan 2010, at 09:16, Stéphane Ducasse wrote: > ask him. I do not see how people can ask him like that while you > discussed with him > > On Jan 8, 2010, at 7:38 PM, Alexandre Bergel wrote: > >>> Do you have information if the "level" is resolution only or >>> effective accuracy? >> >> No idea. You can contact the guy who mentioned this to me: http://www.inf.usi.ch/faculty/binder/ >> >> Cheers, >> Alexandre >> >>> >>> -- >>> Cesar Rabak >>> >>> >>> Em 08/01/2010 15:56, Alexandre Bergel < [hidden email] > >>> escreveu: >>> >>> >>> This is the level of state-of-the-Art Java profilers operate with. >>> >>> Cheers, >>> Alexandre >>> >>> On 8 Jan 2010, at 14:48, [hidden email] wrote: >>> >>>> I don't want to rain in your folks party, but a nanosecond is a >>>> millionth of a millisecond. Also, it is at the timing of silicon >>>> memory access on present microcomputers. >>>> >>>> Isn't enough you increase the accuracy of the timing by one >>>> thousand >>>> looking for a solution to get *microsecond* resolution? >>>> >>>> HTH, >>>> >>>> -- >>>> Cesar Rabak >>>> >>>> >>>> >>>> Em 07/01/2010 20:57, Alexandre Bergel < [hidden email] > >>>> escreveu: >>>> >>>> >>>> Ok, good to know. >>>> >>>> Thanks Igor, >>>> Alexandre >>>> >>>> >>>> On 7 Jan 2010, at 19:29, Igor Stasenko wrote: >>>> >>>>> 2010/1/7 Alexandre Bergel : >>>>>> Hi! >>>>>> >>>>>> I am currently working on a new profiler for Pharo. The idea is >>>>>> to >>>>>> graphically renders a profiling using Mondrian. I hope to release >>>>>> it >>>>>> soon. >>>>>> When discussing with an expert in the field (Walter Binder), the >>>>>> question about time grain was raised. I am partly based on >>>>>> MessageTally, which has a millisecond grain. Anyone is aware of a >>>>>> finer grain? At the level of nanosecond? Does it makes sense >>>>>> since >>>>>> there is no JIT (yet)? >>>>>> >>>>> >>>>> AFAIK, there's no primitives in VM, which provide timing with >>>>> nanosecond precision. >>>>> >>>>> >>>>>> Cheers, >>>>>> Alexandre >>>>>> -- >>>>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>>>> Alexandre Bergel http://www.bergel.eu >>>>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> Pharo-project mailing list >>>>>> [hidden email] >>>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> Best regards, >>>>> Igor Stasenko AKA sig. >>>>> >>>>> _______________________________________________ >>>>> Pharo-project mailing list >>>>> [hidden email] >>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- >>>>> project >>>> >>>> -- >>>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>>> Alexandre Bergel http://www.bergel.eu >>>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>>> >>>> >>>> >>>> >>>> >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> [hidden email] >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>> >>> -- >>> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >>> Alexandre Bergel http://www.bergel.eu >>> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [hidden email] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> -- >> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: >> Alexandre Bergel http://www.bergel.eu >> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. >> >> >> >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> [hidden email] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [hidden email] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
«
Return to Pharo Smalltalk Developers
|
1 view|%1 views
| Free forum by Nabble | Edit this page |