Comments on: dotTrace Profiler 2.0 further improves performance profiling http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/ ReSharper for productivity, dotTrace for performance, dotCover for test coverage Mon, 13 Feb 2012 19:46:09 +0000 http://wordpress.org/?v=2.6.5 By: Ran Davidovitz http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-246 Ran Davidovitz Thu, 14 Sep 2006 05:55:42 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-246 Hi, I have worked with Ants profiler and after evaluating your product I have requested a refund because of few reasons: 1. It has many problems profiling 2.0 ASP.NET application while there are 2 frameworks working (dev team told me to request a refund)... 2. The performance penalty that i paid with ants is enormous. * One thing missing is the ability to see the performance of each line (not only functions)! Thanks - Version two is very nice... Hi,

I have worked with Ants profiler and after evaluating your product I have requested a refund because of few reasons:
1. It has many problems profiling 2.0 ASP.NET application while there are 2 frameworks working (dev team told me to request a refund)…
2. The performance penalty that i paid with ants is enormous.

* One thing missing is the ability to see the performance of each line (not only functions)!

Thanks - Version two is very nice…

]]> By: Alexandra Rusina http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-31 Alexandra Rusina Fri, 23 Jun 2006 08:55:47 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-31 Hi Ward, Have you seen the dotTrace demo? It explains all the interface issues in just 4 minutes. http://www.jetbrains.com/profiler/dotTrace_demo.html Hi Ward,
Have you seen the dotTrace demo? It explains all the interface issues in just 4 minutes.
http://www.jetbrains.com/profiler/dotTrace_demo.html

]]> By: Ward Bekker http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-29 Ward Bekker Thu, 22 Jun 2006 14:02:02 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-29 Hi Oleg, Thank you for you clarification. I'm not certain what the exact conditions where when I noticed the extra overhead, but I recall it was in a situation with multi-threading with a huge amount of small function calls. Good to hear that in most situations dotTrace has much lower overhead! Next time I need to do profiling, I'll try out the EAP. Ward Hi Oleg,

Thank you for you clarification. I’m not certain what the exact conditions where when I noticed the extra overhead, but I recall it was in a situation with multi-threading with a huge amount of small function calls. Good to hear that in most situations dotTrace has much lower overhead! Next time I need to do profiling, I’ll try out the EAP.

Ward

]]> By: oleg.stepanov http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-28 oleg.stepanov Thu, 22 Jun 2006 13:02:46 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-28 It's interesting that you notice more overhead from dotTrace than from Ants profiler. Our measurements show that dotTrace implies ten times as little overhead than Ants does. Which applications do you profile? Regarding the interface: finding out the slowest functions is one click away: just open the Hotspots view. It displays 50 slowest functions and by which call stacks they were called. To investigate a particular function just open it in a separate tab (by pressing Control+T). In this tab you can use Call Tree view to see which functions the current function calls and Backtraces view to see who calls the current function. Hope this helps. If you have other questions you can ask them in the newsgroup or contact me directly at oleg.stepanov (at) jetbrains.com. It’s interesting that you notice more overhead from dotTrace than from Ants profiler. Our measurements show that dotTrace implies ten times as little overhead than Ants does. Which applications do you profile? Regarding the interface: finding out the slowest functions is one click away: just open the Hotspots view. It displays 50 slowest functions and by which call stacks they were called. To investigate a particular function just open it in a separate tab (by pressing Control+T). In this tab you can use Call Tree view to see which functions the current function calls and Backtraces view to see who calls the current function. Hope this helps. If you have other questions you can ask them in the newsgroup or contact me directly at oleg.stepanov (at) jetbrains.com.

]]> By: Ward Bekker http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-27 Ward Bekker Thu, 22 Jun 2006 12:40:55 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-27 I tried out the trail for the 1.1 version of DotTrace, but compared to Redgate's Ants Profiler that I use on a day-to-day bases, it adds much more overhead than Ants Profiler. So, when I'm profiling a slow method, it will be cost like 1.5 times more time in Ants and 2.5 times more in DotTrace. Also, I found the interface of the Ants Profiler much more clear. But maybe that's because I'm not very familliar with DotTrace. Anyway, I had a hard time locating for example the top 10 of slowest lines of code, slowest methods with children etc. Also, in Ants I can see the callers from and to a method, I could not find it how to do this in DotTrace. Ok, my 2 cents...maybe they can be of use... I do like the intergration with Resharper 2.0 btw. I tried out the trail for the 1.1 version of DotTrace, but compared to Redgate’s Ants Profiler that I use on a day-to-day bases, it adds much more overhead than Ants Profiler. So, when I’m profiling a slow method, it will be cost like 1.5 times more time in Ants and 2.5 times more in DotTrace.

Also, I found the interface of the Ants Profiler much more clear. But maybe that’s because I’m not very familliar with DotTrace. Anyway, I had a hard time locating for example the top 10 of slowest lines of code, slowest methods with children etc. Also, in Ants I can see the callers from and to a method, I could not find it how to do this in DotTrace.

Ok, my 2 cents…maybe they can be of use…

I do like the intergration with Resharper 2.0 btw.

]]> By: oleg.stepanov http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-24 oleg.stepanov Wed, 21 Jun 2006 07:00:49 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-24 Brien, That's right, dotTrace currently measures "wall time" - time which was actually spent during function call even if the thread was blocked part of the time. Often it is useful: you want to know that a substantial amount of time was spent waiting for a resource. But sometimes knowing how much time was spend doing real work is also neccessary. Unfortunately just excluding known blocking functions doesn't solve the problem because some blockers are not represented by a managed call. Brien,
That’s right, dotTrace currently measures “wall time” - time which was actually spent during function call even if the thread was blocked part of the time. Often it is useful: you want to know that a substantial amount of time was spent waiting for a resource. But sometimes knowing how much time was spend doing real work is also neccessary. Unfortunately just excluding known blocking functions doesn’t solve the problem because some blockers are not represented by a managed call.

]]> By: Brien http://blogs.jetbrains.com/dotnet/2006/06/dottrace-profilers-user-friendly-performance-profiling-set-to-improve-further-in-v-20/#comment-23 Brien Tue, 20 Jun 2006 17:24:59 +0000 http://blogs.jetbrains.com/dotnet/2006/06/15/#comment-23 The biggest thing missing from 1.1 is the ability to exclude time spent inside calls that block (e.g. Monitor.Wait(), file and network i/o that block) . This makes it difficult to use for high performance applications that do a lot of i/o or have a lot of threads. I requested this feature in the past and I never really got a good response. The biggest thing missing from 1.1 is the ability to exclude time spent inside calls that block (e.g. Monitor.Wait(), file and network i/o that block) . This makes it difficult to use for high performance applications that do a lot of i/o or have a lot of threads.

I requested this feature in the past and I never really got a good response.

]]>