This article is part of the series An Exercise in Performance Tuning in C#.Net.
After a month-long hiatus — too much work, too fast and furious for posting — I’m back to the last posts in the series on real-world performance tuning. The point of these postings hasn’t been to glorify my ability to tune a C# data processing application, but rather to share what I’ve learned in attempts to do so.
Where to next? Turns out my next steps were false starts, at least insofar as tuning is concerned. Still, there were some lessons (or should be) from these dead ends.
The application modifies a few pieces of data from the input file and writes out
a nearly-identical output file. I am re-reading the input file, identifying the
line to alter, stripping out the old values and inserting the new ones (note:
the modified data is an the
obj object, which has preserved the original line
number for ease of going back to it at this stage):
string str = this._fileLines[obj.LineNumber]; str = str.Remove(this._parseDictionary[Enums.Fields.Field1].Position - 1, this._parseDictionary[Enums.Fields.Field1].Length); str = str.Insert(this._parseDictionary[Enums.Fields.Field1].Position - 1, obj.Field1.PadRight(this._parseDictionary[Enums.Fields.Field1].Length, ' '));
I thought this would end up copying data into 1 new string (
string str = ...)
and then two more new strings for every change. So I changed the logic to use
(unsafe) pointers instead, making str into a pointer into the array:
Much to my surprise this made no difference in the performance. I have to wonder
if I was misunderstanding string
str = this._fileLines[obj.LineNumber]. I know
that if you have :
string str1 = "something"; string str2 = str1;
then there will be two memory segments with the value “something”. I thought
this statement would apply to my assignment
this._fileLines[obj.LineNumber]. But given that there was no performance
difference, I now suspect that assignment from the
List<string> [index] is
treated implicitly as a pointer instead of a memory copy.
Posted with : Tech, Microsoft .NET Framework, Performance Tuning