Dear All,
I have an application with multiple threads and multiple pointers to data.
The data needs to be accessed by all threads.
I need to access the pointers/data in the following way:
1) Allocating and writing
2) Changing data
3) Reading data
4) Reading data (non important)
All these actions are today protected by critical-sections. But when the
amount of data increases the synchronisation make the application slower.
Since 75% of the reading of the data is "non important" I wonder if I can
skip the critical-section when reading this data or can this have some
unexpected results? (I do these "non important" reads to provide statistics,
but these statistics does not need to be 100% accurate)
B
