I am trying to profile a single core application (matrix kindaa 2D operation..).
I tried sample profiling with 11 events. When I drill down into my EXE in the sampling results, the window shows 4 threads and data are scaterred between these rows.. Howver, my application is only single-threaded. To reduce confusion, I also nail my main thread to 1 core using the following call:
Since the data is scattered in 4 rows now, certain ratios are not coming out clearly. For example: DTLB_MISSES.ANY and INST_RETIRED.ANY are seen in different rows and hence DTLB Miss Rate is 0 in all the 4 rows... I had to divide the data from 2 different rows to make sure that I get the correct rate...
How do I get past this?
Also, what is the "%" of various events shown in the sampling results? They are all 100% always....May be, if I drill down further there would be variations.... Ha.. Possibly...
Thanks for your time,
This looks like a good idea... I was always wondering how togo-up after I drill-down..... Looks like, "Modules" is the key..... Right?
Is there any other way to drill-upward like a stack(for a lack of better word) after I drill-down..
Thanks! (I was expecting the forum to notify me of replies... I had subscribed to this thread.. Sadly that does not work..)