Does the red line in the gif mean that 10% of the time is spent in the housekeeping of OrderedCollection>>do: or does this include the time spent in evaluating aBlock value:?
It includes execution of the entire method, including time spent evaluating aBlock value:.
I guess the first is true as the end of the list states in the **Leaves** section that an overall 13.9% are spent in do:.
I've never been quite sure what the **Leaves** section means.
Put more general: I didn't find a detailed description on the swiki on how to interpret the results of the time profile.
Can someone please provide a pointer?
My understanding is that, during profiling, a snapshot is taken of the stack every few milliseconds. Whatever time has passed since the last snapshot is added to each frame in the current snapshot. That's why you see 100% in the root-frame because it was present in every snapshot.
- Chris
Hello Chris,
CM> It includes execution of the entire method, including time spent evaluating CM> aBlock value:.
after writing the message I did some testing. I made a method used in aBlock value: (the one below) count up some numbers to make it run significantly longer. This made the percentage of it grow and the the percentage of OrderedCollection>>do: decrease. That would indicate otherwise unless the time counting is more complicated.
CM> I've never been quite sure what the **Leaves** section means.
In my other programming environment (assembler on DSP) a leaf is a routine at the lowest level of nesting but I'm not sure if this is valid here.
CM> My understanding is that, during profiling, a snapshot is taken of the stack CM> every few milliseconds. Whatever time has passed since the last snapshot is CM> added to each frame in the current snapshot. That's why you see 100% in the CM> root-frame because it was present in every snapshot.
Well I imagined that samples are taken only of the top of stack for the time percentage and the deeper entries of the stack are only used to build the indentation.
Trying to find out from the source is above me, especially as opening a debugger seems to stop profiling.
So I'm still stuck about what really is sampled.
Cheers
Herbert mailto:herbertkoenig@gmx.net
squeak-dev@lists.squeakfoundation.org