On 03/30/2010 09:42 AM, Andrew Overholt wrote: > Hi, > > I spoke with Thavidu off-list and suggested that we create a common > profiling layer that can be used by multiple backend implementations > (ex. OProfile, perf). I'd rather see this than a collection of > largely-duplicated plugins. We could then just have a common data > format that we display in the UI. Thoughts? > > Andrew > _______________________________________________ > linuxtools-dev mailing list > linuxtools-dev@eclipse.org > https://dev.eclipse.org/mailman/listinfo/linuxtools-dev
Hi Andrew, Yes, avoiding a proliferation of profiling plugin would be good. It makes sense to factor out a lot of the common code for controlling the collection of data, have a table describing the operations for each. The common data format for the UI might be more painful. There might be a fair amount translation code required for each profilers format to convert to yet another data format. There are also types of data analysis that don't match up between the different performance tools. For perf there are some interesting scripts in python and perf that may not match up with the mapping samples pc values back to code to show code hotspots: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=tree;f=tools/perf/scripts -Will _______________________________________________ linuxtools-dev mailing list linuxtools-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/linuxtools-dev
