Right now, we are rendering each column such that the middle of the column is aligned with the CPU measurement. This could potentially be misleading, as due to the width of the bar, it implies that there was a period of time after the actual CPU measurement that is attributed to that same CPU measurement. 1 2 3 _______ [ * ] [ * ]_______ [ * ][ * ] _______[ * ][ * ] [ * ][ * ][ * ] [__*__][__*__][__*__] A B C D E F In this example, a developer might "attribute" any work done at time B to record 1, when in reality it should be "attributed" to record 2, since the CPU measurement had already been taken by the time B was captured, meaning that the work for B hadn't yet been don and could therefore not have affected the CPU measurement We should be rendering the columns such that the CPU measurement aligns with the trailing edge of the column. 1 2 3 _______ ___ [ *] [ [ *]_______[ [ *][ *][ ____[ *][ *][ *][ *][ *][ __*][____*][____*][__ A B C D E F NOT: the drawing isn't exactly accurate, as the * would overlap the ]. Legend: [ ] = a column for a CPU measurement (500ms wide to account for the fact that CPU measurements are 500ms apart) * = the time when the measurement actually takes place
<rdar://problem/48915271>
Created attachment 364769 [details] Patch
Created attachment 364907 [details] Patch
Comment on attachment 364907 [details] Patch r=me
Comment on attachment 364907 [details] Patch Clearing flags on attachment: 364907 Committed r243166: <https://trac.webkit.org/changeset/243166>
All reviewed patches have been landed. Closing bug.