The instinct is to read it left to right as a timeline, and the x axis is alphabetical — a wide block is total time across the whole sample rather than one long call.
width = share of samples. wide is expensive.
height = stack depth. a tall thin tower is deep and cheap.
x axis = ALPHABETICAL. not chronological.
what to look at: the widest blocks NEAREST THE TOP.
those are doing work rather than delegating it.A wide block low in the stack is a framework entry point and tells you nothing; the same width near the top is a function actually consuming CPU. Producing one before and after a change and comparing shapes is more informative than comparing two numbers, because it shows where the time moved to rather than only that it moved. Reading one correctly takes about five minutes to learn and saves considerably more than that.