
498 VI Debugging and Profiling
Rendering Frame N
DC_0
GPU
CPU
t
DC_1
DC_2DC_3
DC_0
DC_1
DC_2
DC_3
QB QE
Measured time
S
Figure 34.4. Correct timing of overlapped draw calls (DC n) using an asynchronous timer query and a GPU
synchronization point S to defer the execution of the enclosed commands without blocking CPU execution.
of the rendered objects and the surrounding rendering commands. Smaller, less com-
plex objects show smaller divergences, while larger objects offer more opportunities
for overlapped execution and therefore show larger divergences between measured
and actual rendering times. This issue is resolved without stalling CPU execution by
using an