We did this to allow pipelining of GPU/CPU work in nanobench on devices
that don't have GL sync objects (or NV fence). However, we've found
that on the AndroidOne/Mali400 and Nexus7/Tegra3 devices that the
majority of a frame's work can be reordered to complete before an
earlier EGL sync. This makes the results unreliable, especially the
min_ms result. Just accept that we will sync the CPU and GPU on these
devices.
Change-Id: I04d168cbb05504d367f6a06b7b4903163ab2aa79
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/414897
Commit-Queue: Brian Salomon <bsalomon@google.com>
Commit-Queue: Adlai Holler <adlai@google.com>
Auto-Submit: Brian Salomon <bsalomon@google.com>
Reviewed-by: Adlai Holler <adlai@google.com>
7db0e2f942