Profiling ./build/Release/vtile-encode ./test/geometry-test-data/input/counter-clockwise-polygon.json 0 0 0 -i 100000000 with Activity Monitor shows significant (> 2 %) of time taken in std::future and std::async::deferred usage. This says to me that locks are being invoked even when we are running synchronously, which is bad and indicates we may want to re-write to not use futures when we are not using std::launch::async (assuming this also impacts linux):
Profiling
./build/Release/vtile-encode ./test/geometry-test-data/input/counter-clockwise-polygon.json 0 0 0 -i 100000000
withActivity Monitor
shows significant (> 2 %) of time taken instd::future
andstd::async::deferred
usage. This says to me that locks are being invoked even when we are running synchronously, which is bad and indicates we may want to re-write to not use futures when we are not usingstd::launch::async
(assuming this also impacts linux):