You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+2-1Lines changed: 2 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -120,7 +120,8 @@ The instructions below assume you're on Linux.
120
120
121
121
## Benchmark results
122
122
123
-
The following results were obtained by running the [benchmark_decoders.ipynb](./benchmarks/decoders/benchmark_decoders.ipynb) on a lightly-loaded 22-core machine. We first get the operation latency forvarious seek and decode patternsin a loop from a single python thread and then compute the FPS (= 1 / latency). Error bars show the FPS of the p25 and p75 operation latency.
123
+
The following results were obtained by running the [benchmark_decoders.ipynb](./benchmarks/decoders/benchmark_decoders.ipynb) on a lightly-loaded 22-core machine. We first get the operation latency forvarious seek and decode patternsin a loop from a single python thread
124
+
for a single video and then compute the FPS (= 1 / latency). Error bars show the FPS of the p25 and p75 operation latency.
124
125
125
126
If you are running multiple copies of the decoder (for example in a DataLoader) on multiple threads, it is best to use `num_threads=1`for the best performance. If you care about the operation latency (not throughput of many concurrent operations) you can set`num_threads=0` to utilize all cores of your computer.
0 commit comments