Aaron Boxer
Aaron Boxer
I have adjusted the benchmark to be a truer apples-to-apples comparison between synchronous writes and uring writes. We now have uring significantly faster than synchronous on 48 thread system. Most...
This benchmark is basically complete. It now supports synchronous and asynchronous writes, using both buffered and direct IO. Testing on writing a 2.8 GB file show that 1. synchronous and...
A few more interesting data points for uring writes (from fedora 36 / kernel 5.17 system) : 1. direct IO uses ~20% more wall-clock time than buffered 2. direct IO...
> > This benchmark is basically complete. It now supports synchronous and asynchronous writes, using both buffered and direct IO. Testing on writing a 2.8 GB file show that >...
FYI: rendering intent in this profile equals 0x10000, which is invalid.
Yes. libpng seems to have such a routine - you could use that as a guide for a sanity check on profile.
> Hey @boxerab , could you resubmit this against the `release/5.0` branch? done, but I get test errors: ``` cantaloupe_1 | [ERROR] GrokProcessorTest.testProcessWithNonzeroOrientation:14->AbstractProcessorTest.testProcessWithNonzeroOrientation:167->AbstractProcessorTest.doProcessTest:965 ? SourceFormat ``` I seem to recall...
+1 for this feature
better approach would be to serialize directly to disk
Hi Aous, Glad to hear all is well! Thanks a lot for explaining these different approaches to rate control and Q. I may do some experiments on the CUP termination...