Martijn van Beurden

I can test on Windows. Something that might be relevant: what filesystem did you test on? You mentioned ntfs once, are all tests on ntfs?

So, could you summarize what you're seeing? What hardware were you using for the numbers in the post where you said to disregard all earlier measurements? Was that an HDD?...

I've tried to replicate. Here are results for running on Ubuntu 22.04, Intel Celeron N5105, Seagate Barracuda 2.5 5400. ext4 partition. ``` $ for I in {1..20}; do for EXE...

How much space is left on that device when running the tests? If that is more than 20%, fragmentation should not be too much of an issue. On the other...

I've finally taken some time to test this under Windows, and I must say, under the right conditions the difference is horrific, truly. I've compared three compiles, one with the...

Here are some more detailed results [set of 16bit tracks vbuf.pdf](https://github.com/xiph/flac/files/11499339/set.of.16bit.tracks.vbuf.pdf) I'd say using a 32KiB buffer already gives a nice performance improvement. I'm not sure whether it makes much...

Some more results. Took a 32GB NTFS partition on a 250GB SSD (several years old) and filled it for 99%. Then ran the tests on Windows 10. [set of 16bit...

Even more results. I tried to test something a little more realistic, so took a 1TB HDD, filled it for 95% and ran tests. The results are opposite of the...

What kind of audio did you use to get these numbers? Perhaps somewhat rude, but I am currently not inclined to accept this PR. This change is not covered by...

~Please provide some more information on the tests, because I cannot explain the results you're getting. The problem I have with this, is that the function you are optimizing should...