4 ms·
I'll eat my files if this gives 8x on normal applications reading normal sized files from real SSDs or HDDs. The commenter mentions "greater than 8x" on a very
by TheDong 3y ago
I'll eat my files if this gives 8x on normal applications reading normal sized files from real SSDs or HDDs.
The commenter mentions "greater than 8x" on a very artificial benchmark of reading /dev/zero. The original PR author than copies that to the PR description as "8x has been observed" (lossily dropping the greater, but not adding any additional datapoints).
The reality will be that disk IO dwarfs the overhead of syscalls and threadpools in all real cases.
That all that's being saved here: the overhead of having to manage a threadpool, and make a few more syscalls.
Compared to reading a few MB off disk, that overhead will not be noticed, and definitely won't be 8x.
- loeg 3y agoI would expect this to benefit IO operations on NVMe media the most, and slow media like HDD the least. The OS stack imposes a substantial part of the overall latency in tiny random operations against NVMe drives.