bench: add tokio spawn_blocking + std fs benchmark #375
Conversation
Signed-off-by: MrCroxx <[email protected]>
Signed-off-by: MrCroxx <[email protected]>
| let mut buffer = [0u8; BUFFER_SIZE]; | ||
| for &offset in *offsets { | ||
| let file = file.clone(); | ||
| buffer = tokio::task::spawn_blocking(move || { |
There was a problem hiding this comment.
In this function, every read operation is performed in a new spawn_blocking closure. It should be a little faster than read_tokio because it uses seek_read or read_at instead of seek + read.
| for _i in 0..iter { | ||
| let mut file = std::fs::File::open(path).unwrap(); | ||
| let len = *len; | ||
| buffer = tokio::task::spawn_blocking(move || { |
There was a problem hiding this comment.
In this function, all read operations are performed inside the same spawn_blocking closure. Why it's not as fast as read_all_std? Because you open the file inside the for loop. The open-file operation costs a lot.
| let content = content.clone(); | ||
|
|
||
| tokio::task::spawn_blocking(move || { | ||
| for offset in offsets.iter() { |
There was a problem hiding this comment.
In this function, all write operations are performed inside the same spawn_blocking closure. It should be as fast as write_std.
| file.seek(SeekFrom::Start(0)).unwrap(); | ||
| let mut write_len = 0; | ||
| let total_len = content.len(); | ||
| while write_len < total_len { |
There was a problem hiding this comment.
In this function, all write operations are performed inside the same spawn_blocking closure. It should be as fast as write_all_std.
Hi. Thank you for providing the crate
compio. 🥰This PR introduces tokio
spawn_blocking+stdcase for fs benchmark.The benchmark shows that compio works really good on reads. However, the write part is slower than tokio spawn_blocking + std. I'm not sure if I've done it correctly. Could you please help? 🙏
Raw output: