Poor performance shmop read

[remco-pc]

Description

i've tried to use the shmop functions for a large file (json) (4GB) and have saved it in chunks of around 4 MB each.
With this approach i tried to read it and got poor performance; around 200 Mbit/sec only where i expected at least around 10x faster. The chunks cannot be really big currently and i wonder why that is, because if you could do chunks of hundreds of megabyte it would be faster as my current ramdrive which reads at 2.5 GB/sec with a file read (from tmpfs ramdrive)

This was in wsl2 on windows11

PHP Version

Php 8.3.22 (cli)

Operating System

Debian 12

[nielsdos]

The chunks cannot be really big currently and i wonder why that is, because if you could do chunks of hundreds of megabyte it would be faster as my current ramdrive which reads at 2.5 GB/sec with a file read (from tmpfs ramdrive)

I don't understand this part.
Doing file I/O should be slower than using shmop_read. Of course, using small chunks is bad for performance due to call & allocation overhead. Can you share some benchmark code? The only obvious improvement I see that can be done is using fast-ZPP and adding some optimization hints. Note that I only looked at this on my phone though so I didn't test this.