On 2026-03-18 09:57, Zack Weinberg wrote:
I'm currently looking into the possibility that the 'copy_file_range' system call is inefficient for small files
An interesting hypothesis. But on my platform (Fedora 43), for small files the 'cat' bottleneck is the mmap/munmap/mprotect/etc. syscalls for startup. copy_file_range takes only 10% of the CPU time.
Do you happen to remember why we concatenate confdefs.h and the test program, instead of #include-ing confdefs.h from the test program?
Sorry, no.
