Tigran Aivazian <tigran@sco.COM> writes:
> This simple restructuring of sys_wait4() not only makes it more readable
> (because of less pressure on tab'd space) but also saves the precious 16
> bytes of kernel text:
>
> Before
> ======
> text data bss dec hex filename
> 1383321 257386 413440 2054147 1f5803 vmlinux
>
> After
> =====
>
> text data bss dec hex filename
> 1383305 257386 413440 2054131 1f57f3 vmlinux
The patch is more readable, but it looks like it's also a bit slower. If
state is TASK_STOPPED but exit_code is 0, we end up (uselessly) testing state
a second time, since it is volatile. I haven't tried, but I believe this
would be fixed by swapping the two tests (test for TASK_ZOMBIE first).
Also, when I tried this, it looks like the text size for exit.o is the same
before and after, so I'm not sure why the size of vmlinux varies.
--Mike
# md5sum exit.c exit.o exit.c-tigran exit.o-tigran
fe1e76a559e36f31fcecede026819e52 exit.c
89994fee7aa4d746d6441fecf235a1fc exit.o
85465171760971b38a759b3a3995a0b9 exit.c-tigran
1896dcf31e08492cb2a946acfa8f17de exit.o-tigran
# size exit.o exit.o-tigran
text data bss dec hex filename
4165 0 0 4165 1045 exit.o
4165 0 0 4165 1045 exit.o-tigran
-- patched version
.L2019:
movl (%ebx),%eax
movl $1,%ecx
cmpl $8,%eax # TASK_STOPPED
jne .L2022
cmpl $0,84(%ebx) # p->exit_code
je .L2022 # this jumps to a check against TASK_ZOMBIE
# since state is volatile
testb $2,76(%esp)
jne .L2023
testb $16,4(%ebx)
je .L2008
-- Any sufficiently adverse technology is indistinguishable from Microsoft.- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 15 2000 - 21:00:15 EST