Replies: 17 comments 25 replies
-
Command::new("./configure")
.arg("--cc=clang")
.arg("--cxx=clang++")
.current_dir(&build)
.status()
.unwrap();
Command::new("make")
.arg("-C")
.arg(&build)
.arg("CFLAGS=-flto=thin -g -O3 -Wall -Wextra -fno-stack-protector")
.arg("CXXFLAGS=-flto=thin")
.arg("V=1")
.status()
.unwrap(); |
Beta Was this translation helpful? Give feedback.
-
[profile.release]
lto = "thin"
codegen-units = 1 |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
without ThinLTO |
Beta Was this translation helpful? Give feedback.
-
|
with ThinLTO |
Beta Was this translation helpful? Give feedback.
-
#[inline(never)]
#[no_mangle]
pub fn test_func_zzwxh(sqe:&mut io_uring_sqe) {
unsafe {
io_uring_prep_nop(sqe);
}
} |
Beta Was this translation helpful? Give feedback.
-
Those flags are needed to compile liburing nolibc.
Just do
You probably need to alter the |
Beta Was this translation helpful? Give feedback.
-
|
@ammarfaizi2 since you include liburing.h without defining IOURINGINLINE, it will be static inline, so it doesn't prove you can use LTO here. |
Beta Was this translation helpful? Give feedback.
-
|
@ammarfaizi2 try this and see if the LTO works. |
Beta Was this translation helpful? Give feedback.
-
|
Ah yeah, you're right. Now tested with my suggested test_lto code. It still can do LTO. 0000000000003120 <test_lto>:
3120: 48 b8 00 00 00 00 ff movabs $0xffffffff00000000,%rax
3127: ff ff ff
312a: 48 89 07 mov %rax,(%rdi)
312d: 0f 57 c0 xorps %xmm0,%xmm0
3130: 0f 11 47 28 movups %xmm0,0x28(%rdi)
3134: 48 c7 47 38 00 00 00 movq $0x0,0x38(%rdi)
313b: 00
313c: 0f 11 47 08 movups %xmm0,0x8(%rdi)
3140: 48 c7 47 18 00 00 00 movq $0x0,0x18(%rdi)
3147: 00
3148: c3 ret
3149: 0f 1f 80 00 00 00 00 nopl 0x0(%rax) |
Beta Was this translation helpful? Give feedback.
-
Oh yeah, it was a mistake. But still, LTO works. The fact |
Beta Was this translation helpful? Give feedback.
-
|
I test with rust, lto is blocked until use-libc is enabled |
Beta Was this translation helpful? Give feedback.
-
|
Although you believe that |
Beta Was this translation helpful? Give feedback.
-
|
https://discord.com/channels/636084430946959380/636732781086638081/1149426538811510924 |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
Hmm, you know, the more I think about this, I'm almost wondering if it's just outright better to transliterate liburing to Rust and maintain it as a separate fork. This, of course, brings about the possibility of some impl divergence and lag but seeing the contortions I had to go through and what zzwxh is going through, a native impl might be more appropriate. Fundamentally, liburing is just a userspace wrapper around io_uring. I don't see any inherent reason why we couldn't just translate this code. |
Beta Was this translation helpful? Give feedback.
-
|
All of io_uring requires synchronization with the kernel but that really doesn't need to bleed out into the Rust API. The Rust API, as far as I can tell, wouldn't need Send or Sync at all and in fact should be !Send and !Sync. |
Beta Was this translation helpful? Give feedback.
-
|
There is already https://github.com/tokio-rs/io-uring/ which is a low level binding to kernel uring, similar to what liburing is, but native and idiomatic Rust. It has no dependencies other than |
Beta Was this translation helpful? Give feedback.
-
let read_e = opcode::Read::new(types::Fd(fd.as_raw_fd()), buf.as_mut_ptr(), buf.len() as _)
.build()
.user_data(0x42);
// Note that the developer needs to ensure
// that the entry pushed into submission queue is valid (e.g. fd, buffer).
unsafe {
ring.submission()
.push(&read_e)
.expect("submission queue is full");
}look how they massacred my boy! |
Beta Was this translation helpful? Give feedback.
-
|
I tried rewriting liburing again, and encountered some issues, especially within sq_thread_idle where I repeatedly received IORING_SQ_NEED_WAKEUP. I believe this is due to incorrect synchronization, but I'm not sure where to start troubleshooting. |
Beta Was this translation helpful? Give feedback.
-
|
It is also possible that it is a kernel bug or I misunderstood the SQ_POLL mode. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @aaaaaaaaaaasjksj, |
Beta Was this translation helpful? Give feedback.
-
|
https://crates.io/crates/liburing2-sys |
Beta Was this translation helpful? Give feedback.
-
liburing/src/Makefile
Line 48 in 660cefa
llvm ThinLTO prevented until remove
-ffreestanding-fno-builtinWhat are they
-ffreestanding-fno-builtinCan i remove them safely
Can i enable LTO without changing Makefile s
Beta Was this translation helpful? Give feedback.
All reactions