Inline {min,max}_value even in debug builds

[tesuji]

I think it is worth to inline {min,max}_value even in debug builds.
See this godbolt link: https://godbolt.org/z/-COkVS

[Centril]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion

[bors]

⌛ Trying commit 3b49ab6 with merge 61430923b8b2f3e7565c7aa30a3eb96474641bd6...

[bors]

☀️ Try build successful - checks-azure
Build commit: 61430923b8b2f3e7565c7aa30a3eb96474641bd6 (61430923b8b2f3e7565c7aa30a3eb96474641bd6)

[rust-timer]

Queued 61430923b8b2f3e7565c7aa30a3eb96474641bd6 with parent 22bc9e1, future comparison URL.

[jonas-schievink]

Isn't perf only measuring compiler performance (which is always built in release mode)?

[tesuji]

So how can I measure debug build of compiler?

I saw this code generated from debug build (godbolt link: https://godbolt.org/z/9cpH6U)

core::num::<impl u32>::max_value:
        mov     eax, 4294967295
        ret

example::bar:
        push    rax
        call    core::num::<impl u32>::max_value
        mov     dword ptr [rsp + 4], eax
        mov     eax, dword ptr [rsp + 4]
        pop     rcx
        ret

That code is generated from:

pub fn bar() -> u32 {
    u32::max_value()
}

So I think max_value should always be inlined?

[tesuji]

r? @nnethercote (randomly assign)

[nnethercote]

@bors r+

[bors]

📌 Commit 3b49ab6 has been approved by nnethercote

[bors]

⌛ Testing commit 3b49ab6 with merge aa84c93d98bab674e7b7953cc30dd70f6d4cd0c1...

[Centril]

@bors retry rolled up.

[bors]

☔ The latest upstream changes (presumably #65045) made this pull request unmergeable. Please resolve the merge conflicts.

[scottmcm]

The godbolt example says to me to just stop using max_value(), not to add an attribute.

See also rust-lang/rfcs#2700

[tesuji]

@scottmcm What do you think about that case in #64996 and #65016 ?