Skip to content

Prefer asm! over llvm_asm! in core #76669

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Sep 16, 2020
Merged

Prefer asm! over llvm_asm! in core #76669

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
87666e5
Prefer asm! over llvm_asm! in core
tesuji Sep 16, 2020
4dc4e9f
Fix black_box bug detected by Amanieu
tesuji Sep 16, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions library/core/src/hint.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -111,7 +111,7 @@ pub fn spin_loop() {
#[inline]
#[unstable(feature = "test", issue = "50297")]
#[allow(unreachable_code)] // this makes #[cfg] a bit easier below.
pub fn black_box<T>(dummy: T) -> T {
pub fn black_box<T>(mut dummy: T) -> T {
// We need to "use" the argument in some way LLVM can't introspect, and on
// targets that support it we can typically leverage inline assembly to do
// this. LLVM's interpretation of inline assembly is that it's, well, a black
Expand All @@ -121,7 +121,8 @@ pub fn black_box<T>(dummy: T) -> T {
#[cfg(not(miri))] // This is just a hint, so it is fine to skip in Miri.
// SAFETY: the inline assembly is a no-op.
unsafe {
llvm_asm!("" : : "r"(&dummy));
// FIXME: Cannot use `asm!` because it doesn't support MIPS and other architectures.
llvm_asm!("" : : "r"(&mut dummy));
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Incidentally the old asm had bugs: you need volatile and memory clobbers to have the proper black_box effect. Currently rustc automatically adds volatile and LLVM adds memory when volatile is set, but we probably shouldn't rely on this.

}

dummy
Expand Down
20 changes: 17 additions & 3 deletions library/core/src/num/dec2flt/algorithm.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,12 +60,19 @@ mod fpu_precision {
fn set_cw(cw: u16) {
// SAFETY: the `fldcw` instruction has been audited to be able to work correctly with
// any `u16`
unsafe { llvm_asm!("fldcw $0" :: "m" (cw) :: "volatile") }
unsafe {
asm!(
"fldcw ({})",
in(reg) &cw,
// FIXME: We are using ATT syntax to support LLVM 8 and LLVM 9.
options(att_syntax, nostack),
)
}
}

/// Sets the precision field of the FPU to `T` and returns a `FPUControlWord`.
pub fn set_precision<T>() -> FPUControlWord {
let cw = 0u16;
let mut cw = 0_u16;

// Compute the value for the Precision Control field that is appropriate for `T`.
let cw_precision = match size_of::<T>() {
Expand All @@ -78,7 +85,14 @@ mod fpu_precision {
// `FPUControlWord` structure is dropped
// SAFETY: the `fnstcw` instruction has been audited to be able to work correctly with
// any `u16`
unsafe { llvm_asm!("fnstcw $0" : "=*m" (&cw) ::: "volatile") }
unsafe {
asm!(
"fnstcw ({})",
in(reg) &mut cw,
// FIXME: We are using ATT syntax to support LLVM 8 and LLVM 9.
options(att_syntax, nostack),
)
}

// Set the control word to the desired precision. This is achieved by masking away the old
// precision (bits 8 and 9, 0x300) and replacing it with the precision flag computed above.
Expand Down