Skip to content

Apply vectorcall for more places #7336

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
youknowone merged 6 commits into from
Mar 4, 2026
Merged

Apply vectorcall for more places #7336

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
f584c47
Add PyType vectorcall and use vectorcall in all specialized call fall…
youknowone Mar 3, 2026
de9cbc4
Add vectorcall slot for PyMethodDescriptor and PyWrapper
youknowone Mar 3, 2026
02a6998
Add FuncArgs::from_vectorcall_owned and simplify vectorcall fallback …
youknowone Mar 3, 2026
07f129b
Use lazy locals allocation in vectorcall fast path, add debug_assert …
youknowone Mar 3, 2026
c47d1f2
Use from_vectorcall_owned in descriptor/type vectorcall, add stack de…
youknowone Mar 3, 2026
2885f80
Collapse duplicated branch tails and use checked_sub in CALL_KW
youknowone Mar 3, 2026
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
Prev Previous commit
Collapse duplicated branch tails and use checked_sub in CALL_KW 
- Collapse conditional deopt + unconditional vectorcall pattern in
  CallBuiltinClass, CallNonPyGeneral, CallKwNonPy
- Use checked_sub for nargs_usize - kw_count in execute_call_kw_vectorcall
  to prevent silent underflow in release builds
  • Loading branch information
@youknowone
youknowone committed Mar 4, 2026
commit 2885f80785ac9f7ff3303393fdc54ee2996f33d4
31 changes: 15 additions & 16 deletions crates/vm/src/frame.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3931,12 +3931,11 @@ impl ExecutingFrame<'_> {
let nargs: u32 = arg.into();
let callable = self.nth_value(nargs + 1);
let callable_tag = callable as *const PyObject as u32;
if cached_tag == callable_tag && callable.downcast_ref::<PyType>().is_some() {
return self.execute_call_vectorcall(nargs, vm);
if !(cached_tag == callable_tag && callable.downcast_ref::<PyType>().is_some()) {
self.deoptimize(Instruction::Call {
argc: Arg::marker(),
});
}
self.deoptimize(Instruction::Call {
argc: Arg::marker(),
});
self.execute_call_vectorcall(nargs, vm)
}
Instruction::CallAllocAndEnterInit => {
Expand Down Expand Up @@ -4078,12 +4077,11 @@ impl ExecutingFrame<'_> {
let nargs: u32 = arg.into();
let callable = self.nth_value(nargs + 1);
let callable_tag = callable as *const PyObject as u32;
if cached_tag == callable_tag {
return self.execute_call_vectorcall(nargs, vm);
if cached_tag != callable_tag {
self.deoptimize(Instruction::Call {
argc: Arg::marker(),
});
}
self.deoptimize(Instruction::Call {
argc: Arg::marker(),
});
self.execute_call_vectorcall(nargs, vm)
}
Instruction::CallKwPy => {
Expand Down Expand Up @@ -4175,12 +4173,11 @@ impl ExecutingFrame<'_> {
let nargs: u32 = arg.into();
let callable = self.nth_value(nargs + 2);
let callable_tag = callable as *const PyObject as u32;
if cached_tag == callable_tag {
return self.execute_call_kw_vectorcall(nargs, vm);
if cached_tag != callable_tag {
self.deoptimize(Instruction::CallKw {
argc: Arg::marker(),
});
}
self.deoptimize(Instruction::CallKw {
argc: Arg::marker(),
});
self.execute_call_kw_vectorcall(nargs, vm)
}
Instruction::LoadSuperAttrAttr => {
Expand Down Expand Up @@ -5668,7 +5665,9 @@ impl ExecutingFrame<'_> {
.map(|sr| sr.to_pyobj());
let has_self = self_or_null.is_some();

let pos_count = nargs_usize - kw_count;
let pos_count = nargs_usize
.checked_sub(kw_count)
.expect("CALL_KW: kw_count exceeds nargs");
let effective_nargs = if has_self { pos_count + 1 } else { pos_count };

// Build the full args slice: positional (including self) + kwarg values
Expand Down