Skip to content

gh-119726: emit AArch64 trampolines in the data section #121280

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
brandtbucher merged 2 commits into from
Jul 3, 2024
+8 −8
Merged

gh-119726: emit AArch64 trampolines in the data section #121280

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ea726d2
gh-119726: emit AArch64 trampolines in the data section
diegorusso Jun 25, 2024
13e5a2f
Update Misc/NEWS.d/next/Core and Builtins/2024-06-28-18-34-49.gh-issu…
diegorusso Jul 3, 2024
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
2 changes: 2 additions & 0 deletions Misc/NEWS.d/next/Core and Builtins/2024-06-28-18-34-49.gh-issue-119726.Fjv_Ab.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
Optimize code layout for calls to C functions from the JIT on AArch64.
Patch by Diego Russo.
14 changes: 6 additions & 8 deletions Tools/jit/_stencils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,7 +184,7 @@ def pad(self, alignment: int) -> None:
self.disassembly.append(f"{offset:x}: {' '.join(['00'] * padding)}")
self.body.extend([0] * padding)

def emit_aarch64_trampoline(self, hole: Hole, alignment: int) -> None:
def emit_aarch64_trampoline(self, hole: Hole, alignment: int) -> Hole:
"""Even with the large code model, AArch64 Linux insists on 28-bit jumps."""
assert hole.symbol is not None
reuse_trampoline = hole.symbol in self.trampolines
Expand All @@ -194,14 +194,10 @@ def emit_aarch64_trampoline(self, hole: Hole, alignment: int) -> None:
else:
self.pad(alignment)
base = len(self.body)
where = slice(hole.offset, hole.offset + 4)
instruction = int.from_bytes(self.body[where], sys.byteorder)
instruction &= 0xFC000000
instruction |= ((base - hole.offset) >> 2) & 0x03FFFFFF
self.body[where] = instruction.to_bytes(4, sys.byteorder)
new_hole = hole.replace(addend=base, symbol=None, value=HoleValue.DATA)
Copy link
Member

@brandtbucher brandtbucher Jul 3, 2024

Choose a reason for hiding this comment

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

This is so much nicer!


if reuse_trampoline:
return
return new_hole

self.disassembly += [
f"{base + 4 * 0:x}: 58000048 ldr x8, 8",
Expand All @@ -219,6 +215,7 @@ def emit_aarch64_trampoline(self, hole: Hole, alignment: int) -> None:
self.body.extend(code)
self.holes.append(hole.replace(offset=base + 8, kind="R_AARCH64_ABS64"))
self.trampolines[hole.symbol] = base
return new_hole

def remove_jump(self, *, alignment: int = 1) -> None:
"""Remove a zero-length continuation jump, if it exists."""
Expand Down Expand Up @@ -294,8 +291,9 @@ def process_relocations(self, *, alignment: int = 1) -> None:
in {"R_AARCH64_CALL26", "R_AARCH64_JUMP26", "ARM64_RELOC_BRANCH26"}
and hole.value is HoleValue.ZERO
):
self.code.emit_aarch64_trampoline(hole, alignment)
new_hole = self.data.emit_aarch64_trampoline(hole, alignment)
self.code.holes.remove(hole)
self.code.holes.append(new_hole)
self.code.remove_jump(alignment=alignment)
self.code.pad(alignment)
self.data.pad(8)
Expand Down