Skip to content

gh-119726: replace AArch64 trampolines with LDR #121001

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 4 commits into from
Jul 1, 2024
Merged

gh-119726: replace AArch64 trampolines with LDR #121001

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9cb1976
gh-119726: replace AArch64 trampolines with LDR
diegorusso Jun 24, 2024
4cc470c
📜🤖 Added by blurb_it.
blurb-it[bot] Jun 25, 2024
cebe632
Fix black suggestion
diegorusso Jun 25, 2024
45729d7
Improve news explanation.
diegorusso Jun 26, 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/2024-06-25-16-26-44.gh-issue-119726.WqvHxB.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
Improve the speed and memory use of C function calls from JIT code on AArch64.
Patch by Diego Russo
31 changes: 9 additions & 22 deletions Tools/jit/_stencils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,33 +207,20 @@ def emit_aarch64_trampoline(self, hole: Hole, alignment: int) -> None:
return

self.disassembly += [
f"{base + 4 * 0:x}: d2800008 mov x8, #0x0",
f"{base + 4 * 0:016x}: R_AARCH64_MOVW_UABS_G0_NC {hole.symbol}",
f"{base + 4 * 1:x}: f2a00008 movk x8, #0x0, lsl #16",
f"{base + 4 * 1:016x}: R_AARCH64_MOVW_UABS_G1_NC {hole.symbol}",
f"{base + 4 * 2:x}: f2c00008 movk x8, #0x0, lsl #32",
f"{base + 4 * 2:016x}: R_AARCH64_MOVW_UABS_G2_NC {hole.symbol}",
f"{base + 4 * 3:x}: f2e00008 movk x8, #0x0, lsl #48",
f"{base + 4 * 3:016x}: R_AARCH64_MOVW_UABS_G3 {hole.symbol}",
f"{base + 4 * 4:x}: d61f0100 br x8",
f"{base + 4 * 0:x}: 58000048 ldr x8, 8",
f"{base + 4 * 1:x}: d61f0100 br x8",
f"{base + 4 * 2:x}: 00000000",
f"{base + 4 * 2:016x}: R_AARCH64_ABS64 {hole.symbol}",
f"{base + 4 * 3:x}: 00000000",
]
for code in [
0xD2800008.to_bytes(4, sys.byteorder),
0xF2A00008.to_bytes(4, sys.byteorder),
0xF2C00008.to_bytes(4, sys.byteorder),
0xF2E00008.to_bytes(4, sys.byteorder),
0x58000048.to_bytes(4, sys.byteorder),
0xD61F0100.to_bytes(4, sys.byteorder),
0x00000000.to_bytes(4, sys.byteorder),
0x00000000.to_bytes(4, sys.byteorder),
]:
self.body.extend(code)
for i, kind in enumerate(
[
"R_AARCH64_MOVW_UABS_G0_NC",
"R_AARCH64_MOVW_UABS_G1_NC",
"R_AARCH64_MOVW_UABS_G2_NC",
"R_AARCH64_MOVW_UABS_G3",
]
):
self.holes.append(hole.replace(offset=base + 4 * i, kind=kind))
self.holes.append(hole.replace(offset=base + 8, kind="R_AARCH64_ABS64"))
self.trampolines[hole.symbol] = base

def remove_jump(self, *, alignment: int = 1) -> None:
Expand Down
Loading