Commits on Oct 23, 2024

1. fix: don't clobber saved frame pointer in arm64 assembly functions …

   The arm64 neon assembly functions in this repository overwrite the frame
   pointer saved by their callers, leading to crashes from the Go runtime
   execution tracer and profilers which use frame pointer unwinding. For
   historical reasons, on arm64 Go functions save the caller's frame
   pointer register (x29) one word below their stack frame. See
   go.dev/s/regabi#arm64-architecture. The assembly functions here,
   translated from C compiler output, save values at the top of their
   frame, and overwrite the frame pointer saved by the caller. We can fix
   this by decrementing the stack pointer past where that frame pointer is
   saved before saving anything on the stack.
   
   Fixed with this sed script on my macos laptop + manual cleanup to match
   indentation:
   
   ```sed
   /stp[\t ]*x29/i\
   	// The Go ABI saves the frame pointer register one word below the \
   	// caller's frame. Make room so we don't overwrite it. Needs to stay \
   	// 16-byte aligned \
   	SUB $16, RSP
   
   /ldp[\t ]*x29/a\
   	// Put the stack pointer back where it was \
   	ADD $16, RSP
   
   ```
   
   Ran the script from the root of this repository with
   
   	find . -name '*_arm64.s' -exec sed -f fix.sed -i '' {} +
   
   Then manually inspected the assembly for missing SUBs/ADDs at the
   beginning of functions and prior to returns.
   
   Fixes apache#150

   

   nsrip-dd committed Oct 23, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for 7cfd90a
   • Browse repository at this point

   Copy the full SHA

   7cfd90a View commit details

   Browse the repository at this point in the history