[llvm-bugs] [Bug 51033] New: [DebugInfo] Poor support for x87 variable locations

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=51033

            Bug ID: 51033
           Summary: [DebugInfo] Poor support for x87 variable locations
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Keywords: wrong-debug
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: jeremy.morse.llvm at gmail.com
                CC: llvm-bugs at lists.llvm.org
            Blocks: 38768

Observed over in https://reviews.llvm.org/D105657 , both our current variable
location implementation and the instruction referencing stuff I've been working
on are unable to cope with x87 variable locations. Both are unable to actually
express where in registers variable values go.

The root cause is the baroqueness of x87 itself: it refers to registers as a
stack, rather than just naming them directly. If we had a DBG_VALUE that
referred to $st(1), we might need to issue additional DBG_VALUEs as the stack
position of the value changed due to other FP instructions.

It's a significant pain in the neck, and VarLoc LiveDebugValues doesn't try to
solve it. InstrRef LiveDebugValues doesn't try either. I suspect that interest
in this matter is low (what with i386 processors being out of fashion for at
least a decade) but thought I'd open the ticket in case there's something out
there who this is relevant to.


Referenced Bugs:

https://bugs.llvm.org/show_bug.cgi?id=38768
[Bug 38768] [meta][DebugInfo] Umbrella bug for poor debug experiences
-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210708/960552a8/attachment.html>