https://bugs.llvm.org/show_bug.cgi?id=35970

            Bug ID: 35970
           Summary: Implement addrspace machine memory operand printing /
                    parsing in MIR
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedb...@nondot.org
          Reporter: franci...@yahoo.com
                CC: llvm-bugs@lists.llvm.org

Created attachment 19686
  --> https://bugs.llvm.org/attachment.cgi?id=19686&action=edit
Copy of the test.

-debug printing prints the address space of a MachinePointerInfo if it is != 0:

%15:vgpr_32 = BUFFER_LOAD_SSHORT_OFFEN %18, %sgpr96_sgpr97_sgpr98_sgpr99,
%sgpr3, 0, 0, 0, 0, implicit %exec; mem:LD2[<unknown>(addrspace=5)]

When printing MIR and parsing it again we lose the information of the address
space if there is no IR value attached to the MachinePointerInfo.

Example: test/CodeGen/AMDGPU/extload-align.ll

We print:

%15:vgpr_32 = BUFFER_LOAD_SSHORT_OFFEN %18, %sgpr96_sgpr97_sgpr98_sgpr99,
%sgpr3, 0, 0, 0, 0, implicit %exec :: (load 2)

Note the ":: (load 2)" which is missing the addrspace 5.

When we parse it back, the addrspace is assumed to be 0.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs

Reply via email to