[llvm-bugs] [Bug 50348] New: Enable modifiers on V_MOV_B32

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=50348

            Bug ID: 50348
           Summary: Enable modifiers on V_MOV_B32
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: AMDGPU
          Assignee: unassignedbugs at nondot.org
          Reporter: joseph.nash at amd.com
                CC: llvm-bugs at lists.llvm.org

According to the spec, while specified to use arbitrary data formats, FP
modifiers (negation,absolute value, clamp and omod) are allowed on this
instruction. But the code does not allow this.

VOP1Instructions.td:181
defm V_MOV_B32 : VOP1Inst <"v_mov_b32", VOP_I32_I32>;

Part of the limitation is due to the use of a VOP_I32_I32 profile, which
disallows modifiers. 

Due to the use of MOV instructions explicitly in several AMDGPU backend passes,
additional checks may have to be added to those passes if modifiers are allowed
on V_MOV_B32

If enabled, this feature could perhaps result in more optimal code, when a
value needs to have modifiers applied.

-- 
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/20210514/d542733f/attachment.html>