Why do we need MDString for metadata in LLVM

Beginners
#1

Hi, I am new to llvm and have some question about metadata:
Why do we use the MDString class to store string? Why do not just use the string to store?

#2

The short answer as to why MDString is necessary is because everything that can be stuffed into metadata needs to have a common base class (in this case, Metadata). So strings and integers end up needing to be packed into ConstantAsMetadata and MDString, respectively.

#3

Thanks for your reply. You mean the operand of a MDNode is the subclass of metadata.

And I have another question:
What is the MDTuple used for? Is there an example to show? I just can not imagine where should we use the MDTuple.

#4

I think most of the non-debug metadata uses tuples in some way, it’s just a list of other nodes. For example compiling an empty file:

!llvm.module.flags = !{!0, !1, !2, !3, !4, !5, !6, !7}
!llvm.ident = !{!8}

!0 = !{i32 1, !"wchar_size", i32 4}
!1 = !{i32 1, !"branch-target-enforcement", i32 0}
!2 = !{i32 1, !"sign-return-address", i32 0}
!3 = !{i32 1, !"sign-return-address-all", i32 0}
!4 = !{i32 1, !"sign-return-address-with-bkey", i32 0}
!5 = !{i32 7, !"PIC Level", i32 2}
!6 = !{i32 7, !"uwtable", i32 2}
!7 = !{i32 7, !"frame-pointer", i32 1}
!8 = !{!"clang version 15.0.0 (ssh://git@repo hash)"}

All of those !{ ... } are MDTuples I believe.

1 Like
#5

Thanks a lot! I traced with GDB and confirm it. You are right.