[llvm-bugs] [Bug 52418] New: Incomplete support for `--target`in the code-gen layer
llvm-bugs at lists.llvm.org
Fri Nov 5 03:34:50 PDT 2021
Bug ID: 52418
Summary: Incomplete support for `--target`in the code-gen layer
Assignee: unassignedbugs at nondot.org
Reporter: andrzej.warzynski at gmail.com
CC: David.Truby at arm.com, jperier at nvidia.com,
kirankumartp at gmail.com, llvm-bugs at lists.llvm.org,
sscalpone at nvidia.com
This came up during a discussion on https://reviews.llvm.org/D112961.
As per the comment at the top of
"flang/lib/Optimizer/CodeGen/DescriptorModel.h", various assumptions about the
target machine are made in the code-gen layer. This is likely to cause
`--target` (or e.g. `-m64` and `-m32`) to malfunction when using Flang's
interoperability with C (see how various CFI structures are lowered).
Note that `--target` is not available in the driver yet. Also, it is possible
that the assumptions made when lowering the CFI structures will affect other
parts of the code-gen too. However, I suspect that most things will work
correctly when generating code for the host/native target.
I'm creating this issue as a placeholder and as a note for our future selves.
This is currently _not_ being investigated or worked on.
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the llvm-bugs