ROCm integration tests failing in the presense of multiple kernels #53475
Description
https://github.com/llvm/llvm-project/blob/main/mlir/test/Integration/GPU/ROCM/two-modules.mlir and https://github.com/llvm/llvm-project/blob/main/mlir/test/Integration/GPU/ROCM/vector-transferops.mlir , which are distinguished from the other tests in that batch by having more than one kernel present, fail with errors such as the following
ld.lld: error: section: .dynamic is not contiguous with other relro sections
ld.lld: error: section .text virtual address range overlaps with .rodata
>>> .text range is [0x1600, 0x1627]
>>> .rodata range is [0x1600, 0x163F]
ld.lld: error: section .text load address range overlaps with .rodata
>>> .text range is [0x1600, 0x1627]
>>> .rodata range is [0x1600, 0x163F]
<unknown>:0: error: lld invocation error
(for the above failures, %chip is gfx908).
@aganea since you were on the other LLD change that broke us and might know more about how this happened - if I've @-ed the wrong person, apologies.