`exp` on `SIMD.Vec` triggers LLVM error

To reproduce

The Minimal Working Example (MWE) for this bug:

using CUDA
using Adapt
using SIMD
using KernelAbstractions

@inline function vload(::Type{SIMD.Vec{N, T}}, ptr::Core.LLVMPtr{T, AS}) where {N, T, AS}
    alignment = sizeof(T) * N
    vec_ptr = Base.bitcast(Core.LLVMPtr{SIMD.Vec{N, T}, AS}, ptr)
    return unsafe_load(vec_ptr, 1, Val(alignment))
end

@inline function vstore!(ptr::Core.LLVMPtr{T, AS}, x::SIMD.Vec{N, T}) where {N, T, AS}
    alignment = sizeof(T) * N
    vec_ptr = Base.bitcast(Core.LLVMPtr{SIMD.Vec{N, T}, AS}, ptr)
    unsafe_store!(vec_ptr, x, 1, Val(alignment))
    return
end

@kernel function ker!(y, x)
    i = @index(Global)
    v = vload(SIMD.Vec{4, Float32}, pointer(x))
    v = exp(v)
    vstore!(pointer(y), v)
end

function tt(kab)
    x = Adapt.adapt(kab, ones(Float32, 4))
    y = Adapt.adapt(kab, ones(Float32, 4))
    ker!(kab)(y, x; ndrange=1)
    @show y
    return
end
tt(CUDABackend())

Error:

ERROR: LLVM error: Undefined external symbol "expf"
Stacktrace:
  [1] handle_error(reason::Cstring)
    @ LLVM ~/.julia/packages/LLVM/b3kFs/src/core/context.jl:194
  [2] LLVMTargetMachineEmitToMemoryBuffer(T::LLVM.TargetMachine, M::LLVM.Module, codegen::LLVM.API.LLVMCodeGenFileType, ErrorMessage::Base.RefValue{Cstring}, OutMemBuf::Base.RefValue{Ptr{…}})
    @ LLVM.API ~/.julia/packages/LLVM/b3kFs/lib/16/libLLVM.jl:11138
  [3] emit(tm::LLVM.TargetMachine, mod::LLVM.Module, filetype::LLVM.API.LLVMCodeGenFileType)
    @ LLVM ~/.julia/packages/LLVM/b3kFs/src/targetmachine.jl:118
  [4] mcgen(job::GPUCompiler.CompilerJob, mod::LLVM.Module, format::LLVM.API.LLVMCodeGenFileType)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/OGnEB/src/mcgen.jl:75
  [5] mcgen(job::GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, mod::LLVM.Module, format::LLVM.API.LLVMCodeGenFileType)
    @ CUDA ~/.julia/packages/CUDA/1kIOw/src/compiler/compilation.jl:127
  [6] macro expansion
    @ ~/.julia/packages/GPUCompiler/OGnEB/src/driver.jl:400 [inlined]
  [7] emit_asm(job::GPUCompiler.CompilerJob, ir::LLVM.Module; strip::Bool, validate::Bool, format::LLVM.API.LLVMCodeGenFileType)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/OGnEB/src/utils.jl:110
  [8] emit_asm
    @ ~/.julia/packages/GPUCompiler/OGnEB/src/utils.jl:108 [inlined]
  [9] codegen(output::Symbol, job::GPUCompiler.CompilerJob; toplevel::Bool, libraries::Bool, optimize::Bool, cleanup::Bool, validate::Bool, strip::Bool, only_entry::Bool, parent_job::Nothing)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/OGnEB/src/driver.jl:120
...

Start by running the provided Julia MWE with CUDA, Adapt, SIMD, and KernelAbstractions to reproduce the LLVM error. Trace compilation of the exp(v) call through the GPUCompiler path and inspect why expf remains an undefined external symbol. Done means the kernel compiles and runs without this LLVM error.