llvm error on windows amd

[paulxshen]

hello:

using BFloat16s
a = randn(BFloat16, 1000)
a + a

Julia Version 1.11.5
Commit 760b2e5b73 (2025-04-14 06:53 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: 16 × AMD Ryzen 7 8845HS w/ Radeon 780M Graphics
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver3)
Threads: 4 default, 0 interactive, 2 GC (on 16 virtual cores)
Environment:
  JULIA_PKG_PRECOMPILE_AUTO = 0

LLVM ERROR: Cannot select: 0x27ffa87e6a0: v16bf16 = X86ISD::VFPROUND 0x27ffa87e240, C:\Users\default.LAPTOP-HMRU58MH\.julia\packages\BFloat16s\LIFRh\src\bfloat16.jl:236 @[ broadcast.jl:678 @[ broadcast.jl:651 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ]
  0x27ffa87e240: v16f32 = fadd 0x27ffa88c4e0, 0x27ffa87ec50, C:\Users\default.LAPTOP-HMRU58MH\.julia\packages\BFloat16s\LIFRh\src\bfloat16.jl:236 @[ broadcast.jl:678 @[ broadcast.jl:651 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ]
    0x27ffa88c4e0: v16f32 = fp_extend 0x27ffa8800e0, C:\Users\default.LAPTOP-HMRU58MH\.julia\packages\BFloat16s\LIFRh\src\bfloat16.jl:236 @[ broadcast.jl:678 @[ broadcast.jl:651 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ]
      0x27ffa8800e0: v16bf16,ch = load<(load (s256) from %ir.uglygep165 + 32, align 2, !tbaa !192, !alias.scope !248, !noalias !197)> 0x27ffa23a310, 0x27ffa88c780, undef:i64, C:\Users\default.LAPTOP-HMRU58MH\.julia\packages\BFloat16s\LIFRh\src\bfloat16.jl:236 @[ broadcast.jl:678 @[ broadcast.jl:651 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ]
        0x27ffa88c780: i64 = add 0x27ffa87fc10, Constant:i64<32>, C:\Users\default.LAPTOP-HMRU58MH\.julia\packages\BFloat16s\LIFRh\src\bfloat16.jl:236 @[ broadcast.jl:678 @[ broadcast.jl:651 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ]
          0x27ffa87fc10: i64 = add 0x27ffa88c710, 0x27ffa87e470, essentials.jl:917 @[ broadcast.jl:644 @[ broadcast.jl:674 @[ broadcast.jl:650 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ] ]
            0x27ffa88c710: i64,ch = CopyFromReg 0x27ffa23a310, Register:i64 %28, essentials.jl:917 @[ broadcast.jl:644 @[ broadcast.jl:674 @[ broadcast.jl:650 @[ broadcast.jl:610 @[ broadcast.jl:973 @[ simdloop.jl:77 @[ broadcast.jl:972 @[ broadcast.jl:925 @[ broadcast.jl:897 @[ broadcast.jl:872 @[ broadcast.jl:861 @[ arraymath.jl:16 ] ] ] ] ] ] ] ] ] ] ] ]
...

but scalar works

sin(BFloat16(2))
BFloat16(0.91015625)

thoughts? thanks :)