[llvm] [AMDGPU] Add shuffle optimizer pass (PR #155824)

[Matt Arsenault via llvm-commits]

================
@@ -3095,10 +3095,6 @@ def int_amdgcn_wmma_f32_16x16x16_bf8_bf8 : AMDGPUWmmaIntrinsic<llvm_anyint_ty, l
 // A and B are <16 x iu4>.
 def int_amdgcn_wmma_i32_16x16x32_iu4     : AMDGPUWmmaIntrinsicIU<llvm_anyint_ty, llvm_anyint_ty>;
 
-def int_amdgcn_swmmac_f32_16x16x32_f16     : AMDGPUSWmmacIntrinsicIdx<llvm_anyfloat_ty, llvm_anyfloat_ty, llvm_anyfloat_ty, llvm_anyint_ty>;
-def int_amdgcn_swmmac_f32_16x16x32_bf16    : AMDGPUSWmmacIntrinsicIdx<llvm_anyint_ty, llvm_anyint_ty, llvm_anyfloat_ty, llvm_anyint_ty>;
-def int_amdgcn_swmmac_f16_16x16x32_f16     : AMDGPUSWmmacIntrinsicIdx<llvm_anyfloat_ty, llvm_anyfloat_ty, llvm_anyfloat_ty, llvm_anyint_ty>;
-def int_amdgcn_swmmac_bf16_16x16x32_bf16   : AMDGPUSWmmacIntrinsicIdx<llvm_anyint_ty, llvm_anyint_ty, llvm_anyint_ty, llvm_anyint_ty>;
----------------
arsenm wrote:

Why are these removed? 

https://github.com/llvm/llvm-project/pull/155824