[PATCH] D72445: [ARM,MVE] Add missing IntrNoMem flag on IR intrinsics.

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

I agree that IntrNoMem makes a very sensible default.



================
Comment at: llvm/include/llvm/IR/IntrinsicsARM.td:898
 defm int_arm_mve_vcvt_narrow: MVEPredicated<[llvm_v8f16_ty],
    [llvm_v8f16_ty, llvm_v4f32_ty, llvm_i32_ty], llvm_v4i1_ty, [IntrNoMem]>;
 
----------------
Can we remove it here?


================
Comment at: llvm/include/llvm/IR/IntrinsicsARM.td:1050
   [llvm_i32_ty, llvm_i32_ty], [LLVMMatchType<0>, LLVMMatchType<0>],
    LLVMMatchType<0>, llvm_anyvector_ty, [IntrNoMem]>;
 
----------------
Same Q for this one and the ones below too.


================
Comment at: llvm/include/llvm/IR/IntrinsicsARM.td:1075
 
 def int_arm_cls: Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrNoMem]>;
 def int_arm_cls64: Intrinsic<[llvm_i32_ty], [llvm_i64_ty], [IntrNoMem]>;
----------------
Completely of topic but how come there is a cls intrinsics in the middle of the mve's? Can you move it up to near something similar, whilst you are here.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72445/new/

https://reviews.llvm.org/D72445