[clang] [llvm] [AMDGPU] Add suffix _d4 to tensor load/store with 4 groups D#, NFC (PR #184176)

[Changpeng Fang via cfe-commits]

================
@@ -2077,8 +2077,8 @@ class VIMAGE_TENSOR_Pseudo<string opName, bit _UpTo2D = 0> :
 }
 
 let SubtargetPredicate = isGFX125xOnly in {
-def TENSOR_LOAD_TO_LDS    : VIMAGE_TENSOR_Pseudo<"tensor_load_to_lds">;
-def TENSOR_STORE_FROM_LDS : VIMAGE_TENSOR_Pseudo<"tensor_store_from_lds">;
+def TENSOR_LOAD_TO_LDS_D4    : VIMAGE_TENSOR_Pseudo<"tensor_load_to_lds">;
----------------
changpeng wrote:

> Compiler invented suffixes should use lowercase

Done. Using _d2 and _d4 as suffixes now

https://github.com/llvm/llvm-project/pull/184176