[llvm] [AMDGPU] Using feature to define vmem cacheline size (PR #207884)

[Matt Arsenault via llvm-commits]

================
@@ -414,7 +414,7 @@ class SubtargetFeatureDataCacheLineSize <int Value> : SubtargetFeature <
   "Data (VMEM) cache line size in bytes."
 >;
 
-def FeatureDataCacheLineSize128 : SubtargetFeatureDataCacheLineSize<128>;
+def FeatureCanPrefetchData128 : SubtargetFeatureDataCacheLineSize<128>;
----------------
arsenm wrote:

I meant the TTI hook should change to `if (HasPreftch) return FeatureDerivedcachelineSize`, not give the feature double duty 

https://github.com/llvm/llvm-project/pull/207884