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

[via llvm-commits]

================
@@ -2149,6 +2158,7 @@ def FeatureISAVersion12 : FeatureSet<
    FeatureRealTrue16Insts,
    FeatureMTBUFInsts,
    FeatureFormattedMUBUFInsts,
+   FeatureDataCacheLineSize128
----------------
Shoreshen wrote:

Hi @arsenm , I think the semantic match for this llvm.prefetch is vmem data cache instruction. Which is actually matching feature `VmemPrefInsts`.

If function returns 0, then no prefetch intrinsic will be created.

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