[flang-commits] [llvm] [libcxxabi] [clang-tools-extra] [lldb] [clang] [lld] [compiler-rt] [flang] [libcxx] [AMDGPU] GFX12: select @llvm.prefetch intrinsic (PR #74576)

[Jay Foad via flang-commits]

================
@@ -959,6 +967,32 @@ def : GCNPat <
 }
 } // let OtherPredicates = [HasShaderCyclesRegister]
 
+def SIMM24bitPtr : ImmLeaf <iPTR,
+  [{return isInt<24>(Imm);}]
+>;
+
+multiclass SMPrefetchPat<string type, int cache_type> {
+  def : GCNPat <
+    (smrd_prefetch (SMRDImm i64:$sbase, i32:$offset), timm, timm, (i32 cache_type)),
+    (!cast<SM_Prefetch_Pseudo>("S_PREFETCH_"#type) $sbase, $offset, (i32 SGPR_NULL), (i8 0))
+  >;
+
+  def : GCNPat <
+    (smrd_prefetch (i64 SReg_64:$sbase), timm, timm, (i32 cache_type)),
+    (!cast<SM_Prefetch_Pseudo>("S_PREFETCH_"#type) $sbase, 0, (i32 SGPR_NULL), (i8 0))
+  >;
+
+  def : GCNPat <
+    (prefetch SIMM24bitPtr:$offset, timm, timm, (i32 cache_type)),
+    (!cast<SM_Prefetch_Pseudo>("S_PREFETCH_"#type#"_PC_REL") (as_i32timm $offset), (i32 SGPR_NULL), (i8 0))
+  > {
+    let AddedComplexity = 10;
+  }
----------------
jayfoad wrote:

But that is how `llvm.prefetch` is defined: "`address` is the address to be prefetched". A different operation should use a different intrinsic.

https://github.com/llvm/llvm-project/pull/74576