[llvm] [AMDGPU] Explicit instruction prefetch for large entry functions (PR #227995)

[Lukas Sommer via llvm-commits]

================
@@ -220,6 +225,15 @@ void AMDGPUAsmPrinter::emitFunctionBodyStart() {
     HSAMetadataStream->emitKernel(*MF, CurrentProgramInfo);
 }
 
+void AMDGPUAsmPrinter::emitFunctionBodyEnd() {
+  // Emit the label for the prefetch end symbol if ICache prefetch is enabled.
+  // This symbol was created in emitFunctionBodyStart for this function.
+  if (PrefetchEndSym) {
+    OutStreamer->emitLabel(PrefetchEndSym);
+  }
----------------
sommerlukas wrote:

Done.

https://github.com/llvm/llvm-project/pull/227995