Update llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp

Co-authored-by: Matt Arsenault <[email protected]>

Author: jrbyrnes

llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp

@@ -1086,7 +1086,7 @@ bool AVGPRRewriteScheduleStage::initGCNSchedStage() {
   // either VGPR or AGPR. However, for the unified RF case, we should only be
   // using AGPR if strictly necessary. That is, if the required number of VGPRs
   // exceeds the addressable limit.  Otherwise, we should be stricly using VGPRs
-  // to minimize cross RC copies. Thus, if we are underc this limit, we should
+  // to minimize cross RC copies. Thus, if we are under this limit, we should
   // constrain AVReg- > VReg.
   // TODO: AVReg constraining for non unified case.
   if (!ST.hasGFX90AInsts() || DAG.RegionsWithAVRegs.empty() ||