[NFC] Remove unused ModuleOp argument from emitRedundantThreadPredicate function (#6601)

Signed-off-by: Anatoly Myachev <anatoly.myachev@intel.com>

Author: anmyachev

third_party/nvidia/lib/TritonNVIDIAGPUToLLVM/LoadStoreOpToLLVM.cpp

@@ -52,7 +52,7 @@ Value maybeAnd(RewriterBase &rewriter, Location loc, Value a, Value b) {
 // according to freeVarsMask. The predicate may be null to indicate no
 // predication is required.
 Value emitRedundantThreadPredicate(
-    ModuleOp moduleOp, const llvm::MapVector<StringAttr, int32_t> &freeVarMasks,
+    const llvm::MapVector<StringAttr, int32_t> &freeVarMasks,
     ConversionPatternRewriter &rewriter, Location loc,
     const NVIDIA::TargetInfo &targetInfo) {
   auto b = TritonLLVMOpBuilder(loc, rewriter);
@@ -452,14 +452,13 @@ struct StoreOpConversion : public ConvertOpToLLVMPattern<triton::StoreOp>,
                        << mask << "\n";
     }
 
-    auto moduleOp = op->getParentOfType<ModuleOp>();
     const size_t dtsize =
         std::max<int>(1, valueElemTy.getIntOrFloatBitWidth() / 8);
     const size_t valueElemNBits = dtsize * 8;
 
     auto freeVarMasks = getFreeVariableMasks(ptr.getType());
-    Value threadPred = emitRedundantThreadPredicate(moduleOp, freeVarMasks,
-                                                    rewriter, loc, targetInfo);
+    Value threadPred =
+        emitRedundantThreadPredicate(freeVarMasks, rewriter, loc, targetInfo);
     uint32_t regMask = freeVarMasks[str_attr("reg")];
 
     const int numVecs = elemsPerThread / vec;
@@ -615,8 +614,8 @@ struct AtomicCASOpConversion
                        << " elemsPerThread = " << elemsPerThread << "\n";
 
     auto freeVarMasks = getFreeVariableMasks(op.getPtr().getType());
-    Value threadPred = emitRedundantThreadPredicate(moduleOp, freeVarMasks,
-                                                    rewriter, loc, targetInfo);
+    Value threadPred =
+        emitRedundantThreadPredicate(freeVarMasks, rewriter, loc, targetInfo);
     uint32_t regMask = freeVarMasks[str_attr("reg")];
 
     auto vecTy = vec_ty(valueElemTy, vec);
@@ -831,8 +830,8 @@ struct AtomicRMWOpConversion
                        << " numElems = " << numElems;
 
     auto freeVarMasks = getFreeVariableMasks(ptr.getType());
-    Value threadPred = emitRedundantThreadPredicate(moduleOp, freeVarMasks,
-                                                    rewriter, loc, targetInfo);
+    Value threadPred =
+        emitRedundantThreadPredicate(freeVarMasks, rewriter, loc, targetInfo);
     uint32_t regMask = freeVarMasks[str_attr("reg")];
 
     auto packedTy = vec_ty(valueElemTy, packed);
@@ -1112,14 +1111,13 @@ struct AsyncCopyGlobalToLocalOpConversion
              << vecBytes << " bytes";
     }
 
-    auto moduleOp = op->getParentOfType<ModuleOp>();
     auto freeVarMasks = getFreeVariableMasks(srcTy);
     // NOTE(@peterbell10): We load redundant data on different CTAs, so the data
     // is available in each CTAs respective shared memory. Otherwise, we would
     // need an additional broadcast step to copy the data between CTAs.
     freeVarMasks[str_attr("block")] = 0;
-    Value threadPred = emitRedundantThreadPredicate(moduleOp, freeVarMasks,
-                                                    rewriter, loc, targetInfo);
+    Value threadPred =
+        emitRedundantThreadPredicate(freeVarMasks, rewriter, loc, targetInfo);
     uint32_t regMask = freeVarMasks[str_attr("reg")];
 
     for (int i = 0; i < shmemAddrs.size(); i++) {