diff --git a/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td b/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td
--- a/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td
+++ b/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td
@@ -217,7 +217,7 @@
   }];
 
   let builders = [
-    OpBuilder<(ins "Value":$output, "Value":$value)>
+    OpBuilder<(ins "Value":$value, "Value":$output)>
   ];
 
   let verifier = [{ return ::verify(*this); }];
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -813,7 +813,7 @@
 
   auto fillValue = rewriter.create<ConstantOp>(loc, fillValueAttr);
   auto filledTensor =
-      rewriter.create<linalg::FillOp>(loc, initTensor, fillValue).result();
+      rewriter.create<linalg::FillOp>(loc, fillValue, initTensor).result();
 
   SmallVector<AffineExpr, 2> srcExprs;
   SmallVector<AffineExpr, 2> dstExprs;
@@ -1018,7 +1018,7 @@
     auto initTensor = rewriter.create<linalg::InitTensorOp>(
         loc, outputTy.getShape(), outputTy.getElementType());
     Value zeroTensor =
-        rewriter.create<linalg::FillOp>(loc, initTensor, zero).getResult(0);
+        rewriter.create<linalg::FillOp>(loc, zero, initTensor).getResult(0);
     rewriter.replaceOpWithNewOp<linalg::BatchMatmulOp>(
         op, TypeRange{op.getType()}, ValueRange{adaptor.a(), adaptor.b()},
         ValueRange{zeroTensor});
@@ -1093,7 +1093,6 @@
   }
 };
 
-
 class ReshapeConverter : public OpConversionPattern<tosa::ReshapeOp> {
 public:
   using OpConversionPattern<tosa::ReshapeOp>::OpConversionPattern;
@@ -1737,7 +1736,7 @@
     Value zeroVal = rewriter.create<ConstantOp>(
         loc, rewriter.getZeroAttr(resultType.getElementType()));
     Value result =
-        rewriter.create<linalg::FillOp>(loc, init, zeroVal).getResult(0);
+        rewriter.create<linalg::FillOp>(loc, zeroVal, init).getResult(0);
 
     for (auto arg : args) {
       sizes[axis] = rewriter.create<memref::DimOp>(loc, arg, axisValue);
@@ -1981,7 +1980,7 @@
     auto fillValueIdx = rewriter.create<ConstantOp>(
         loc, rewriter.getIntegerAttr(outElementTy, 0));
     auto filledTensorIdx =
-        rewriter.create<linalg::FillOp>(loc, initTensorIdx, fillValueIdx)
+        rewriter.create<linalg::FillOp>(loc, fillValueIdx, initTensorIdx)
             .result();
 
     // Second fill the output buffer for the running max.
@@ -1999,7 +1998,7 @@
 
     auto fillValueMax = rewriter.create<ConstantOp>(loc, fillValueMaxAttr);
     auto filledTensorMax =
-        rewriter.create<linalg::FillOp>(loc, initTensorMax, fillValueMax)
+        rewriter.create<linalg::FillOp>(loc, fillValueMax, initTensorMax)
             .result();
 
     // We need to reduce along the arg-max axis, with parallel operations along
@@ -2288,7 +2287,7 @@
         loc, resultTy.getShape(), resultTy.getElementType());
 
     Value filledInitTensor =
-        rewriter.create<linalg::FillOp>(loc, initTensor, initialValue).result();
+        rewriter.create<linalg::FillOp>(loc, initialValue, initTensor).result();
 
     Value fakeWindowDims =
         rewriter.create<linalg::InitTensorOp>(loc, kernel, outElementTy);
diff --git a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
--- a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
+++ b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
@@ -426,8 +426,8 @@
   b.create<linalg::YieldOp>(block.getArgument(0));
 }
 
-void FillOp::build(OpBuilder &builder, OperationState &result, Value output,
-                   Value value) {
+void FillOp::build(OpBuilder &builder, OperationState &result, Value value,
+                   Value output) {
   build(builder, result, output.getType().dyn_cast<RankedTensorType>(), value,
         output);
   fillStructuredOpRegion<FillOp>(builder, *result.regions.front(),
@@ -1966,7 +1966,7 @@
     auto newInit = rewriter.create<TensorReshapeOp>(
         loc, reshapeOp.getResultType(), oldFill.output(),
         reshapeOp.reassociation());
-    rewriter.replaceOpWithNewOp<FillOp>(reshapeOp, newInit, oldFill.value());
+    rewriter.replaceOpWithNewOp<FillOp>(reshapeOp, oldFill.value(), newInit);
 
     return success();
   }
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Bufferize.cpp b/mlir/lib/Dialect/Linalg/Transforms/Bufferize.cpp
--- a/mlir/lib/Dialect/Linalg/Transforms/Bufferize.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Bufferize.cpp
@@ -187,7 +187,7 @@
       return rewriter.notifyMatchFailure(op,
                                          "operand must be of a tensor type");
 
-    rewriter.create<FillOp>(op.getLoc(), adaptor.output(), adaptor.value());
+    rewriter.create<FillOp>(op.getLoc(), adaptor.value(), adaptor.output());
     rewriter.replaceOp(op, adaptor.output());
 
     return success();
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp b/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
--- a/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
@@ -283,7 +283,7 @@
             .Default([](auto) { return Value(); });
     if (!fillVal)
       return {};
-    b.create<linalg::FillOp>(promotionInfo->fullLocalView, fillVal);
+    b.create<linalg::FillOp>(fillVal, promotionInfo->fullLocalView);
   }
 
   // Copy data into the promoted buffers. Use callback if provided.
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp b/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
--- a/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
@@ -676,7 +676,7 @@
 
   // Initialize tensor with the pad value
   Value tmpTensor =
-      rewriter.create<linalg::FillOp>(loc, initTensor, padValue).result();
+      rewriter.create<linalg::FillOp>(loc, padValue, initTensor).result();
 
   // Copy original contents into new tensor
   // Uses linalg.generic, but could be done with tensor.insert_slice
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
--- a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
@@ -756,7 +756,7 @@
     // pattern.)
     auto padValue = padOp.getConstantPaddingValue();
     if (padValue)
-      return rewriter.create<FillOp>(padOp.getLoc(), dest, padValue).result();
+      return rewriter.create<FillOp>(padOp.getLoc(), padValue, dest).result();
 
     // Fill could not be vectorized: Lower to tensor::GenerateOp with region.
     auto generateOp = rewriter.create<tensor::GenerateOp>(
diff --git a/mlir/lib/Dialect/Vector/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/VectorTransforms.cpp
--- a/mlir/lib/Dialect/Vector/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/VectorTransforms.cpp
@@ -2440,7 +2440,7 @@
         b.create<scf::YieldOp>(loc, viewAndIndices);
       },
       [&](OpBuilder &b, Location loc) {
-        b.create<linalg::FillOp>(loc, alloc, xferOp.padding());
+        b.create<linalg::FillOp>(loc, xferOp.padding(), alloc);
         // Take partial subview of memref which guarantees no dimension
         // overflows.
         Value memRefSubView = createSubViewIntersection(
diff --git a/mlir/test/lib/Dialect/Linalg/TestLinalgTransforms.cpp b/mlir/test/lib/Dialect/Linalg/TestLinalgTransforms.cpp
--- a/mlir/test/lib/Dialect/Linalg/TestLinalgTransforms.cpp
+++ b/mlir/test/lib/Dialect/Linalg/TestLinalgTransforms.cpp
@@ -309,10 +309,11 @@
   auto floatType = src.getType().cast<MemRefType>().getElementType();
   if (!floatType.isa<FloatType>())
     return failure();
-  if (!isOutput)
-    b.create<FillOp>(
-        src.getLoc(), dst,
-        b.create<ConstantOp>(src.getLoc(), FloatAttr::get(floatType, 42.0)));
+  if (!isOutput) {
+    Value cst =
+        b.create<ConstantOp>(src.getLoc(), FloatAttr::get(floatType, 42.0));
+    b.create<FillOp>(src.getLoc(), cst, dst);
+  }
   b.create<CopyOp>(src.getLoc(), src, dst);
   return success();
 }