Index: mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
===================================================================
--- mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
+++ mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
@@ -1117,7 +1117,8 @@
 
   let arguments = (ins Arg<GPU_MMAMemRef, "", [MemRead]>:$srcMemref,
                   Variadic<Index>:$indices,
-                  IndexAttr:$leadDimension);
+                  IndexAttr:$leadDimension,
+                  OptionalAttr<BoolAttr>:$transpose);
 
   let results = (outs GPU_MMAMatrix:$res);
 
Index: mlir/lib/Conversion/GPUToSPIRV/WmmaOpsToSPIRV.cpp
===================================================================
--- mlir/lib/Conversion/GPUToSPIRV/WmmaOpsToSPIRV.cpp
+++ mlir/lib/Conversion/GPUToSPIRV/WmmaOpsToSPIRV.cpp
@@ -87,8 +87,12 @@
     auto i32Type = rewriter.getI32Type();
     auto strideValue = rewriter.create<spirv::ConstantOp>(
         loc, i32Type, IntegerAttr::get(i32Type, stride));
+    bool useColMajor = false;
+    if (auto transposeLoad = subgroupMmaLoadMatrixOp.getTranspose()) {
+      useColMajor = transposeLoad.value();
+    }
     auto coloumnMajor = rewriter.create<spirv::ConstantOp>(
-        loc, rewriter.getI1Type(), rewriter.getBoolAttr(false));
+        loc, rewriter.getI1Type(), rewriter.getBoolAttr(useColMajor));
     rewriter.replaceOpWithNewOp<spirv::NVCooperativeMatrixLoadOp>(
         subgroupMmaLoadMatrixOp, coopType, bufferPtr, strideValue, coloumnMajor,
         spirv::MemoryAccessAttr());
Index: mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
===================================================================
--- mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
+++ mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
@@ -127,11 +127,12 @@
   AffineExpr zero = b.getAffineConstantExpr(0);
   auto broadcastInnerDim = AffineMap::get(map.getNumDims(), 0, {zero, innerDim},
                                           readOp.getContext());
+  bool isTranspose = !map.isMinorIdentity() && map.getNumDims() == 2;
 
   if (!useNvGpu) {
-    // TODO: Support transpose once it is added to GPU dialect ops.
-    // For now we only support (d0, d1) -> (d0, d1) and (d0, d1) -> (0, d1).
-    return map.isMinorIdentity() || map == broadcastInnerDim;
+    bool result =
+        map.isMinorIdentity() || map == broadcastInnerDim || isTranspose;
+    return result;
   }
 
   return true;
@@ -445,9 +446,10 @@
       gpu::MMAMatrixType::get(op.getVectorType().getShape(),
                               op.getVectorType().getElementType(), fragType);
   OpBuilder b(op);
+  bool isTranspose = !map.isMinorIdentity() && map.getNumDims() == 2;
   Value load = b.create<gpu::SubgroupMmaLoadMatrixOp>(
       op.getLoc(), type, op.getSource(), op.getIndices(),
-      b.getIndexAttr(*stride));
+      b.getIndexAttr(*stride), b.getBoolAttr(isTranspose));
   valueMapping[op.getResult()] = load;
 }
 
Index: mlir/test/Conversion/VectorToGPU/vector-to-mma-ops.mlir
===================================================================
--- mlir/test/Conversion/VectorToGPU/vector-to-mma-ops.mlir
+++ mlir/test/Conversion/VectorToGPU/vector-to-mma-ops.mlir
@@ -5,11 +5,12 @@
 #map2 = affine_map<(d0, d1, d2) -> (d1, d2)>
 #map3 = affine_map<(d0, d1, d2) -> (d0, d1)>
 #map4 = affine_map<(d0) -> (d0, 0)>
+#map5 = affine_map<(d0, d1) -> (d0, d1)>
 
 // CHECK-LABEL: func @matmul
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
-//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<16x16xf16>
 func.func @matmul(%arg0: memref<16x16xf16>, %arg1: memref<16x16xf16>, %arg2: memref<16x16xf16>) {
@@ -26,8 +27,8 @@
 
 // CHECK-LABEL: func @matmul_cst
 //   CHECK-DAG:   %[[CST:.+]] = arith.constant 0.000000e+00 : f16
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
 //   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_constant_matrix %[[CST]] : !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<16x16xf16>
@@ -45,8 +46,8 @@
 // CHECK-LABEL: func @matmul_broadcast
 //  CHECK-SAME:   (%{{.*}}: memref<16x16xf16>, %{{.*}}: memref<16x16xf16>, %{{.*}}: memref<16x16xf16>, %[[F:.*]]: f16)
 //   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_constant_matrix %[[F]] : !gpu.mma_matrix<16x16xf16, "COp">
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<16x16xf16>
 func.func @matmul_broadcast(%arg0: memref<16x16xf16>, %arg1: memref<16x16xf16>, %arg2: memref<16x16xf16>, %f: f16) {
@@ -61,10 +62,10 @@
 }
 
 // CHECK-LABEL: func @matmul_loop
-//       CHECK:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//       CHECK:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index, transpose = false} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[ACC:.+]] = scf.for {{.*}} iter_args(%[[ACC1:.+]] = %[[C]]) -> (!gpu.mma_matrix<16x16xf16, "COp">) {
-//   CHECK-DAG:     %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:     %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:     %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index, transpose = false} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:     %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 128 : index, transpose = false} : memref<128x128xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
 //  CHECK-NEXT:     %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[ACC1]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //  CHECK-NEXT:     scf.yield %[[D]] : !gpu.mma_matrix<16x16xf16, "COp">
 //  CHECK-NEXT:   }
@@ -88,8 +89,8 @@
 // CHECK-LABEL: func @matmul_fused_elementwise
 //   CHECK-DAG:   %[[CST_0:.+]] = arith.constant 0.000000e+00 : f16
 //   CHECK-DAG:   %[[CST_1:.+]] = arith.constant 1.000000e+00 : f16
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
 //   CHECK-DAG:   %[[C0:.+]] = gpu.subgroup_mma_constant_matrix %[[CST_0]] : !gpu.mma_matrix<16x16xf16, "COp">
 //   CHECK-DAG:   %[[C1:.+]] = gpu.subgroup_mma_constant_matrix %[[CST_1]] : !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C0]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
@@ -110,11 +111,11 @@
 
 // CHECK-LABEL: func @matmul_fused_broadcast
 //   CHECK-DAG:   %[[CST_0:.+]] = arith.constant 0.000000e+00 : f16
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
 //   CHECK-DAG:   %[[C0:.+]] = gpu.subgroup_mma_constant_matrix %[[CST_0]] : !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C0]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
-//       CHECK:   %[[E:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] {leadDimension = 0 : index} : memref<16x16x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//       CHECK:   %[[E:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] {leadDimension = 0 : index, transpose = false} : memref<16x16x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[F:.+]] = gpu.subgroup_mma_elementwise divf %[[D]], %[[E]] : (!gpu.mma_matrix<16x16xf16, "COp">, !gpu.mma_matrix<16x16xf16, "COp">) -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[F]], %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<16x16xf16>
 func.func @matmul_fused_broadcast(%arg0: memref<16x16xf16>, %arg1: memref<16x16xf16>,
@@ -135,9 +136,9 @@
 
 // CHECK-LABEL: func @matmul_3Dmemref
 //   CHECK-DAG:   %[[C0:.+]] = arith.constant 0 : index
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]]] {leadDimension = 0 : index} : memref<16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
-//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index, transpose = false} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]]] {leadDimension = 0 : index, transpose = false} : memref<16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index, transpose = false} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<2x16x16xf16>
 func.func @matmul_3Dmemref(%arg0: memref<2x16x16xf16>, %arg1: memref<16xf16>, %arg2: memref<2x16x16xf16>) {
@@ -154,9 +155,9 @@
 
 // CHECK-LABEL: func @matmul_memref_strided
 //   CHECK-DAG:   %[[C0:.+]] = arith.constant 0 : index
-//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 32 : index} : memref<2x16x16xf16, #{{.*}}> -> !gpu.mma_matrix<16x16xf16, "AOp">
-//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]]] {leadDimension = 0 : index} : memref<16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
-//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 32 : index, transpose = false} : memref<2x16x16xf16, #{{.*}}> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]]] {leadDimension = 0 : index, transpose = false} : memref<16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index, transpose = false} : memref<2x16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
 //       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%[[C0]], %[[C0]], %[[C0]]] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<2x16x16xf16>
 func.func @matmul_memref_strided(%arg0: memref<2x16x16xf16, affine_map<(d0, d1, d2) -> (d0 * 512 + d1 * 32 + d2)>>, %arg1: memref<16xf16>, %arg2: memref<2x16x16xf16>) {
@@ -170,3 +171,21 @@
   vector.transfer_write %D, %arg2[%c0, %c0, %c0] {in_bounds = [true, true]} : vector<16x16xf16>, memref<2x16x16xf16>
   return
 }
+
+// CHECK-LABEL: func @matmul_transposed
+//   CHECK-DAG:   %[[A:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "AOp">
+//   CHECK-DAG:   %[[B:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = true} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "BOp">
+//   CHECK-DAG:   %[[C:.+]] = gpu.subgroup_mma_load_matrix %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index, transpose = false} : memref<16x16xf16> -> !gpu.mma_matrix<16x16xf16, "COp">
+//       CHECK:   %[[D:.+]] = gpu.subgroup_mma_compute %[[A]], %[[B]], %[[C]] : !gpu.mma_matrix<16x16xf16, "AOp">, !gpu.mma_matrix<16x16xf16, "BOp"> -> !gpu.mma_matrix<16x16xf16, "COp">
+//       CHECK:   gpu.subgroup_mma_store_matrix %[[D]], %{{.*}}[%{{.*}}, %{{.*}}] {leadDimension = 16 : index} : !gpu.mma_matrix<16x16xf16, "COp">, memref<16x16xf16>
+func.func @matmul_transposed(%arg0: memref<16x16xf16>, %arg1: memref<16x16xf16>, %arg2: memref<16x16xf16>) {
+  %cst_0 = arith.constant dense<0.000000e+00> : vector<16x16xf16>
+  %c0 = arith.constant 0 : index
+  %cst = arith.constant 0.000000e+00 : f16
+  %A = vector.transfer_read %arg0[%c0, %c0], %cst {in_bounds = [true, true]} : memref<16x16xf16>, vector<16x16xf16>
+  %B = vector.transfer_read %arg1[%c0, %c0], %cst {permutation_map = #map5, in_bounds = [true, true]} : memref<16x16xf16>, vector<16x16xf16>
+  %C = vector.transfer_read %arg2[%c0, %c0], %cst {in_bounds = [true, true]} : memref<16x16xf16>, vector<16x16xf16>
+  %D = vector.contract {indexing_maps = [#map1, #map2, #map3], iterator_types = ["parallel", "parallel", "reduction"], kind = #vector.kind<add>} %A, %B, %C : vector<16x16xf16>, vector<16x16xf16> into vector<16x16xf16>
+  vector.transfer_write %D, %arg2[%c0, %c0] {in_bounds = [true, true]} : vector<16x16xf16>, memref<16x16xf16>
+  return
+}