diff --git a/include/ttmlir/Dialect/TTIR/IR/TTIRBase.td b/include/ttmlir/Dialect/TTIR/IR/TTIRBase.td
index 57f3dc37d3..b71541b8dc 100644
--- a/include/ttmlir/Dialect/TTIR/IR/TTIRBase.td
+++ b/include/ttmlir/Dialect/TTIR/IR/TTIRBase.td
@@ -6,6 +6,7 @@
 #define TTMLIR_TTMLIR_DIALECT_TTIR_TTIRDIALECT_TD
 
 include "mlir/IR/OpBase.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
 
 //===----------------------------------------------------------------------===//
 // TTIR dialect definition.
@@ -38,6 +39,6 @@ def TTIR_Dialect : Dialect {
 //===----------------------------------------------------------------------===//
 
 class TTIR_Op<string mnemonic, list<Trait> traits = []> :
-        Op<TTIR_Dialect, mnemonic, traits>;
+        Op<TTIR_Dialect, mnemonic, !listconcat(traits, [Pure])>;
 
 #endif
diff --git a/include/ttmlir/Dialect/TTNN/IR/TTNNBase.td b/include/ttmlir/Dialect/TTNN/IR/TTNNBase.td
index 34d3daf9cc..f6f764d01a 100644
--- a/include/ttmlir/Dialect/TTNN/IR/TTNNBase.td
+++ b/include/ttmlir/Dialect/TTNN/IR/TTNNBase.td
@@ -6,6 +6,7 @@
 #define TTMLIR_TTMLIR_DIALECT_TTNN_TTNNDIALECT_TD
 
 include "mlir/IR/OpBase.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
 include "ttmlir/Dialect/TTNN/IR/TTNNOpModelInterface.td"
 include "ttmlir/Dialect/TTNN/IR/TTNNWorkaroundInterface.td"
 
@@ -45,6 +46,6 @@ def TTNN_Dialect : Dialect {
 //===----------------------------------------------------------------------===//
 
 class TTNN_Op<string mnemonic, list<Trait> traits = []> :
-        Op<TTNN_Dialect, mnemonic, !listconcat(traits, [TTNN_OpModelInterface, TTNN_WorkaroundInterface])>;
+        Op<TTNN_Dialect, mnemonic, !listconcat(traits, [Pure, TTNN_OpModelInterface, TTNN_WorkaroundInterface])>;
 
 #endif
diff --git a/include/ttmlir/Dialect/TTNN/IR/TTNNOps.td b/include/ttmlir/Dialect/TTNN/IR/TTNNOps.td
index d918691611..41eb9dff75 100644
--- a/include/ttmlir/Dialect/TTNN/IR/TTNNOps.td
+++ b/include/ttmlir/Dialect/TTNN/IR/TTNNOps.td
@@ -794,9 +794,7 @@ def TTNN_ClampOp : TTNN_Op<"clamp"> {
     let hasVerifier = 1;
 }
 
-// Note: NoMemoryEffect is used to indicate that operation can be removed if it is not used.
-// Removal of this operation is done by the dead code elimination pass (RemoveDeadValuesPass).
-def TTNN_EmptyOp : TTNN_Op<"empty", [NoMemoryEffect]> {
+def TTNN_EmptyOp : TTNN_Op<"empty"> {
     let summary = "Empty op.";
     let description = [{
       Tensor empty operation
diff --git a/test/ttmlir/Dialect/TTIR/test_remove_dead_values_pass.mlir b/test/ttmlir/Dialect/TTIR/test_remove_dead_values_pass.mlir
new file mode 100644
index 0000000000..8b6df4d0f2
--- /dev/null
+++ b/test/ttmlir/Dialect/TTIR/test_remove_dead_values_pass.mlir
@@ -0,0 +1,22 @@
+// RUN: ttmlir-opt --remove-dead-values %s | FileCheck %s
+#any_device = #tt.operand_constraint<dram|l1|scalar|tile|any_device|any_device_tile>
+module attributes {} {
+  func.func @forward(%arg0: tensor<64x128xf32>, %arg1: tensor<64x128xf32>) -> tensor<64x128xf32> {
+    %0 = tensor.empty() : tensor<64x128xf32>
+    // CHECK: %[[C:.*]] = "ttir.multiply"[[C:.*]]
+    %1 = "ttir.multiply"(%arg0, %arg1, %0) <{operandSegmentSizes = array<i32: 2, 1>, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<64x128xf32>, tensor<64x128xf32>, tensor<64x128xf32>) -> tensor<64x128xf32>
+    %2 = tensor.empty() : tensor<64x128xf32>
+    // CHECK-NOT: %[[C:.*]] = "ttir.add"[[C:.*]]
+    %3 = "ttir.add"(%arg0, %arg1, %2) <{operandSegmentSizes = array<i32: 2, 1>, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<64x128xf32>, tensor<64x128xf32>, tensor<64x128xf32>) -> tensor<64x128xf32>
+    %4 = tensor.empty() : tensor<64x128xf32>
+    // CHECK-NOT: %[[C:.*]] = "ttir.subtract"[[C:.*]]
+    %5 = "ttir.subtract"(%arg0, %arg1, %4) <{operandSegmentSizes = array<i32: 2, 1>, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<64x128xf32>, tensor<64x128xf32>, tensor<64x128xf32>) -> tensor<64x128xf32>
+    %6 = tensor.empty() : tensor<64x128xf32>
+    // CHECK-NOT: %[[C:.*]] = "ttir.div"[[C:.*]]
+    %7 = "ttir.div"(%arg0, %arg1, %6) <{operandSegmentSizes = array<i32: 2, 1>, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<64x128xf32>, tensor<64x128xf32>, tensor<64x128xf32>) -> tensor<64x128xf32>
+    %8 = tensor.empty() : tensor<64x128xf32>
+    // CHECK-NOT: %[[C:.*]] = "ttir.eq"[[C:.*]]
+    %9 = "ttir.eq"(%arg0, %arg1, %8) <{operandSegmentSizes = array<i32: 2, 1>, operand_constraints = [#any_device, #any_device, #any_device]}> : (tensor<64x128xf32>, tensor<64x128xf32>, tensor<64x128xf32>) -> tensor<64x128xf32>
+    return %1 : tensor<64x128xf32>
+  }
+}
diff --git a/test/ttmlir/Dialect/TTNN/test_remove_dead_values_pass.mlir b/test/ttmlir/Dialect/TTNN/test_remove_dead_values_pass.mlir
new file mode 100644
index 0000000000..f3231730f5
--- /dev/null
+++ b/test/ttmlir/Dialect/TTNN/test_remove_dead_values_pass.mlir
@@ -0,0 +1,77 @@
+// RUN: ttmlir-opt --remove-dead-values %s | FileCheck %s
+#device = #tt.device<workerGrid = #tt.grid<8x8, (d0, d1) -> (0, d0, d1)>, l1Map = (d0, d1)[s0, s1] -> (0, d0 floordiv s0, d1 floordiv s1, (d0 mod s0) * s1 + d1 mod s1), dramMap = (d0, d1)[s0, s1] -> (0, 0, ((((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) floordiv 8192) mod 12, (((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) floordiv 98304 + (((d0 floordiv s0) * 8 + d1 floordiv s1) * (s1 * s0) + (d0 mod s0) * s1 + d1 mod s1) mod 8192), meshShape = , chipIds = [0]>
+#dram = #ttnn.buffer_type<dram>
+#system_desc = #tt.system_desc<[{role = host, target_triple = "x86_64-pc-linux-gnu"}], [{arch = <wormhole_b0>, grid = 8x8, l1_size = 1499136, num_dram_channels = 12, dram_channel_size = 1073741824, noc_l1_address_align_bytes = 16, pcie_address_align_bytes = 32, noc_dram_address_align_bytes = 32, l1_unreserved_base = 1024, erisc_l1_unreserved_base = 1024, dram_unreserved_base = 1024, dram_unreserved_end = 1073741824, physical_cores = {worker = [ 0x0,  0x1,  0x2,  0x3,  0x4,  0x5,  0x6,  0x7,  1x0,  1x1,  1x2,  1x3,  1x4,  1x5,  1x6,  1x7,  2x0,  2x1,  2x2,  2x3,  2x4,  2x5,  2x6,  2x7,  3x0,  3x1,  3x2,  3x3,  3x4,  3x5,  3x6,  3x7,  4x0,  4x1,  4x2,  4x3,  4x4,  4x5,  4x6,  4x7,  5x0,  5x1,  5x2,  5x3,  5x4,  5x5,  5x6,  5x7,  6x0,  6x1,  6x2,  6x3,  6x4,  6x5,  6x6,  6x7,  7x0,  7x1,  7x2,  7x3,  7x4,  7x5,  7x6,  7x7] dram = [ 8x0,  9x0,  10x0,  8x1,  9x1,  10x1,  8x2,  9x2,  10x2,  8x3,  9x3,  10x3]}, supported_data_types = [<f32>, <f16>, <bf16>, <bfp_f8>, <bfp_bf8>, <bfp_f4>, <bfp_bf4>, <bfp_f2>, <bfp_bf2>, <u32>, <u16>, <u8>], supported_tile_sizes = [ 4x16,  16x16,  32x16,  4x32,  16x32,  32x32], num_cbs = 32}], [0], [3 : i32], [ 0x0x0x0]>
+#system_memory = #ttnn.buffer_type<system_memory>
+#ttnn_layout = #ttnn.ttnn_layout<(d0, d1) -> (d0, d1), <1x1>, memref<64x128xf32, #system_memory>>
+#ttnn_layout1 = #ttnn.ttnn_layout<(d0, d1) -> (d0, d1), <1x1>, memref<2x4x!tt.tile<32x32, f32>, #dram>, interleaved>
+#ttnn_layout2 = #ttnn.ttnn_layout<(d0, d1) -> (d0, d1), <1x1>, memref<64x128xf32, #dram>, interleaved>
+module attributes {tt.device = #device, tt.system_desc = #system_desc} {
+  func.func @forward(%arg0: tensor<64x128xf32, #ttnn_layout>, %arg1: tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout> {
+    %0 = "ttnn.get_device"() <{mesh_shape = #ttnn<mesh_shape 1x1>}> : () -> !tt.device<#device>
+    %1 = "ttnn.to_layout"(%arg0) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %2 = "ttnn.to_device"(%1, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%1) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %3 = "ttnn.to_layout"(%arg1) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %4 = "ttnn.to_device"(%3, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%3) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %5 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes<f32>, layout = #ttnn.layout<row_major>, memory_config = #ttnn.memory_config<<interleaved>, #dram, <<64x128>>>, shape = #ttnn.shape<64x128>}> : (!tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout2>
+    // CHECK: %[[C:.*]] = "ttnn.multiply"[[C:.*]]
+    %6 = "ttnn.multiply"(%2, %4, %5) <{operandSegmentSizes = array<i32: 2, 1>}> : (tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout2>
+    "ttnn.deallocate"(%4) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    "ttnn.deallocate"(%2) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %7 = "ttnn.to_layout"(%arg0) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %8 = "ttnn.to_device"(%7, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%7) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %9 = "ttnn.to_layout"(%arg1) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %10 = "ttnn.to_device"(%9, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%9) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %11 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes<f32>, layout = #ttnn.layout<row_major>, memory_config = #ttnn.memory_config<<interleaved>, #dram, <<64x128>>>, shape = #ttnn.shape<64x128>}> : (!tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout2>
+    // CHECK-NOT: %[[C:.*]] = "ttnn.add"[[C:.*]]
+    %12 = "ttnn.add"(%8, %10, %11) <{operandSegmentSizes = array<i32: 2, 1>}> : (tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout2>
+    "ttnn.deallocate"(%11) <{force = false}> : (tensor<64x128xf32, #ttnn_layout2>) -> ()
+    "ttnn.deallocate"(%10) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    "ttnn.deallocate"(%8) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %13 = "ttnn.to_layout"(%arg0) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %14 = "ttnn.to_device"(%13, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%13) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %15 = "ttnn.to_layout"(%arg1) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %16 = "ttnn.to_device"(%15, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%15) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %17 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes<f32>, layout = #ttnn.layout<row_major>, memory_config = #ttnn.memory_config<<interleaved>, #dram, <<64x128>>>, shape = #ttnn.shape<64x128>}> : (!tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout2>
+    // CHECK-NOT: %[[C:.*]] = "ttnn.subtract"[[C:.*]]
+    %18 = "ttnn.subtract"(%14, %16, %17) <{operandSegmentSizes = array<i32: 2, 1>}> : (tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout2>
+    "ttnn.deallocate"(%17) <{force = false}> : (tensor<64x128xf32, #ttnn_layout2>) -> ()
+    "ttnn.deallocate"(%16) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    "ttnn.deallocate"(%14) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %19 = "ttnn.to_layout"(%arg0) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %20 = "ttnn.to_device"(%19, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%19) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %21 = "ttnn.to_layout"(%arg1) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %22 = "ttnn.to_device"(%21, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%21) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %23 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes<f32>, layout = #ttnn.layout<row_major>, memory_config = #ttnn.memory_config<<interleaved>, #dram, <<64x128>>>, shape = #ttnn.shape<64x128>}> : (!tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout2>
+    // CHECK-NOT: %[[C:.*]] = "ttnn.div"[[C:.*]]
+    %24 = "ttnn.div"(%20, %22, %23) <{operandSegmentSizes = array<i32: 2, 1>}> : (tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout2>
+    "ttnn.deallocate"(%23) <{force = false}> : (tensor<64x128xf32, #ttnn_layout2>) -> ()
+    "ttnn.deallocate"(%22) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    "ttnn.deallocate"(%20) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %25 = "ttnn.to_layout"(%arg0) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %26 = "ttnn.to_device"(%25, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%25) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %27 = "ttnn.to_layout"(%arg1) <{layout = #ttnn.layout<tile>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout1>
+    %28 = "ttnn.to_device"(%27, %0) <{memory_config = #ttnn.memory_config<<interleaved>, #dram, <<2x4>>>}> : (tensor<64x128xf32, #ttnn_layout1>, !tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout1>
+    "ttnn.deallocate"(%27) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %29 = "ttnn.empty"(%0) <{dtype = #tt.supportedDataTypes<f32>, layout = #ttnn.layout<row_major>, memory_config = #ttnn.memory_config<<interleaved>, #dram, <<64x128>>>, shape = #ttnn.shape<64x128>}> : (!tt.device<#device>) -> tensor<64x128xf32, #ttnn_layout2>
+    // CHECK-NOT: %[[C:.*]] = "ttnn.eq"[[C:.*]]
+    %30 = "ttnn.eq"(%26, %28, %29) <{operandSegmentSizes = array<i32: 2, 1>}> : (tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout1>, tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout2>
+    "ttnn.deallocate"(%29) <{force = false}> : (tensor<64x128xf32, #ttnn_layout2>) -> ()
+    "ttnn.deallocate"(%28) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    "ttnn.deallocate"(%26) <{force = false}> : (tensor<64x128xf32, #ttnn_layout1>) -> ()
+    %31 = "ttnn.from_device"(%6) : (tensor<64x128xf32, #ttnn_layout2>) -> tensor<64x128xf32, #ttnn_layout>
+    "ttnn.deallocate"(%5) <{force = false}> : (tensor<64x128xf32, #ttnn_layout2>) -> ()
+    %32 = "ttnn.to_layout"(%31) <{layout = #ttnn.layout<row_major>}> : (tensor<64x128xf32, #ttnn_layout>) -> tensor<64x128xf32, #ttnn_layout>
+    "ttnn.deallocate"(%31) <{force = false}> : (tensor<64x128xf32, #ttnn_layout>) -> ()
+    return %32 : tensor<64x128xf32, #ttnn_layout>
+  }
+}