diff --git a/src/backend/PipelineLayout.cpp b/src/backend/PipelineLayout.cpp
index d25c2acce0..e47855791c 100644
--- a/src/backend/PipelineLayout.cpp
+++ b/src/backend/PipelineLayout.cpp
@@ -34,7 +34,7 @@ namespace backend {
 
     // PipelineLayoutBase
 
-    PipelineLayoutBase::PipelineLayoutBase(DeviceBase* device,
+    PipelineLayoutBase::PipelineLayoutBase(DeviceBase*,
                                            const nxt::PipelineLayoutDescriptor* descriptor) {
         ASSERT(descriptor->numBindGroupLayouts <= kMaxBindGroups);
         for (uint32_t group = 0; group < descriptor->numBindGroupLayouts; ++group) {
@@ -42,15 +42,6 @@ namespace backend {
                 reinterpret_cast<BindGroupLayoutBase*>(descriptor->bindGroupLayouts[group].Get());
             mMask.set(group);
         }
-        // TODO(kainino@chromium.org): It shouldn't be necessary to construct default bind
-        // group layouts here. Remove these and fix things so that they are not needed.
-        for (uint32_t group = descriptor->numBindGroupLayouts; group < kMaxBindGroups; ++group) {
-            auto builder = device->CreateBindGroupLayoutBuilder();
-            mBindGroupLayouts[group] = builder->GetResult();
-            // Remove the external ref objects are created with
-            mBindGroupLayouts[group]->Release();
-            builder->Release();
-        }
     }
 
     const BindGroupLayoutBase* PipelineLayoutBase::GetBindGroupLayout(size_t group) const {
diff --git a/src/backend/PipelineLayout.h b/src/backend/PipelineLayout.h
index 262dafb502..e11629614e 100644
--- a/src/backend/PipelineLayout.h
+++ b/src/backend/PipelineLayout.h
@@ -37,6 +37,7 @@ namespace backend {
         PipelineLayoutBase(DeviceBase* device, const nxt::PipelineLayoutDescriptor* descriptor);
 
         const BindGroupLayoutBase* GetBindGroupLayout(size_t group) const;
+        // XXX: rename to GetBindGroupLayoutsMask?
         const std::bitset<kMaxBindGroups> GetBindGroupsLayoutMask() const;
 
         // Utility functions to compute inherited bind groups.
diff --git a/src/backend/d3d12/PipelineLayoutD3D12.cpp b/src/backend/d3d12/PipelineLayoutD3D12.cpp
index 2b6674f972..3ee78ed2b0 100644
--- a/src/backend/d3d12/PipelineLayoutD3D12.cpp
+++ b/src/backend/d3d12/PipelineLayoutD3D12.cpp
@@ -17,6 +17,7 @@
 #include "backend/d3d12/BindGroupLayoutD3D12.h"
 #include "backend/d3d12/DeviceD3D12.h"
 #include "common/Assert.h"
+#include "common/BitSetIterator.h"
 
 using Microsoft::WRL::ComPtr;
 
@@ -42,7 +43,7 @@ namespace backend { namespace d3d12 {
         uint32_t parameterIndex = 0;
         uint32_t rangeIndex = 0;
 
-        for (uint32_t group = 0; group < kMaxBindGroups; ++group) {
+        for (uint32_t group : IterateBitSet(GetBindGroupsLayoutMask())) {
             const BindGroupLayout* bindGroupLayout = ToBackend(GetBindGroupLayout(group));
 
             // Set the root descriptor table parameter and copy ranges. Ranges are offset by the
diff --git a/src/backend/metal/PipelineLayoutMTL.mm b/src/backend/metal/PipelineLayoutMTL.mm
index 0a85d4f381..49be220ba2 100644
--- a/src/backend/metal/PipelineLayoutMTL.mm
+++ b/src/backend/metal/PipelineLayoutMTL.mm
@@ -16,6 +16,7 @@
 
 #include "backend/BindGroupLayout.h"
 #include "backend/metal/DeviceMTL.h"
+#include "common/BitSetIterator.h"
 
 namespace backend { namespace metal {
 
@@ -28,7 +29,7 @@ namespace backend { namespace metal {
             uint32_t samplerIndex = 0;
             uint32_t textureIndex = 0;
 
-            for (size_t group = 0; group < kMaxBindGroups; ++group) {
+            for (uint32_t group : IterateBitSet(GetBindGroupsLayoutMask())) {
                 const auto& groupInfo = GetBindGroupLayout(group)->GetBindingInfo();
                 for (size_t binding = 0; binding < kMaxBindingsPerGroup; ++binding) {
                     if (!(groupInfo.visibilities[binding] & StageBit(stage))) {
diff --git a/src/backend/opengl/PipelineGL.cpp b/src/backend/opengl/PipelineGL.cpp
index a71702b7dd..491e1b3521 100644
--- a/src/backend/opengl/PipelineGL.cpp
+++ b/src/backend/opengl/PipelineGL.cpp
@@ -19,6 +19,7 @@
 #include "backend/opengl/PersistentPipelineStateGL.h"
 #include "backend/opengl/PipelineLayoutGL.h"
 #include "backend/opengl/ShaderModuleGL.h"
+#include "common/BitSetIterator.h"
 
 #include <iostream>
 #include <set>
@@ -125,7 +126,7 @@ namespace backend { namespace opengl {
         const auto& layout = ToBackend(parent->GetLayout());
         const auto& indices = layout->GetBindingIndexInfo();
 
-        for (uint32_t group = 0; group < kMaxBindGroups; ++group) {
+        for (uint32_t group : IterateBitSet(layout->GetBindGroupsLayoutMask())) {
             const auto& groupInfo = layout->GetBindGroupLayout(group)->GetBindingInfo();
 
             for (uint32_t binding = 0; binding < kMaxBindingsPerGroup; ++binding) {
diff --git a/src/backend/opengl/PipelineLayoutGL.cpp b/src/backend/opengl/PipelineLayoutGL.cpp
index 44a08d89b0..77942d6690 100644
--- a/src/backend/opengl/PipelineLayoutGL.cpp
+++ b/src/backend/opengl/PipelineLayoutGL.cpp
@@ -16,6 +16,7 @@
 
 #include "backend/BindGroupLayout.h"
 #include "backend/opengl/DeviceGL.h"
+#include "common/BitSetIterator.h"
 
 namespace backend { namespace opengl {
 
@@ -26,7 +27,7 @@ namespace backend { namespace opengl {
         GLuint sampledTextureIndex = 0;
         GLuint ssboIndex = 0;
 
-        for (size_t group = 0; group < kMaxBindGroups; ++group) {
+        for (uint32_t group : IterateBitSet(GetBindGroupsLayoutMask())) {
             const auto& groupInfo = GetBindGroupLayout(group)->GetBindingInfo();
 
             for (size_t binding = 0; binding < kMaxBindingsPerGroup; ++binding) {