Google Git
Sign in
dawn / dawn / 26dc5f6989a35758ee466e8b2df5cebc983058ab^! / .
commit26dc5f6989a35758ee466e8b2df5cebc983058ab[log] [tgz]
authorDavid Neto <dneto@google.com>Tue Dec 26 19:04:03 2023 +0000
committerDawn LUCI CQ <dawn-scoped@luci-project-accounts.iam.gserviceaccount.com>Tue Dec 26 19:04:03 2023 +0000
treef9100caf68db7fb94f964caac54527247e1b05a5
parent9fbef511d79e33454bd1b1a5349b67f0035c160c [diff]
msl ast_writer: Prevent emitted loops from erasure during optimizations

emit __asm__(""); at the top of each loop body

Bug: tint:2125
Change-Id: I93a17f30dec08f3eb57443d5c6f5aeb3df975938
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/166860
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>

diff --git a/src/tint/lang/msl/writer/ast_printer/ast_printer.cc b/src/tint/lang/msl/writer/ast_printer/ast_printer.cc
index 8787d09..4f90706 100644
--- a/src/tint/lang/msl/writer/ast_printer/ast_printer.cc
+++ b/src/tint/lang/msl/writer/ast_printer/ast_printer.cc

@@ -2123,6 +2123,7 @@
 
     TINT_SCOPED_ASSIGNMENT(emit_continuing_, emit_continuing);
     Line() << "while (true) {";
+    EmitLoopPreserver();
     {
         ScopedIndent si(this);
         if (!EmitStatements(stmt->body->statements)) {
@@ -2193,6 +2194,7 @@
 
         TINT_SCOPED_ASSIGNMENT(emit_continuing_, emit_continuing);
         Line() << "while (true) {";
+        EmitLoopPreserver();
         IncrementIndent();
         TINT_DEFER({
             DecrementIndent();
@@ -2233,6 +2235,7 @@
             }
             out << " {";
         }
+        EmitLoopPreserver();
         {
             auto emit_continuing = [] { return true; };
             TINT_SCOPED_ASSIGNMENT(emit_continuing_, emit_continuing);
@@ -2266,6 +2269,7 @@
     bool emit_as_loop = cond_pre.lines.size() > 0;
     if (emit_as_loop) {
         Line() << "while (true) {";
+        EmitLoopPreserver();
         IncrementIndent();
         TINT_DEFER({
             DecrementIndent();
@@ -2288,6 +2292,7 @@
             }
             out << " {";
         }
+        EmitLoopPreserver();
         if (!EmitStatementsWithIndent(stmt->body->statements)) {
             return false;
         }
@@ -3026,6 +3031,14 @@
     return true;
 }
 
+void ASTPrinter::EmitLoopPreserver() {
+    IncrementIndent();
+    // This statement prevents the MSL compiler from erasing a loop during
+    // optimimzations.
+    Line() << R"(__asm__("");)";
+    DecrementIndent();
+}
+
 template <typename F>
 bool ASTPrinter::CallBuiltinHelper(StringStream& out,
                                    const ast::CallExpression* call,

diff --git a/src/tint/lang/msl/writer/ast_printer/ast_printer.h b/src/tint/lang/msl/writer/ast_printer/ast_printer.h
index 9c9ba7c..af251db 100644
--- a/src/tint/lang/msl/writer/ast_printer/ast_printer.h
+++ b/src/tint/lang/msl/writer/ast_printer/ast_printer.h

@@ -378,6 +378,10 @@
                               const ast::CallExpression* expr,
                               const sem::BuiltinFn* builtin);
 
+    /// Emits a code sequence that preserves a loop during
+    /// optimizations even if the loop is infinite.
+    void EmitLoopPreserver();
+
     /// Handles generating a builtin name
     /// @param builtin the semantic info for the builtin
     /// @returns the name or "" if not valid

diff --git a/src/tint/lang/msl/writer/ast_printer/ast_printer_test.cc b/src/tint/lang/msl/writer/ast_printer/ast_printer_test.cc
index b62df62..f7b869e 100644
--- a/src/tint/lang/msl/writer/ast_printer/ast_printer_test.cc
+++ b/src/tint/lang/msl/writer/ast_printer/ast_printer_test.cc

@@ -232,6 +232,7 @@
 
 void comp_main_inner(uint local_invocation_index, threadgroup tint_array<float2x2, 4>* const tint_symbol) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float2x2(float2(0.0f), float2(0.0f));
   }

diff --git a/src/tint/lang/msl/writer/ast_printer/continue_test.cc b/src/tint/lang/msl/writer/ast_printer/continue_test.cc
index 508c69d..a79ddc3 100644
--- a/src/tint/lang/msl/writer/ast_printer/continue_test.cc
+++ b/src/tint/lang/msl/writer/ast_printer/continue_test.cc

@@ -43,6 +43,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(loop)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     if (false) {
       break;
     }

diff --git a/src/tint/lang/msl/writer/ast_printer/loop_test.cc b/src/tint/lang/msl/writer/ast_printer/loop_test.cc
index f4a6662..450592b 100644
--- a/src/tint/lang/msl/writer/ast_printer/loop_test.cc
+++ b/src/tint/lang/msl/writer/ast_printer/loop_test.cc

@@ -50,6 +50,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(l)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     break;
   }
 )");
@@ -70,6 +71,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(l)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     break;
     {
       a_statement();
@@ -93,6 +95,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(l)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     break;
     {
       a_statement();
@@ -126,7 +129,9 @@
 
     ASSERT_TRUE(gen.EmitStatement(outer)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     while (true) {
+      __asm__("");
       break;
       {
         a_statement();
@@ -166,6 +171,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(outer)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     float lhs = 2.5f;
     float other = 0.0f;
     break;
@@ -191,6 +197,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  for(; ; ) {
+    __asm__("");
     return;
   }
 )");
@@ -211,6 +218,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  for(int i = 0; ; ) {
+    __asm__("");
     return;
   }
 )");
@@ -244,6 +252,7 @@
       f(2);
     }
     for(; ; ) {
+      __asm__("");
       return;
     }
   }
@@ -265,6 +274,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  for(; true; ) {
+    __asm__("");
     return;
   }
 )");
@@ -287,6 +297,7 @@
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(),
               R"(  for(; ; i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     return;
   }
 )");
@@ -315,6 +326,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(loop)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while (true) {
+    __asm__("");
     return;
     {
       f(1);
@@ -342,6 +354,7 @@
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(),
               R"(  for(int i = 0; true; i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     a_statement();
   }
 )");
@@ -376,6 +389,7 @@
       f(2);
     }
     while (true) {
+      __asm__("");
       if (!(true)) { break; }
       return;
       {
@@ -401,6 +415,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while(true) {
+    __asm__("");
     return;
   }
 )");
@@ -420,6 +435,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while(true) {
+    __asm__("");
     continue;
   }
 )");
@@ -442,6 +458,7 @@
 
     ASSERT_TRUE(gen.EmitStatement(f)) << gen.Diagnostics();
     EXPECT_EQ(gen.Result(), R"(  while((t && false)) {
+    __asm__("");
     return;
   }
 )");

diff --git a/test/tint/array/assign_to_function_var.wgsl.expected.msl b/test/tint/array/assign_to_function_var.wgsl.expected.msl
index 7c0cd4c..6f9b5bb 100644
--- a/test/tint/array/assign_to_function_var.wgsl.expected.msl
+++ b/test/tint/array/assign_to_function_var.wgsl.expected.msl

@@ -55,6 +55,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<int4, 4>* const tint_symbol_8, const constant S* const tint_symbol_9, device S* const tint_symbol_10) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_8))[i] = int4(0);
   }

diff --git a/test/tint/array/assign_to_private_var.wgsl.expected.msl b/test/tint/array/assign_to_private_var.wgsl.expected.msl
index bfe306f..4046b9e 100644
--- a/test/tint/array/assign_to_private_var.wgsl.expected.msl
+++ b/test/tint/array/assign_to_private_var.wgsl.expected.msl

@@ -55,6 +55,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<int4, 4>* const tint_symbol_8, const constant S* const tint_symbol_9, device S* const tint_symbol_10) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_8))[i] = int4(0);
   }

diff --git a/test/tint/array/assign_to_storage_var.wgsl.expected.msl b/test/tint/array/assign_to_storage_var.wgsl.expected.msl
index b5d6233..94abf80 100644
--- a/test/tint/array/assign_to_storage_var.wgsl.expected.msl
+++ b/test/tint/array/assign_to_storage_var.wgsl.expected.msl

@@ -57,6 +57,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<int4, 4>* const tint_symbol_10, device S* const tint_symbol_11, const constant S* const tint_symbol_12, device S* const tint_symbol_13, device S_nested* const tint_symbol_14) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_10))[i] = int4(0);
   }

diff --git a/test/tint/array/assign_to_workgroup_var.wgsl.expected.msl b/test/tint/array/assign_to_workgroup_var.wgsl.expected.msl
index 3ad0e3b..d644a06 100644
--- a/test/tint/array/assign_to_workgroup_var.wgsl.expected.msl
+++ b/test/tint/array/assign_to_workgroup_var.wgsl.expected.msl

@@ -53,11 +53,13 @@
 
 void tint_symbol_inner(uint local_invocation_index, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<int4, 4>* const tint_symbol_10, threadgroup tint_array<int4, 4>* const tint_symbol_11, threadgroup tint_array<tint_array<tint_array<int, 2>, 3>, 4>* const tint_symbol_12, const constant S* const tint_symbol_13, device S* const tint_symbol_14) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_10))[i] = int4(0);
     (*(tint_symbol_11))[i] = int4(0);
   }
   for(uint idx_1 = local_invocation_index; (idx_1 < 24u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i_1 = (idx_1 / 6u);
     uint const i_2 = ((idx_1 % 6u) / 2u);
     uint const i_3 = (idx_1 % 2u);

diff --git a/test/tint/array/strides.spvasm.expected.msl b/test/tint/array/strides.spvasm.expected.msl
index 5de9554..38ff24a 100644
--- a/test/tint/array/strides.spvasm.expected.msl
+++ b/test/tint/array/strides.spvasm.expected.msl

@@ -34,12 +34,14 @@
 
 void assign_and_preserve_padding_3(device tint_array<strided_arr, 2>* const dest, tint_array<strided_arr, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_4(&((*(dest))[i]), value[i]);
   }
 }
 
 void assign_and_preserve_padding_2(device tint_array<tint_array<strided_arr, 2>, 3>* const dest, tint_array<tint_array<strided_arr, 2>, 3> value) {
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_3(&((*(dest))[i]), value[i]);
   }
 }
@@ -50,6 +52,7 @@
 
 void assign_and_preserve_padding(device tint_array<strided_arr_1, 4>* const dest, tint_array<strided_arr_1, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/storage/dynamic_index/read.wgsl.expected.msl b/test/tint/buffer/storage/dynamic_index/read.wgsl.expected.msl
index 641edaa..a545fd1 100644
--- a/test/tint/buffer/storage/dynamic_index/read.wgsl.expected.msl
+++ b/test/tint/buffer/storage/dynamic_index/read.wgsl.expected.msl

@@ -57,6 +57,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -65,6 +66,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -73,6 +75,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -81,6 +84,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/dynamic_index/read_f16.wgsl.expected.msl b/test/tint/buffer/storage/dynamic_index/read_f16.wgsl.expected.msl
index 714f1de..61b1043 100644
--- a/test/tint/buffer/storage/dynamic_index/read_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/dynamic_index/read_f16.wgsl.expected.msl

@@ -80,6 +80,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -88,6 +89,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -96,6 +98,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -104,6 +107,7 @@
 half2x3 tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -112,6 +116,7 @@
 half3x3 tint_unpack_vec3_in_composite_4(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -120,6 +125,7 @@
 half4x3 tint_unpack_vec3_in_composite_5(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -128,6 +134,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_6(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/dynamic_index/write.wgsl.expected.msl b/test/tint/buffer/storage/dynamic_index/write.wgsl.expected.msl
index c71462e..31962aa 100644
--- a/test/tint/buffer/storage/dynamic_index/write.wgsl.expected.msl
+++ b/test/tint/buffer/storage/dynamic_index/write.wgsl.expected.msl

@@ -103,6 +103,7 @@
 
 void assign_and_preserve_padding_3(device tint_array<tint_packed_vec3_f32_array_element, 2>* const dest, tint_array<float3, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     (*(dest))[i].elements = packed_float3(value[i]);
   }
 }

diff --git a/test/tint/buffer/storage/dynamic_index/write_f16.wgsl.expected.msl b/test/tint/buffer/storage/dynamic_index/write_f16.wgsl.expected.msl
index b5afec9..a984832 100644
--- a/test/tint/buffer/storage/dynamic_index/write_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/dynamic_index/write_f16.wgsl.expected.msl

@@ -158,6 +158,7 @@
 
 void assign_and_preserve_padding_6(device tint_array<tint_packed_vec3_f32_array_element, 2>* const dest, tint_array<float3, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     (*(dest))[i].elements = packed_float3(value[i]);
   }
 }

diff --git a/test/tint/buffer/storage/static_index/read.wgsl.expected.msl b/test/tint/buffer/storage/static_index/read.wgsl.expected.msl
index 6293a49..15c7f94 100644
--- a/test/tint/buffer/storage/static_index/read.wgsl.expected.msl
+++ b/test/tint/buffer/storage/static_index/read.wgsl.expected.msl

@@ -61,6 +61,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -69,6 +70,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -77,6 +79,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -85,6 +88,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/static_index/read_f16.wgsl.expected.msl b/test/tint/buffer/storage/static_index/read_f16.wgsl.expected.msl
index 66ae4dd..bbca88a 100644
--- a/test/tint/buffer/storage/static_index/read_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/static_index/read_f16.wgsl.expected.msl

@@ -86,6 +86,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -94,6 +95,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -102,6 +104,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -110,6 +113,7 @@
 half2x3 tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -118,6 +122,7 @@
 half3x3 tint_unpack_vec3_in_composite_4(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -126,6 +131,7 @@
 half4x3 tint_unpack_vec3_in_composite_5(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -134,6 +140,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_6(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/static_index/write.wgsl.expected.msl b/test/tint/buffer/storage/static_index/write.wgsl.expected.msl
index 7f0d7eb..93c261e 100644
--- a/test/tint/buffer/storage/static_index/write.wgsl.expected.msl
+++ b/test/tint/buffer/storage/static_index/write.wgsl.expected.msl

@@ -105,6 +105,7 @@
 
 void assign_and_preserve_padding_3(device tint_array<tint_packed_vec3_f32_array_element, 2>* const dest, tint_array<float3, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     (*(dest))[i].elements = packed_float3(value[i]);
   }
 }

diff --git a/test/tint/buffer/storage/static_index/write_f16.wgsl.expected.msl b/test/tint/buffer/storage/static_index/write_f16.wgsl.expected.msl
index 5ffea48..6141100 100644
--- a/test/tint/buffer/storage/static_index/write_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/static_index/write_f16.wgsl.expected.msl

@@ -162,6 +162,7 @@
 
 void assign_and_preserve_padding_6(device tint_array<tint_packed_vec3_f32_array_element, 2>* const dest, tint_array<float3, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     (*(dest))[i].elements = packed_float3(value[i]);
   }
 }
@@ -174,6 +175,7 @@
 
 void assign_and_preserve_padding_8(device tint_array<Inner, 4>* const dest, tint_array<Inner, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_7(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/storage/types/mat2x3_f16.wgsl.expected.msl b/test/tint/buffer/storage/types/mat2x3_f16.wgsl.expected.msl
index 2e9605e..d42c35e 100644
--- a/test/tint/buffer/storage/types/mat2x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat2x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/types/mat2x3_f32.wgsl.expected.msl b/test/tint/buffer/storage/types/mat2x3_f32.wgsl.expected.msl
index a03434b..326df90 100644
--- a/test/tint/buffer/storage/types/mat2x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat2x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/types/mat3x3_f16.wgsl.expected.msl b/test/tint/buffer/storage/types/mat3x3_f16.wgsl.expected.msl
index c688ab9..e5901a6 100644
--- a/test/tint/buffer/storage/types/mat3x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat3x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/types/mat3x3_f32.wgsl.expected.msl b/test/tint/buffer/storage/types/mat3x3_f32.wgsl.expected.msl
index 924c9d6..bb4912a 100644
--- a/test/tint/buffer/storage/types/mat3x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat3x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/types/mat4x3_f16.wgsl.expected.msl b/test/tint/buffer/storage/types/mat4x3_f16.wgsl.expected.msl
index 5ced5ff..602e00c 100644
--- a/test/tint/buffer/storage/types/mat4x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat4x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/storage/types/mat4x3_f32.wgsl.expected.msl b/test/tint/buffer/storage/types/mat4x3_f32.wgsl.expected.msl
index 78ba3ce..49f2d3d 100644
--- a/test/tint/buffer/storage/types/mat4x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/storage/types/mat4x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/dynamic_index/read.wgsl.expected.msl b/test/tint/buffer/uniform/dynamic_index/read.wgsl.expected.msl
index 63bacfe..a10666c 100644
--- a/test/tint/buffer/uniform/dynamic_index/read.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/dynamic_index/read.wgsl.expected.msl

@@ -57,6 +57,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -65,6 +66,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -73,6 +75,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -81,6 +84,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/dynamic_index/read_f16.wgsl.expected.msl b/test/tint/buffer/uniform/dynamic_index/read_f16.wgsl.expected.msl
index 9ccbb22..3a0ba94 100644
--- a/test/tint/buffer/uniform/dynamic_index/read_f16.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/dynamic_index/read_f16.wgsl.expected.msl

@@ -80,6 +80,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -88,6 +89,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -96,6 +98,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -104,6 +107,7 @@
 half2x3 tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -112,6 +116,7 @@
 half3x3 tint_unpack_vec3_in_composite_4(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -120,6 +125,7 @@
 half4x3 tint_unpack_vec3_in_composite_5(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -128,6 +134,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_6(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/static_index/read.wgsl.expected.msl b/test/tint/buffer/uniform/static_index/read.wgsl.expected.msl
index 047bafa..37f1ec2 100644
--- a/test/tint/buffer/uniform/static_index/read.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/static_index/read.wgsl.expected.msl

@@ -62,6 +62,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -70,6 +71,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -78,6 +80,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -86,6 +89,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/static_index/read_f16.wgsl.expected.msl b/test/tint/buffer/uniform/static_index/read_f16.wgsl.expected.msl
index 242e1f5..71a600e 100644
--- a/test/tint/buffer/uniform/static_index/read_f16.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/static_index/read_f16.wgsl.expected.msl

@@ -85,6 +85,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -93,6 +94,7 @@
 float3x3 tint_unpack_vec3_in_composite_1(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -101,6 +103,7 @@
 float4x3 tint_unpack_vec3_in_composite_2(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -109,6 +112,7 @@
 half2x3 tint_unpack_vec3_in_composite_3(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -117,6 +121,7 @@
 half3x3 tint_unpack_vec3_in_composite_4(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -125,6 +130,7 @@
 half4x3 tint_unpack_vec3_in_composite_5(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -133,6 +139,7 @@
 tint_array<float3, 2> tint_unpack_vec3_in_composite_6(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   tint_array<float3, 2> result = {};
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x2_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x2_f32/to_workgroup.wgsl.expected.msl
index ed5fb0c..f431584 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x2_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x2_f32/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float2x2, 4>* const tint_symbol, const constant tint_array<float2x2, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float2x2(float2(0.0f), float2(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 0589421..5d1772c 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
index aceac06..ac667d1 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_builtin.wgsl.expected.msl
index 83c7143..5eb17bf 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_fn.wgsl.expected.msl
index d2da737..7b82dad 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_private.wgsl.expected.msl
index b83c6e8..31b51ba 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_storage.wgsl.expected.msl
index afbab95..1a97187 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -42,6 +44,7 @@
 
 void assign_and_preserve_padding(device tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4>* const dest, tint_array<half2x3, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_workgroup.wgsl.expected.msl
index 2d1d8b1..f1fc6fa 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f16/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4> in) {
   tint_array<half2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -41,6 +43,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<half2x3, 4>* const tint_symbol, const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 2>, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = half2x3(half3(0.0h), half3(0.0h));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index 4a958fe..ad35522 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
index e3655fb..9cbc007 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_builtin.wgsl.expected.msl
index ac6a186..63beae2 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_fn.wgsl.expected.msl
index 471b2da..f6a7911 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_private.wgsl.expected.msl
index 1d8250e..1f73bde 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_storage.wgsl.expected.msl
index 6adc305..1e1c612 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -42,6 +44,7 @@
 
 void assign_and_preserve_padding(device tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4>* const dest, tint_array<float2x3, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_workgroup.wgsl.expected.msl
index 3ff0a56..87008b8 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float2x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4> in) {
   tint_array<float2x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -41,6 +43,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float2x3, 4>* const tint_symbol, const constant tint_array<tint_array<tint_packed_vec3_f32_array_element, 2>, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float2x3(float3(0.0f), float3(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x4_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x4_f16/to_workgroup.wgsl.expected.msl
index 01e1741..ec6ff10 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x4_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x4_f16/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<half2x4, 4>* const tint_symbol, const constant tint_array<half2x4, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = half2x4(half4(0.0h), half4(0.0h));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat2x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat2x4_f32/to_workgroup.wgsl.expected.msl
index 3ff1615..ab18a13 100644
--- a/test/tint/buffer/uniform/std140/array/mat2x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat2x4_f32/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float2x4, 4>* const tint_symbol, const constant tint_array<float2x4, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float2x4(float4(0.0f), float4(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index f90ac9e..b32555e 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
index 376cd23..43646f4 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_builtin.wgsl.expected.msl
index 1a1377b..0306b17 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_fn.wgsl.expected.msl
index 24d26c0..7061225 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_private.wgsl.expected.msl
index fcdd7d4..2c3063f 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_storage.wgsl.expected.msl
index e77dbf1..86fd953 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -43,6 +45,7 @@
 
 void assign_and_preserve_padding(device tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4>* const dest, tint_array<float3x3, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_workgroup.wgsl.expected.msl
index e1653cd..a917036 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float3x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4> in) {
   tint_array<float3x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -41,6 +43,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float3x3, 4>* const tint_symbol, const constant tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float3x3(float3(0.0f), float3(0.0f), float3(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat3x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat3x4_f32/to_workgroup.wgsl.expected.msl
index 7351a09..3a2f692 100644
--- a/test/tint/buffer/uniform/std140/array/mat3x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat3x4_f32/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float3x4, 4>* const tint_symbol, const constant tint_array<float3x4, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float3x4(float4(0.0f), float4(0.0f), float4(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x2_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x2_f16/to_workgroup.wgsl.expected.msl
index 0d29883..527f902 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x2_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x2_f16/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<half4x2, 4>* const tint_symbol, const constant tint_array<half4x2, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = half4x2(half2(0.0h), half2(0.0h), half2(0.0h), half2(0.0h));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x2_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x2_f32/to_workgroup.wgsl.expected.msl
index 4f1df19..e523c29 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x2_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x2_f32/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float4x2, 4>* const tint_symbol, const constant tint_array<float4x2, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float4x2(float2(0.0f), float2(0.0f), float2(0.0f), float2(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 6b79003..78cb8a1 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
index 6a4bf26..0e2a39e 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_builtin.wgsl.expected.msl
index d29c64f..8303fbc 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_fn.wgsl.expected.msl
index 2eb0517..cfa263e 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_private.wgsl.expected.msl
index dc505d8..6b6b343 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_storage.wgsl.expected.msl
index daef935..3d4a519 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -44,6 +46,7 @@
 
 void assign_and_preserve_padding(device tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* const dest, tint_array<half4x3, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_workgroup.wgsl.expected.msl
index 67f78b0..414b146 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f16/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<half4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4> in) {
   tint_array<half4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -41,6 +43,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<half4x3, 4>* const tint_symbol, const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = half4x3(half3(0.0h), half3(0.0h), half3(0.0h), half3(0.0h));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index 7b444ce..4e4afdb 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
index ade51a6..d5d1899 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_builtin.wgsl.expected.msl
index 83f80b7..3448203 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_fn.wgsl.expected.msl
index 471d98e..1596e48 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_private.wgsl.expected.msl
index 88eb93a..2f0ef79 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_storage.wgsl.expected.msl
index bdc44ad..661ffcb 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -30,6 +31,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -44,6 +46,7 @@
 
 void assign_and_preserve_padding(device tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4>* const dest, tint_array<float4x3, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_workgroup.wgsl.expected.msl
index e178ef7..c420ac2 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -34,6 +35,7 @@
 tint_array<float4x3, 4> tint_unpack_vec3_in_composite_1(tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4> in) {
   tint_array<float4x3, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite(in[i]);
   }
   return result;
@@ -41,6 +43,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float4x3, 4>* const tint_symbol, const constant tint_array<tint_array<tint_packed_vec3_f32_array_element, 4>, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float4x3(float3(0.0f), float3(0.0f), float3(0.0f), float3(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x4_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x4_f16/to_workgroup.wgsl.expected.msl
index e22baa8..79d94e5 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x4_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x4_f16/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<half4x4, 4>* const tint_symbol, const constant tint_array<half4x4, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = half4x4(half4(0.0h), half4(0.0h), half4(0.0h), half4(0.0h));
   }

diff --git a/test/tint/buffer/uniform/std140/array/mat4x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/array/mat4x4_f32/to_workgroup.wgsl.expected.msl
index eb1b0fe..78fac42 100644
--- a/test/tint/buffer/uniform/std140/array/mat4x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/array/mat4x4_f32/to_workgroup.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<float4x4, 4>* const tint_symbol, const constant tint_array<float4x4, 4>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i] = float4x4(float4(0.0f), float4(0.0f), float4(0.0f), float4(0.0f));
   }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_storage.wgsl.expected.msl
index 1cc33f8..5d62b61 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_storage.wgsl.expected.msl

@@ -30,6 +30,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_workgroup.wgsl.expected.msl
index 5b50787..fb0567d 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x2_f16/to_workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_storage.wgsl.expected.msl
index 8165cd0..e63293d 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_workgroup.wgsl.expected.msl
index f2303f8..d2a1def 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x2_f32/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index dcaae72..b8a19e0 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -35,6 +35,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
index 0e5d8a6..f40977b 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -49,6 +50,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -67,6 +69,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_builtin.wgsl.expected.msl
index e71cfaa..e1c5d9c 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_builtin.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_fn.wgsl.expected.msl
index 0fc03a4..1b2e965 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_fn.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_private.wgsl.expected.msl
index f58017c..37755f3 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_private.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_storage.wgsl.expected.msl
index fe2e439..c727e17 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_workgroup.wgsl.expected.msl
index 3235501..74f9824 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f16/to_workgroup.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -64,6 +66,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index cacb8ae..d4b482b 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -35,6 +35,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
index 526d744..55314d1 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -49,6 +50,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -67,6 +69,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_builtin.wgsl.expected.msl
index 5734e29..a68f229 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_builtin.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_fn.wgsl.expected.msl
index 05f8a07..3907fba 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_fn.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_private.wgsl.expected.msl
index 2b17191..089f859 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_private.wgsl.expected.msl

@@ -41,6 +41,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_storage.wgsl.expected.msl
index 40c8ff5..3fe251b 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_workgroup.wgsl.expected.msl
index bf3193b..447ce74 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x3_f32/to_workgroup.wgsl.expected.msl

@@ -41,6 +41,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -64,6 +66,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_storage.wgsl.expected.msl
index 719237a..28a9ca2 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_workgroup.wgsl.expected.msl
index 03151ea..7607f1a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x4_f16/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_storage.wgsl.expected.msl
index 52d9404..7aac521 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_workgroup.wgsl.expected.msl
index 5ddb92f..e93e10a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat2x4_f32/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_storage.wgsl.expected.msl
index 369c335..312ef3a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_storage.wgsl.expected.msl

@@ -30,6 +30,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_workgroup.wgsl.expected.msl
index 3b789e9..62fba42 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x2_f16/to_workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_storage.wgsl.expected.msl
index 3b94d66..cc8345c 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_workgroup.wgsl.expected.msl
index 210510c..fd7fd71 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x2_f32/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 984567f..34e7e96 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -35,6 +35,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl
index a707c60..1e7ec6a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -49,6 +50,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -67,6 +69,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_builtin.wgsl.expected.msl
index b1bbaf9..d245d11 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_builtin.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_fn.wgsl.expected.msl
index 3520432..f16b13b 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_fn.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_private.wgsl.expected.msl
index 2e942f5..dcc3d06 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_private.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_storage.wgsl.expected.msl
index f9022fc..43ba4a6 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -72,6 +74,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_workgroup.wgsl.expected.msl
index 454e443..92099fe 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f16/to_workgroup.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -64,6 +66,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index 29361e5..bce9b6d 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -35,6 +35,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
index 4b0a579..e991c88 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -49,6 +50,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -67,6 +69,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_builtin.wgsl.expected.msl
index 3fdbccc..4bb8278 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_builtin.wgsl.expected.msl

@@ -30,6 +30,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_fn.wgsl.expected.msl
index 82b8bfc..c0ffebc 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_fn.wgsl.expected.msl

@@ -30,6 +30,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -52,6 +53,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_private.wgsl.expected.msl
index 166da04..647d181 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_private.wgsl.expected.msl

@@ -40,6 +40,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -56,6 +57,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_storage.wgsl.expected.msl
index 27d1919..634ffb0 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_storage.wgsl.expected.msl

@@ -30,6 +30,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -52,6 +53,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_workgroup.wgsl.expected.msl
index 4987aac..c9f0553 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x3_f32/to_workgroup.wgsl.expected.msl

@@ -40,6 +40,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -56,6 +57,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -63,6 +65,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_storage.wgsl.expected.msl
index c9bca39..ce88e3e 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_workgroup.wgsl.expected.msl
index 065dd48..c9e815c 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x4_f16/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_storage.wgsl.expected.msl
index e349da5..663a39a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_storage.wgsl.expected.msl

@@ -30,6 +30,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_workgroup.wgsl.expected.msl
index 421719c..4ebd2e0 100644
--- a/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat3x4_f32/to_workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_storage.wgsl.expected.msl
index b80bfe1..4838775 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_storage.wgsl.expected.msl

@@ -30,6 +30,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_workgroup.wgsl.expected.msl
index 8c98cfc..ad88575 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x2_f16/to_workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_storage.wgsl.expected.msl
index 16ec5d1..e71864b 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_workgroup.wgsl.expected.msl
index 5931951..37b7b89 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x2_f32/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 7cabe51..20857b0 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -35,6 +35,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -71,6 +73,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
index 2c2086d..c0fcb1e 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -49,6 +50,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -67,6 +69,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_builtin.wgsl.expected.msl
index eea26e8..22b576a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_builtin.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_fn.wgsl.expected.msl
index 8ff5b7c..83803d9 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_fn.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_private.wgsl.expected.msl
index a1ac5a2..33ab4b6 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_private.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_storage.wgsl.expected.msl
index 6f84630..f93db23 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -73,6 +75,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_workgroup.wgsl.expected.msl
index c460177..446962e 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f16/to_workgroup.wgsl.expected.msl

@@ -41,6 +41,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -64,6 +66,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index e7bc473..0a1a62d 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -34,6 +34,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -52,6 +53,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -70,6 +72,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
index 479a398..c75a238 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -30,6 +30,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -48,6 +49,7 @@
 tint_array<Inner, 4> tint_unpack_vec3_in_composite_2(tint_array<Inner_tint_packed_vec3, 4> in) {
   tint_array<Inner, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -66,6 +68,7 @@
 tint_array<Outer, 4> tint_unpack_vec3_in_composite_4(tint_array<Outer_tint_packed_vec3, 4> in) {
   tint_array<Outer, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_3(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_builtin.wgsl.expected.msl
index ddd7ad1..7223575 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_builtin.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_fn.wgsl.expected.msl
index e8c56ff..8da0a09 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_fn.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_private.wgsl.expected.msl
index c57a58d..17a7d1a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_private.wgsl.expected.msl

@@ -41,6 +41,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_storage.wgsl.expected.msl
index 37070d6..4f9187a 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -53,6 +54,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -73,6 +75,7 @@
 
 void assign_and_preserve_padding(device tint_array<S_tint_packed_vec3, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_workgroup.wgsl.expected.msl
index c86e1dd..d719c22 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x3_f32/to_workgroup.wgsl.expected.msl

@@ -41,6 +41,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;
@@ -57,6 +58,7 @@
 tint_array<S, 4> tint_unpack_vec3_in_composite_2(tint_array<S_tint_packed_vec3, 4> in) {
   tint_array<S, 4> result = {};
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = tint_unpack_vec3_in_composite_1(in[i]);
   }
   return result;
@@ -64,6 +66,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S_tint_packed_vec3, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_storage.wgsl.expected.msl
index 4d16bf6..8223731 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_workgroup.wgsl.expected.msl
index 0b6c9ea..d43d742 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x4_f16/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_storage.wgsl.expected.msl
index bc5986c..54e83266 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_storage.wgsl.expected.msl

@@ -31,6 +31,7 @@
 
 void assign_and_preserve_padding(device tint_array<S, 4>* const dest, tint_array<S, 4> value) {
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_workgroup.wgsl.expected.msl
index 832baa2..efb661e 100644
--- a/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/struct/mat4x4_f32/to_workgroup.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup tint_array<S, 4>* const tint_symbol_1, const constant tint_array<S, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol = S{};
     (*(tint_symbol_1))[i] = tint_symbol;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 8d82142..66b9eae 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
index 6b09268..351639f 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_builtin.wgsl.expected.msl
index bc97605..6e97983 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_fn.wgsl.expected.msl
index f0ce93c..4eeec0f 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_private.wgsl.expected.msl
index f08c837..fcce5d9 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_storage.wgsl.expected.msl
index bcd5c22..ffce46c 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_workgroup.wgsl.expected.msl
index f501690..a1204b8 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f16/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index f79edc4..9dd941f 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
index 69dc89f..209a0d9 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_builtin.wgsl.expected.msl
index 997bc9c..b5246cb 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_fn.wgsl.expected.msl
index 35a013f..4fa9980 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_private.wgsl.expected.msl
index 76a7694..9a67d9e 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_storage.wgsl.expected.msl
index 28c4697..d1733cd 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_workgroup.wgsl.expected.msl
index 87a4247..5504f34 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat2x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 76a8717..818c5ed 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl
index 3eddae1..36cfd6c 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_builtin.wgsl.expected.msl
index 2416843..284120d 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_fn.wgsl.expected.msl
index 1c6b945..5183cdf 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_private.wgsl.expected.msl
index 197c3fe..df4f251 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_storage.wgsl.expected.msl
index a2c2764..c79089d 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_workgroup.wgsl.expected.msl
index b6b09a3..41a611d 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f16/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index e35f7f6..0d94d71 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
index 52ae088..50dc2cf 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_builtin.wgsl.expected.msl
index 01c6635..6879754 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_fn.wgsl.expected.msl
index ca1223e..7200e84 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_private.wgsl.expected.msl
index 9bbecf4..2fb02e7 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_storage.wgsl.expected.msl
index ebe399b..65a9461 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_workgroup.wgsl.expected.msl
index 8a82992..a84c6f9 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat3x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
index 27b22ff..2c20f74 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
index e324c0d..bc5f7d7 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_builtin.wgsl.expected.msl
index 4c19ba6..1661c25 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_fn.wgsl.expected.msl
index f8bf151..d3588d7 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_private.wgsl.expected.msl
index 9124188..c3a8656 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_storage.wgsl.expected.msl
index e94bb82..5b76f29 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_workgroup.wgsl.expected.msl
index b96b38e..45134fd 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f16/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
index 40b38c5..26cdbea 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/dynamic_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
index 90afeda..016dc82 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/static_index_via_ptr.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_builtin.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_builtin.wgsl.expected.msl
index 2fd10cb..e42d9f8 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_builtin.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_builtin.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_fn.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_fn.wgsl.expected.msl
index 6c3583e..f9fd46e 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_fn.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_fn.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_private.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_private.wgsl.expected.msl
index 2564332..f0cf1ae 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_private.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_private.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_storage.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_storage.wgsl.expected.msl
index d1bb402..39e493a 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_storage.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_storage.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_workgroup.wgsl.expected.msl b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_workgroup.wgsl.expected.msl
index 93bb82a..ebc4026 100644
--- a/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_workgroup.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/std140/unnested/mat4x3_f32/to_workgroup.wgsl.expected.msl

@@ -26,6 +26,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat2x3_f16.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat2x3_f16.wgsl.expected.msl
index 11535e4..66ab882 100644
--- a/test/tint/buffer/uniform/types/mat2x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat2x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 2> in) {
   half2x3 result = half2x3(0.0h);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat2x3_f32.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat2x3_f32.wgsl.expected.msl
index 1297b5b..fcb9344 100644
--- a/test/tint/buffer/uniform/types/mat2x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat2x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float2x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 2> in) {
   float2x3 result = float2x3(0.0f);
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat3x3_f16.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat3x3_f16.wgsl.expected.msl
index f155226..93b4fe7 100644
--- a/test/tint/buffer/uniform/types/mat3x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat3x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat3x3_f32.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat3x3_f32.wgsl.expected.msl
index 5a8898e..cf4c6b4 100644
--- a/test/tint/buffer/uniform/types/mat3x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat3x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat4x3_f16.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat4x3_f16.wgsl.expected.msl
index fb113d8..7dbe667 100644
--- a/test/tint/buffer/uniform/types/mat4x3_f16.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat4x3_f16.wgsl.expected.msl

@@ -22,6 +22,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/buffer/uniform/types/mat4x3_f32.wgsl.expected.msl b/test/tint/buffer/uniform/types/mat4x3_f32.wgsl.expected.msl
index c7184e4..57bc4c6 100644
--- a/test/tint/buffer/uniform/types/mat4x3_f32.wgsl.expected.msl
+++ b/test/tint/buffer/uniform/types/mat4x3_f32.wgsl.expected.msl

@@ -22,6 +22,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/bug/chromium/1403752.wgsl.expected.msl b/test/tint/bug/chromium/1403752.wgsl.expected.msl
index d427a04..4ae1c69 100644
--- a/test/tint/bug/chromium/1403752.wgsl.expected.msl
+++ b/test/tint/bug/chromium/1403752.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void d() {
   int j = 0;
   for(; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/bug/chromium/1449538.wgsl.expected.msl b/test/tint/bug/chromium/1449538.wgsl.expected.msl
index 81a4429..e80c545 100644
--- a/test/tint/bug/chromium/1449538.wgsl.expected.msl
+++ b/test/tint/bug/chromium/1449538.wgsl.expected.msl

@@ -3,20 +3,28 @@
 using namespace metal;
 void f() {
   for(int i0520 = 0; false; ) {
+    __asm__("");
   }
   for(int i62 = 0; false; ) {
+    __asm__("");
   }
   for(int i0520 = 0; false; ) {
+    __asm__("");
   }
   for(int i62 = 0; false; ) {
+    __asm__("");
   }
   for(int i62 = 0; false; ) {
+    __asm__("");
   }
   for(int i60 = 0; false; ) {
+    __asm__("");
   }
   for(int i62 = 0; false; ) {
+    __asm__("");
   }
   for(int i60 = 0; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/bug/fxc/dyn_array_idx/read/workgroup.wgsl.expected.msl b/test/tint/bug/fxc/dyn_array_idx/read/workgroup.wgsl.expected.msl
index b8557fd..28c36c5 100644
--- a/test/tint/bug/fxc/dyn_array_idx/read/workgroup.wgsl.expected.msl
+++ b/test/tint/bug/fxc/dyn_array_idx/read/workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup S* const tint_symbol, device Result* const tint_symbol_1, const constant UBO* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 64u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol)).data[i] = 0;
   }

diff --git a/test/tint/bug/fxc/dyn_array_idx/write/workgroup.wgsl.expected.msl b/test/tint/bug/fxc/dyn_array_idx/write/workgroup.wgsl.expected.msl
index fa5f687..8c88e24 100644
--- a/test/tint/bug/fxc/dyn_array_idx/write/workgroup.wgsl.expected.msl
+++ b/test/tint/bug/fxc/dyn_array_idx/write/workgroup.wgsl.expected.msl

@@ -28,6 +28,7 @@
 
 void f_inner(uint local_invocation_index, threadgroup S* const tint_symbol, const constant UBO* const tint_symbol_1, device Result* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 64u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol)).data[i] = 0;
   }

diff --git a/test/tint/bug/fxc/gradient_in_varying_loop/1112.wgsl.expected.msl b/test/tint/bug/fxc/gradient_in_varying_loop/1112.wgsl.expected.msl
index b831db0..3987e5d 100644
--- a/test/tint/bug/fxc/gradient_in_varying_loop/1112.wgsl.expected.msl
+++ b/test/tint/bug/fxc/gradient_in_varying_loop/1112.wgsl.expected.msl

@@ -13,6 +13,7 @@
   float3 const random = tint_symbol_4.sample(tint_symbol_5, vUV).rgb;
   int i = 0;
   while (true) {
+    __asm__("");
     if ((i < 1)) {
     } else {
       break;

diff --git a/test/tint/bug/fxc/indexed_assign_to_array_in_struct/1206.wgsl.expected.msl b/test/tint/bug/fxc/indexed_assign_to_array_in_struct/1206.wgsl.expected.msl
index b1e90cf..817bc40 100644
--- a/test/tint/bug/fxc/indexed_assign_to_array_in_struct/1206.wgsl.expected.msl
+++ b/test/tint/bug/fxc/indexed_assign_to_array_in_struct/1206.wgsl.expected.msl

@@ -35,6 +35,7 @@
 tint_array<float3, 8> tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 8> in) {
   tint_array<float3, 8> result = {};
   for(uint i = 0u; (i < 8u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_loop.wgsl.expected.msl b/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_loop.wgsl.expected.msl
index 91fd58a..e81bc5f 100644
--- a/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_loop.wgsl.expected.msl
+++ b/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_loop.wgsl.expected.msl

@@ -10,6 +10,7 @@
 
 void foo(thread tint_private_vars_struct* const tint_private_vars) {
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     (*(tint_private_vars)).v2f[i] = 1.0f;
     (*(tint_private_vars)).v3i[i] = 1;
     (*(tint_private_vars)).v4u[i] = 1u;
@@ -20,6 +21,7 @@
 kernel void tint_symbol() {
   thread tint_private_vars_struct tint_private_vars = {};
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     foo(&(tint_private_vars));
   }
   return;

diff --git a/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_no_loop.wgsl.expected.msl b/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_no_loop.wgsl.expected.msl
index 1c06a5e..9d73779 100644
--- a/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_no_loop.wgsl.expected.msl
+++ b/test/tint/bug/fxc/vector_assignment_in_loop/loop_call_with_no_loop.wgsl.expected.msl

@@ -19,6 +19,7 @@
 kernel void tint_symbol() {
   thread tint_private_vars_struct tint_private_vars = {};
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     foo(&(tint_private_vars));
   }
   return;

diff --git a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_all.wgsl.expected.msl b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_all.wgsl.expected.msl
index eb3e7d2..664550c 100644
--- a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_all.wgsl.expected.msl
+++ b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_all.wgsl.expected.msl

@@ -15,6 +15,7 @@
   bool3 v3b = false;
   bool4 v4b = false;
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     v2f[i] = 1.0f;
     v3f[i] = 1.0f;
     v4f[i] = 1.0f;

diff --git a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_repeated.wgsl.expected.msl b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_repeated.wgsl.expected.msl
index 9b58f40..2d918b9 100644
--- a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_repeated.wgsl.expected.msl
+++ b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_repeated.wgsl.expected.msl

@@ -11,6 +11,7 @@
   bool2 v2b = false;
   bool2 v2b_2 = false;
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     v2f[i] = 1.0f;
     v3i[i] = 1;
     v4u[i] = 1u;

diff --git a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_some.wgsl.expected.msl b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_some.wgsl.expected.msl
index fa9d188..5425ca7 100644
--- a/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_some.wgsl.expected.msl
+++ b/test/tint/bug/fxc/vector_assignment_in_loop/loop_types_some.wgsl.expected.msl

@@ -15,6 +15,7 @@
   bool3 v3b = false;
   bool4 v4b = false;
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     v2f[i] = 1.0f;
     v2i[i] = 1;
     v2u[i] = 1u;

diff --git a/test/tint/bug/tint/1064.wgsl.expected.msl b/test/tint/bug/tint/1064.wgsl.expected.msl
index 194a6cf..454d69c 100644
--- a/test/tint/bug/tint/1064.wgsl.expected.msl
+++ b/test/tint/bug/tint/1064.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 fragment void tint_symbol() {
   while (true) {
+    __asm__("");
     if (false) {
     } else {
       break;

diff --git a/test/tint/bug/tint/1081.wgsl.expected.msl b/test/tint/bug/tint/1081.wgsl.expected.msl
index 5b9ce54..c1181ce 100644
--- a/test/tint/bug/tint/1081.wgsl.expected.msl
+++ b/test/tint/bug/tint/1081.wgsl.expected.msl

@@ -23,6 +23,7 @@
 int tint_symbol_inner(int3 x, thread tint_private_vars_struct* const tint_private_vars) {
   int y = x[0];
   while (true) {
+    __asm__("");
     int const r = f(y, tint_private_vars);
     if ((r == 0)) {
       break;

diff --git a/test/tint/bug/tint/1121.wgsl.expected.msl b/test/tint/bug/tint/1121.wgsl.expected.msl
index 339e957..f0db93a 100644
--- a/test/tint/bug/tint/1121.wgsl.expected.msl
+++ b/test/tint/bug/tint/1121.wgsl.expected.msl

@@ -85,7 +85,9 @@
   int const TILE_COUNT_X = 2;
   int const TILE_COUNT_Y = 2;
   for(int y = 0; (y < TILE_COUNT_Y); y = as_type<int>((as_type<uint>(y) + as_type<uint>(1)))) {
+    __asm__("");
     for(int x = 0; (x < TILE_COUNT_X); x = as_type<int>((as_type<uint>(x) + as_type<uint>(1)))) {
+      __asm__("");
       int2 tilePixel0Idx = int2(as_type<int>((as_type<uint>(x) * as_type<uint>(TILE_SIZE))), as_type<int>((as_type<uint>(y) * as_type<uint>(TILE_SIZE))));
       float2 floorCoord = (((2.0f * float2(tilePixel0Idx)) / (*(tint_symbol_3)).fullScreenSize.xy) - float2(1.0f));
       float2 ceilCoord = (((2.0f * float2(as_type<int2>((as_type<uint2>(tilePixel0Idx) + as_type<uint2>(int2(TILE_SIZE)))))) / (*(tint_symbol_3)).fullScreenSize.xy) - float2(1.0f));
@@ -97,6 +99,7 @@
       frustumPlanes[3] = float4(0.0f, -1.0f, (viewCeilCoord[1] / viewNear), 0.0f);
       float dp = 0.0f;
       for(uint i = 0u; (i < 6u); i = (i + 1u)) {
+        __asm__("");
         float4 p = 0.0f;
         if ((frustumPlanes[i][0] > 0.0f)) {
           p[0] = boxMax[0];

diff --git a/test/tint/bug/tint/1321.wgsl.expected.msl b/test/tint/bug/tint/1321.wgsl.expected.msl
index aa1daf6..7d3878a 100644
--- a/test/tint/bug/tint/1321.wgsl.expected.msl
+++ b/test/tint/bug/tint/1321.wgsl.expected.msl

@@ -24,6 +24,7 @@
     int const tint_symbol_1 = foo();
     int const a_save = tint_symbol_1;
     while (true) {
+      __asm__("");
       {
         float const x = arr[a_save];
         break;

diff --git a/test/tint/bug/tint/1474-a.wgsl.expected.msl b/test/tint/bug/tint/1474-a.wgsl.expected.msl
index 593ea73..4cb74c7 100644
--- a/test/tint/bug/tint/1474-a.wgsl.expected.msl
+++ b/test/tint/bug/tint/1474-a.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 kernel void tint_symbol() {
   while(true) {
+    __asm__("");
     if (true) {
       break;
     } else {

diff --git a/test/tint/bug/tint/1538.wgsl.expected.msl b/test/tint/bug/tint/1538.wgsl.expected.msl
index ca25b03..6e5e6e1 100644
--- a/test/tint/bug/tint/1538.wgsl.expected.msl
+++ b/test/tint/bug/tint/1538.wgsl.expected.msl

@@ -20,6 +20,7 @@
 
 int f() {
   while (true) {
+    __asm__("");
     g();
     break;
   }
@@ -29,6 +30,7 @@
 
 kernel void tint_symbol(device tint_array<uint, 1>* tint_symbol_1 [[buffer(0)]]) {
   while (true) {
+    __asm__("");
     if (((*(tint_symbol_1))[0] == 0u)) {
       break;
     }

diff --git a/test/tint/bug/tint/1557.wgsl.expected.msl b/test/tint/bug/tint/1557.wgsl.expected.msl
index 3b62b3c..e61acec 100644
--- a/test/tint/bug/tint/1557.wgsl.expected.msl
+++ b/test/tint/bug/tint/1557.wgsl.expected.msl

@@ -8,6 +8,7 @@
 void g() {
   int j = 0;
   while (true) {
+    __asm__("");
     if ((j >= 1)) {
       break;
     }

diff --git a/test/tint/bug/tint/1604.wgsl.expected.msl b/test/tint/bug/tint/1604.wgsl.expected.msl
index ba49092..d2f041d 100644
--- a/test/tint/bug/tint/1604.wgsl.expected.msl
+++ b/test/tint/bug/tint/1604.wgsl.expected.msl

@@ -5,6 +5,7 @@
   switch(*(tint_symbol_1)) {
     case 0: {
       while (true) {
+        __asm__("");
         return;
       }
       break;

diff --git a/test/tint/bug/tint/1605.wgsl.expected.msl b/test/tint/bug/tint/1605.wgsl.expected.msl
index 18e89e5..e4a6e0b 100644
--- a/test/tint/bug/tint/1605.wgsl.expected.msl
+++ b/test/tint/bug/tint/1605.wgsl.expected.msl

@@ -3,7 +3,9 @@
 using namespace metal;
 bool func_3(const constant int* const tint_symbol_1) {
   for(int i = 0; (i < *(tint_symbol_1)); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     for(int j = -1; (j == 1); j = as_type<int>((as_type<uint>(j) + as_type<uint>(1)))) {
+      __asm__("");
       return false;
     }
   }

diff --git a/test/tint/bug/tint/1739.wgsl.expected.msl b/test/tint/bug/tint/1739.wgsl.expected.msl
index d626b35..f034375 100644
--- a/test/tint/bug/tint/1739.wgsl.expected.msl
+++ b/test/tint/bug/tint/1739.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/bug/tint/1764.wgsl.expected.msl b/test/tint/bug/tint/1764.wgsl.expected.msl
index 06b186f..b1cc428 100644
--- a/test/tint/bug/tint/1764.wgsl.expected.msl
+++ b/test/tint/bug/tint/1764.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup tint_array<int, 246>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 246u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_1))[i] = 0;
   }

diff --git a/test/tint/bug/tint/2010.spvasm.expected.msl b/test/tint/bug/tint/2010.spvasm.expected.msl
index b24f5bf..f46cd3a 100644
--- a/test/tint/bug/tint/2010.spvasm.expected.msl
+++ b/test/tint/bug/tint/2010.spvasm.expected.msl

@@ -47,6 +47,7 @@
   uint const x_52 = (*(tint_private_vars)).x_3[0];
   x_54 = 0u;
   while (true) {
+    __asm__("");
     uint x_55 = 0u;
     x_58 = (*(tint_symbol_3)).field0.field0;
     if ((x_54 < x_58)) {
@@ -79,6 +80,7 @@
   x_85 = x_76.xyxy;
   x_88 = 1u;
   while (true) {
+    __asm__("");
     float4 x_111 = 0.0f;
     float4 x_86 = 0.0f;
     uint x_89 = 0u;
@@ -129,6 +131,7 @@
     atomic_store_explicit(tint_symbol_14, 0u, memory_order_relaxed);
   }
   for(uint idx = local_invocation_index; (idx < 4096u); idx = (idx + 32u)) {
+    __asm__("");
     uint const i = idx;
     S const tint_symbol_1 = S{};
     (*(tint_symbol_15))[i] = tint_symbol_1;

diff --git a/test/tint/bug/tint/2059.wgsl.expected.msl b/test/tint/bug/tint/2059.wgsl.expected.msl
index 98ac37e..9bc2e44 100644
--- a/test/tint/bug/tint/2059.wgsl.expected.msl
+++ b/test/tint/bug/tint/2059.wgsl.expected.msl

@@ -63,6 +63,7 @@
 
 void assign_and_preserve_padding_3(device tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 1>* const dest, tint_array<float3x3, 1> value) {
   for(uint i = 0u; (i < 1u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding(&((*(dest))[i]), value[i]);
   }
 }
@@ -77,6 +78,7 @@
 
 void assign_and_preserve_padding_6(device tint_array<S_tint_packed_vec3, 1>* const dest, tint_array<S, 1> value) {
   for(uint i = 0u; (i < 1u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }
@@ -87,6 +89,7 @@
 
 void assign_and_preserve_padding_7(device tint_array<S2_tint_packed_vec3, 1>* const dest, tint_array<S2, 1> value) {
   for(uint i = 0u; (i < 1u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_2(&((*(dest))[i]), value[i]);
   }
 }
@@ -94,6 +97,7 @@
 kernel void tint_symbol(device tint_array<tint_packed_vec3_f32_array_element, 3>* tint_symbol_8 [[buffer(0)]], device S_tint_packed_vec3* tint_symbol_9 [[buffer(1)]], device S2_tint_packed_vec3* tint_symbol_10 [[buffer(2)]], device S3_tint_packed_vec3* tint_symbol_11 [[buffer(3)]], device S4_tint_packed_vec3* tint_symbol_12 [[buffer(4)]], device tint_array<tint_array<tint_packed_vec3_f32_array_element, 3>, 1>* tint_symbol_13 [[buffer(5)]], device tint_array<S_tint_packed_vec3, 1>* tint_symbol_14 [[buffer(6)]], device tint_array<S2_tint_packed_vec3, 1>* tint_symbol_15 [[buffer(7)]]) {
   float3x3 m = float3x3(0.0f);
   for(uint c = 0u; (c < 3u); c = (c + 1u)) {
+    __asm__("");
     m[c] = float3(float(((c * 3u) + 1u)), float(((c * 3u) + 2u)), float(((c * 3u) + 3u)));
   }
   {

diff --git a/test/tint/bug/tint/221.wgsl.expected.msl b/test/tint/bug/tint/221.wgsl.expected.msl
index 9a2fe97..83beacb 100644
--- a/test/tint/bug/tint/221.wgsl.expected.msl
+++ b/test/tint/bug/tint/221.wgsl.expected.msl

@@ -26,6 +26,7 @@
 kernel void tint_symbol(device Buf* tint_symbol_1 [[buffer(0)]]) {
   uint i = 0u;
   while (true) {
+    __asm__("");
     if ((i >= (*(tint_symbol_1)).count)) {
       break;
     }

diff --git a/test/tint/bug/tint/534.wgsl.expected.msl b/test/tint/bug/tint/534.wgsl.expected.msl
index bd4ec92..2b51953 100644
--- a/test/tint/bug/tint/534.wgsl.expected.msl
+++ b/test/tint/bug/tint/534.wgsl.expected.msl

@@ -46,6 +46,7 @@
   uint4 srcColorBits = 0u;
   uint4 dstColorBits = tint_ftou(dstColor);
   for(uint i = 0u; (i < (*(tint_symbol_3)).channelCount); i = (i + 1u)) {
+    __asm__("");
     uint const tint_symbol_1 = i;
     srcColorBits[tint_symbol_1] = ConvertToFp16FloatValue(srcColor[i]);
     success = (success && (srcColorBits[i] == dstColorBits[i]));

diff --git a/test/tint/bug/tint/744.wgsl.expected.msl b/test/tint/bug/tint/744.wgsl.expected.msl
index 7c01635..d539045 100644
--- a/test/tint/bug/tint/744.wgsl.expected.msl
+++ b/test/tint/bug/tint/744.wgsl.expected.msl

@@ -30,6 +30,7 @@
   uint const dimOutter = (*(tint_symbol_1)).outShape[1];
   uint result = 0u;
   for(uint i = 0u; (i < dimInner); i = (i + 1u)) {
+    __asm__("");
     uint const a = (i + (resultCell[0] * dimInner));
     uint const b = (resultCell[1] + (i * dimOutter));
     result = (result + ((*(tint_symbol_2)).numbers[a] * (*(tint_symbol_3)).numbers[b]));

diff --git a/test/tint/bug/tint/757.wgsl.expected.msl b/test/tint/bug/tint/757.wgsl.expected.msl
index f9ee13b..e6671f0 100644
--- a/test/tint/bug/tint/757.wgsl.expected.msl
+++ b/test/tint/bug/tint/757.wgsl.expected.msl

@@ -27,6 +27,7 @@
   flatIndex = (flatIndex * 1u);
   float4 texel = tint_symbol_1.read(uint2(int2(GlobalInvocationID.xy)), 0, 0);
   for(uint i = 0u; (i < 1u); i = (i + 1u)) {
+    __asm__("");
     (*(tint_symbol_2)).values[(flatIndex + i)] = texel[0];
   }
 }

diff --git a/test/tint/bug/tint/914.wgsl.expected.msl b/test/tint/bug/tint/914.wgsl.expected.msl
index 7dd3ccd..adb0758 100644
--- a/test/tint/bug/tint/914.wgsl.expected.msl
+++ b/test/tint/bug/tint/914.wgsl.expected.msl

@@ -53,6 +53,7 @@
 
 void tint_symbol_inner(uint3 local_id, uint3 global_id, uint local_invocation_index, threadgroup tint_array<tint_array<float, 64>, 64>* const tint_symbol_11, threadgroup tint_array<tint_array<float, 64>, 64>* const tint_symbol_12, const constant Uniforms* const tint_symbol_13, const device Matrix* const tint_symbol_14, const device Matrix* const tint_symbol_15, device Matrix* const tint_symbol_16) {
   for(uint idx = local_invocation_index; (idx < 4096u); idx = (idx + 256u)) {
+    __asm__("");
     uint const i = (idx / 64u);
     uint const i_1 = (idx % 64u);
     (*(tint_symbol_11))[i][i_1] = 0.0f;
@@ -68,6 +69,7 @@
   float ACached = 0.0f;
   tint_array<float, 4> BCached = {};
   for(uint index = 0u; (index < 16u); index = (index + 1u)) {
+    __asm__("");
     acc[index] = 0.0f;
   }
   uint const ColPerThreadA = 4u;
@@ -75,8 +77,11 @@
   uint const RowPerThreadB = 4u;
   uint const tileRowB = (local_id[1] * RowPerThreadB);
   for(uint t = 0u; (t < numTiles); t = (t + 1u)) {
+    __asm__("");
     for(uint innerRow = 0u; (innerRow < 4u); innerRow = (innerRow + 1u)) {
+      __asm__("");
       for(uint innerCol = 0u; (innerCol < ColPerThreadA); innerCol = (innerCol + 1u)) {
+        __asm__("");
         uint const inputRow = (tileRow + innerRow);
         uint const inputCol = (tileColA + innerCol);
         uint const tint_symbol_1 = inputRow;
@@ -85,7 +90,9 @@
       }
     }
     for(uint innerRow = 0u; (innerRow < RowPerThreadB); innerRow = (innerRow + 1u)) {
+      __asm__("");
       for(uint innerCol = 0u; (innerCol < 4u); innerCol = (innerCol + 1u)) {
+        __asm__("");
         uint const inputRow = (tileRowB + innerRow);
         uint const inputCol = (tileCol + innerCol);
         uint const tint_symbol_3 = innerCol;
@@ -95,12 +102,16 @@
     }
     threadgroup_barrier(mem_flags::mem_threadgroup);
     for(uint k = 0u; (k < 64u); k = (k + 1u)) {
+      __asm__("");
       for(uint inner = 0u; (inner < 4u); inner = (inner + 1u)) {
+        __asm__("");
         BCached[inner] = (*(tint_symbol_12))[k][(tileCol + inner)];
       }
       for(uint innerRow = 0u; (innerRow < 4u); innerRow = (innerRow + 1u)) {
+        __asm__("");
         ACached = (*(tint_symbol_11))[(tileRow + innerRow)][k];
         for(uint innerCol = 0u; (innerCol < 4u); innerCol = (innerCol + 1u)) {
+          __asm__("");
           uint const index = ((innerRow * 4u) + innerCol);
           acc[index] = (acc[index] + (ACached * BCached[innerCol]));
         }
@@ -109,7 +120,9 @@
     threadgroup_barrier(mem_flags::mem_threadgroup);
   }
   for(uint innerRow = 0u; (innerRow < 4u); innerRow = (innerRow + 1u)) {
+    __asm__("");
     for(uint innerCol = 0u; (innerCol < 4u); innerCol = (innerCol + 1u)) {
+      __asm__("");
       uint const index = ((innerRow * 4u) + innerCol);
       mm_write((globalRow + innerRow), (globalCol + innerCol), acc[index], tint_symbol_13, tint_symbol_16);
     }

diff --git a/test/tint/bug/tint/942.wgsl.expected.msl b/test/tint/bug/tint/942.wgsl.expected.msl
index 6453075..81b81f1 100644
--- a/test/tint/bug/tint/942.wgsl.expected.msl
+++ b/test/tint/bug/tint/942.wgsl.expected.msl

@@ -29,6 +29,7 @@
 
 void tint_symbol_inner(uint3 WorkGroupID, uint3 LocalInvocationID, uint local_invocation_index, threadgroup tint_array<tint_array<float3, 256>, 4>* const tint_symbol_1, const constant Params* const tint_symbol_2, texture2d<float, access::sample> tint_symbol_3, const constant Flip* const tint_symbol_4, sampler tint_symbol_5, texture2d<float, access::write> tint_symbol_6) {
   for(uint idx = local_invocation_index; (idx < 1024u); idx = (idx + 64u)) {
+    __asm__("");
     uint const i_1 = (idx / 256u);
     uint const i_2 = (idx % 256u);
     (*(tint_symbol_1))[i_1][i_2] = float3(0.0f);
@@ -38,7 +39,9 @@
   uint2 const dims = uint2(tint_symbol_3.get_width(0), tint_symbol_3.get_height(0));
   uint2 const baseIndex = (((WorkGroupID.xy * uint2((*(tint_symbol_2)).blockDim, 4u)) + (LocalInvocationID.xy * uint2(4u, 1u))) - uint2(filterOffset, 0u));
   for(uint r = 0u; (r < 4u); r = (r + 1u)) {
+    __asm__("");
     for(uint c = 0u; (c < 4u); c = (c + 1u)) {
+      __asm__("");
       uint2 loadIndex = (baseIndex + uint2(c, r));
       if (((*(tint_symbol_4)).value != 0u)) {
         loadIndex = loadIndex.yx;
@@ -48,7 +51,9 @@
   }
   threadgroup_barrier(mem_flags::mem_threadgroup);
   for(uint r = 0u; (r < 4u); r = (r + 1u)) {
+    __asm__("");
     for(uint c = 0u; (c < 4u); c = (c + 1u)) {
+      __asm__("");
       uint2 writeIndex = (baseIndex + uint2(c, r));
       if (((*(tint_symbol_4)).value != 0u)) {
         writeIndex = writeIndex.yx;
@@ -57,6 +62,7 @@
       if ((((center >= filterOffset) && (center < (256u - filterOffset))) && all((writeIndex < dims)))) {
         float3 acc = float3(0.0f);
         for(uint f = 0u; (f < (*(tint_symbol_2)).filterDim); f = (f + 1u)) {
+          __asm__("");
           uint i = ((center + f) - filterOffset);
           acc = (acc + ((1.0f / float((*(tint_symbol_2)).filterDim)) * (*(tint_symbol_1))[r][i]));
         }

diff --git a/test/tint/bug/tint/948.wgsl.expected.msl b/test/tint/bug/tint/948.wgsl.expected.msl
index ae3f8b8..57ef1a9 100644
--- a/test/tint/bug/tint/948.wgsl.expected.msl
+++ b/test/tint/bug/tint/948.wgsl.expected.msl

@@ -99,6 +99,7 @@
   stageUnits = (float2(1.0f) / x_111);
   i = 0;
   while (true) {
+    __asm__("");
     int const x_122 = i;
     if ((x_122 < 2)) {
     } else {
@@ -135,6 +136,7 @@
       (*(tint_private_vars)).mt = fmod((x_181 * x_184), 1.0f);
       f = 0.0f;
       while (true) {
+        __asm__("");
         float const x_193 = f;
         if ((x_193 < 8.0f)) {
         } else {

diff --git a/test/tint/bug/tint/949.wgsl.expected.msl b/test/tint/bug/tint/949.wgsl.expected.msl
index fb8ec0f..5c21f24 100644
--- a/test/tint/bug/tint/949.wgsl.expected.msl
+++ b/test/tint/bug/tint/949.wgsl.expected.msl

@@ -331,6 +331,7 @@
   currSampledHeight = 1.0f;
   i = 0;
   while (true) {
+    __asm__("");
     int const x_388 = i;
     if ((x_388 < 15)) {
     } else {

diff --git a/test/tint/bug/tint/990.wgsl.expected.msl b/test/tint/bug/tint/990.wgsl.expected.msl
index 1725260..ff2feee 100644
--- a/test/tint/bug/tint/990.wgsl.expected.msl
+++ b/test/tint/bug/tint/990.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void f() {
   int i = 0;
   for(; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/builtins/atomicStore/array/aliased_arrays.spvasm.expected.msl b/test/tint/builtins/atomicStore/array/aliased_arrays.spvasm.expected.msl
index 397daf1..6407f10 100644
--- a/test/tint/builtins/atomicStore/array/aliased_arrays.spvasm.expected.msl
+++ b/test/tint/builtins/atomicStore/array/aliased_arrays.spvasm.expected.msl

@@ -30,6 +30,7 @@
   uint idx = 0u;
   idx = local_invocation_index_2;
   while (true) {
+    __asm__("");
     if (!((idx < 6u))) {
       break;
     }
@@ -54,6 +55,7 @@
 
 void compute_main_inner_1(uint local_invocation_index_1_param, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_2) {
   for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 6u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i = (idx_1 / 2u);
     uint const i_1 = (idx_1 % 2u);
     uint const i_2 = (idx_1 % 1u);

diff --git a/test/tint/builtins/atomicStore/array/aliased_arrays.wgsl.expected.msl b/test/tint/builtins/atomicStore/array/aliased_arrays.wgsl.expected.msl
index 51e5245..7deaf57 100644
--- a/test/tint/builtins/atomicStore/array/aliased_arrays.wgsl.expected.msl
+++ b/test/tint/builtins/atomicStore/array/aliased_arrays.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void compute_main_inner(uint local_invocation_index, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol) {
   for(uint idx = local_invocation_index; (idx < 6u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = (idx / 2u);
     uint const i_1 = (idx % 2u);
     uint const i_2 = (idx % 1u);

diff --git a/test/tint/builtins/atomicStore/array/array.spvasm.expected.msl b/test/tint/builtins/atomicStore/array/array.spvasm.expected.msl
index d45e49b..856968a 100644
--- a/test/tint/builtins/atomicStore/array/array.spvasm.expected.msl
+++ b/test/tint/builtins/atomicStore/array/array.spvasm.expected.msl

@@ -22,6 +22,7 @@
   uint idx = 0u;
   idx = local_invocation_index_2;
   while (true) {
+    __asm__("");
     if (!((idx < 4u))) {
       break;
     }
@@ -44,6 +45,7 @@
 
 void compute_main_inner_1(uint local_invocation_index_1_param, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<atomic_uint, 4>* const tint_symbol_2) {
   for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 4u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i = idx_1;
     atomic_store_explicit(&((*(tint_symbol_2))[i]), 0u, memory_order_relaxed);
   }

diff --git a/test/tint/builtins/atomicStore/array/array.wgsl.expected.msl b/test/tint/builtins/atomicStore/array/array.wgsl.expected.msl
index 5f20d27..d3036aa 100644
--- a/test/tint/builtins/atomicStore/array/array.wgsl.expected.msl
+++ b/test/tint/builtins/atomicStore/array/array.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void compute_main_inner(uint local_invocation_index, threadgroup tint_array<atomic_uint, 4>* const tint_symbol) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     atomic_store_explicit(&((*(tint_symbol))[i]), 0u, memory_order_relaxed);
   }

diff --git a/test/tint/builtins/atomicStore/array/arrays.spvasm.expected.msl b/test/tint/builtins/atomicStore/array/arrays.spvasm.expected.msl
index 397daf1..6407f10 100644
--- a/test/tint/builtins/atomicStore/array/arrays.spvasm.expected.msl
+++ b/test/tint/builtins/atomicStore/array/arrays.spvasm.expected.msl

@@ -30,6 +30,7 @@
   uint idx = 0u;
   idx = local_invocation_index_2;
   while (true) {
+    __asm__("");
     if (!((idx < 6u))) {
       break;
     }
@@ -54,6 +55,7 @@
 
 void compute_main_inner_1(uint local_invocation_index_1_param, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol_2) {
   for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 6u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i = (idx_1 / 2u);
     uint const i_1 = (idx_1 % 2u);
     uint const i_2 = (idx_1 % 1u);

diff --git a/test/tint/builtins/atomicStore/array/arrays.wgsl.expected.msl b/test/tint/builtins/atomicStore/array/arrays.wgsl.expected.msl
index 51e5245..7deaf57 100644
--- a/test/tint/builtins/atomicStore/array/arrays.wgsl.expected.msl
+++ b/test/tint/builtins/atomicStore/array/arrays.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void compute_main_inner(uint local_invocation_index, threadgroup tint_array<tint_array<tint_array<atomic_uint, 1>, 2>, 3>* const tint_symbol) {
   for(uint idx = local_invocation_index; (idx < 6u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = (idx / 2u);
     uint const i_1 = (idx % 2u);
     uint const i_2 = (idx % 1u);

diff --git a/test/tint/builtins/atomicStore/struct/array_of_struct.spvasm.expected.msl b/test/tint/builtins/atomicStore/struct/array_of_struct.spvasm.expected.msl
index cbb81fd..81f7e0e 100644
--- a/test/tint/builtins/atomicStore/struct/array_of_struct.spvasm.expected.msl
+++ b/test/tint/builtins/atomicStore/struct/array_of_struct.spvasm.expected.msl

@@ -34,6 +34,7 @@
   uint idx = 0u;
   idx = local_invocation_index_2;
   while (true) {
+    __asm__("");
     if (!((idx < 10u))) {
       break;
     }
@@ -58,6 +59,7 @@
 
 void compute_main_inner_1(uint local_invocation_index_1_param, thread tint_private_vars_struct* const tint_private_vars, threadgroup tint_array<S_atomic, 10>* const tint_symbol_2) {
   for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 10u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i = idx_1;
     (*(tint_symbol_2))[i].x = 0;
     atomic_store_explicit(&((*(tint_symbol_2))[i].a), 0u, memory_order_relaxed);

diff --git a/test/tint/builtins/atomicStore/struct/array_of_struct.wgsl.expected.msl b/test/tint/builtins/atomicStore/struct/array_of_struct.wgsl.expected.msl
index 978edb1..6d38f69 100644
--- a/test/tint/builtins/atomicStore/struct/array_of_struct.wgsl.expected.msl
+++ b/test/tint/builtins/atomicStore/struct/array_of_struct.wgsl.expected.msl

@@ -22,6 +22,7 @@
 
 void compute_main_inner(uint local_invocation_index, threadgroup tint_array<S, 10>* const tint_symbol) {
   for(uint idx = local_invocation_index; (idx < 10u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol))[i].x = 0;
     atomic_store_explicit(&((*(tint_symbol))[i].a), 0u, memory_order_relaxed);

diff --git a/test/tint/builtins/atomicStore/struct/struct_of_array.spvasm.expected.msl b/test/tint/builtins/atomicStore/struct/struct_of_array.spvasm.expected.msl
index 0d116fb..9f335ca 100644
--- a/test/tint/builtins/atomicStore/struct/struct_of_array.spvasm.expected.msl
+++ b/test/tint/builtins/atomicStore/struct/struct_of_array.spvasm.expected.msl

@@ -36,6 +36,7 @@
   (*(tint_symbol)).y = 0u;
   idx = local_invocation_index_2;
   while (true) {
+    __asm__("");
     if (!((idx < 10u))) {
       break;
     }
@@ -62,6 +63,7 @@
     (*(tint_symbol_2)).y = 0u;
   }
   for(uint idx_1 = local_invocation_index_1_param; (idx_1 < 10u); idx_1 = (idx_1 + 1u)) {
+    __asm__("");
     uint const i = idx_1;
     atomic_store_explicit(&((*(tint_symbol_2)).a[i]), 0u, memory_order_relaxed);
   }

diff --git a/test/tint/builtins/atomicStore/struct/struct_of_array.wgsl.expected.msl b/test/tint/builtins/atomicStore/struct/struct_of_array.wgsl.expected.msl
index 00fa7d5..edaee4cb 100644
--- a/test/tint/builtins/atomicStore/struct/struct_of_array.wgsl.expected.msl
+++ b/test/tint/builtins/atomicStore/struct/struct_of_array.wgsl.expected.msl

@@ -26,6 +26,7 @@
     (*(tint_symbol)).y = 0u;
   }
   for(uint idx = local_invocation_index; (idx < 10u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     atomic_store_explicit(&((*(tint_symbol)).a[i]), 0u, memory_order_relaxed);
   }

diff --git a/test/tint/builtins/gen/literal/textureLoad/1bfdfb.wgsl.expected.msl b/test/tint/builtins/gen/literal/textureLoad/1bfdfb.wgsl.expected.msl
index e2c5422..7f0579a 100644
--- a/test/tint/builtins/gen/literal/textureLoad/1bfdfb.wgsl.expected.msl
+++ b/test/tint/builtins/gen/literal/textureLoad/1bfdfb.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/gen/literal/textureLoad/8acf41.wgsl.expected.msl b/test/tint/builtins/gen/literal/textureLoad/8acf41.wgsl.expected.msl
index 0f06bb1..16deb0c 100644
--- a/test/tint/builtins/gen/literal/textureLoad/8acf41.wgsl.expected.msl
+++ b/test/tint/builtins/gen/literal/textureLoad/8acf41.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/gen/literal/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl b/test/tint/builtins/gen/literal/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl
index 082481f..e7f661f 100644
--- a/test/tint/builtins/gen/literal/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl
+++ b/test/tint/builtins/gen/literal/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/gen/var/textureLoad/1bfdfb.wgsl.expected.msl b/test/tint/builtins/gen/var/textureLoad/1bfdfb.wgsl.expected.msl
index 01899f1..1682906 100644
--- a/test/tint/builtins/gen/var/textureLoad/1bfdfb.wgsl.expected.msl
+++ b/test/tint/builtins/gen/var/textureLoad/1bfdfb.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/gen/var/textureLoad/8acf41.wgsl.expected.msl b/test/tint/builtins/gen/var/textureLoad/8acf41.wgsl.expected.msl
index 7256ccb..5652b46 100644
--- a/test/tint/builtins/gen/var/textureLoad/8acf41.wgsl.expected.msl
+++ b/test/tint/builtins/gen/var/textureLoad/8acf41.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/gen/var/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl b/test/tint/builtins/gen/var/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl
index a3ba7fd..6700304 100644
--- a/test/tint/builtins/gen/var/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl
+++ b/test/tint/builtins/gen/var/textureSampleBaseClampToEdge/7c04e6.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/textureLoad/texture_external_param.wgsl.expected.msl b/test/tint/builtins/textureLoad/texture_external_param.wgsl.expected.msl
index 5c92b89..2f324bc 100644
--- a/test/tint/builtins/textureLoad/texture_external_param.wgsl.expected.msl
+++ b/test/tint/builtins/textureLoad/texture_external_param.wgsl.expected.msl

@@ -45,6 +45,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/builtins/textureStore/loop_continuing_read_write_texture.wgsl.expected.msl b/test/tint/builtins/textureStore/loop_continuing_read_write_texture.wgsl.expected.msl
index 32d7ed3..82b5b7d 100644
--- a/test/tint/builtins/textureStore/loop_continuing_read_write_texture.wgsl.expected.msl
+++ b/test/tint/builtins/textureStore/loop_continuing_read_write_texture.wgsl.expected.msl

@@ -5,6 +5,7 @@
   {
     int i = 0;
     while (true) {
+      __asm__("");
       if (!((i < 3))) {
         break;
       }

diff --git a/test/tint/builtins/workgroupUniformLoad/for_loop.wgsl.expected.msl b/test/tint/builtins/workgroupUniformLoad/for_loop.wgsl.expected.msl
index 57923b6..5819f07 100644
--- a/test/tint/builtins/workgroupUniformLoad/for_loop.wgsl.expected.msl
+++ b/test/tint/builtins/workgroupUniformLoad/for_loop.wgsl.expected.msl

@@ -12,6 +12,7 @@
   {
     int i = 0;
     while (true) {
+      __asm__("");
       int const tint_symbol = i;
       int const tint_symbol_1 = tint_workgroupUniformLoad(tint_symbol_4);
       if (!((tint_symbol < tint_symbol_1))) {

diff --git a/test/tint/diagnostic_filtering/for_loop_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/for_loop_attribute.wgsl.expected.msl
index 90587efd..f722f40 100644
--- a/test/tint/diagnostic_filtering/for_loop_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/for_loop_attribute.wgsl.expected.msl

@@ -20,6 +20,7 @@
 void tint_symbol_inner(float x) {
   float4 v = float4(0.0f);
   for(; ((x > v[0]) && (dfdx(1.0f) > 0.0f)); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/diagnostic_filtering/for_loop_body_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/for_loop_body_attribute.wgsl.expected.msl
index 4335d2f..918a01b 100644
--- a/test/tint/diagnostic_filtering/for_loop_body_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/for_loop_body_attribute.wgsl.expected.msl

@@ -20,6 +20,7 @@
 void tint_symbol_inner(float x, texture2d<float, access::sample> tint_symbol_3, sampler tint_symbol_4) {
   float4 v = float4(0.0f);
   for(; (x > v[0]); ) {
+    __asm__("");
     v = tint_symbol_3.sample(tint_symbol_4, float2(0.0f));
   }
 }

diff --git a/test/tint/diagnostic_filtering/loop_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/loop_attribute.wgsl.expected.msl
index 5de56a9..86fc7d2 100644
--- a/test/tint/diagnostic_filtering/loop_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/loop_attribute.wgsl.expected.msl

@@ -19,6 +19,7 @@
 
 void tint_symbol_inner(float x) {
   while (true) {
+    __asm__("");
     {
       if ((x > 0.0f)) { break; }
     }

diff --git a/test/tint/diagnostic_filtering/loop_body_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/loop_body_attribute.wgsl.expected.msl
index 796b68c..063fabb 100644
--- a/test/tint/diagnostic_filtering/loop_body_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/loop_body_attribute.wgsl.expected.msl

@@ -19,6 +19,7 @@
 
 void tint_symbol_inner(float x) {
   while (true) {
+    __asm__("");
     {
       if ((x > 0.0f)) { break; }
     }

diff --git a/test/tint/diagnostic_filtering/loop_continuing_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/loop_continuing_attribute.wgsl.expected.msl
index c57d216..741fbad 100644
--- a/test/tint/diagnostic_filtering/loop_continuing_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/loop_continuing_attribute.wgsl.expected.msl

@@ -19,6 +19,7 @@
 
 void tint_symbol_inner(float x) {
   while (true) {
+    __asm__("");
     {
       if ((x > 0.0f)) { break; }
     }

diff --git a/test/tint/diagnostic_filtering/while_loop_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/while_loop_attribute.wgsl.expected.msl
index df56d48..7858cd2 100644
--- a/test/tint/diagnostic_filtering/while_loop_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/while_loop_attribute.wgsl.expected.msl

@@ -20,6 +20,7 @@
 void tint_symbol_inner(float x) {
   float4 v = float4(0.0f);
   while(((x > 0.0f) && (dfdx(1.0f) > 0.0f))) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/diagnostic_filtering/while_loop_body_attribute.wgsl.expected.msl b/test/tint/diagnostic_filtering/while_loop_body_attribute.wgsl.expected.msl
index 657eb2d..4eb60f8 100644
--- a/test/tint/diagnostic_filtering/while_loop_body_attribute.wgsl.expected.msl
+++ b/test/tint/diagnostic_filtering/while_loop_body_attribute.wgsl.expected.msl

@@ -20,6 +20,7 @@
 void tint_symbol_inner(float x, texture2d<float, access::sample> tint_symbol_3, sampler tint_symbol_4) {
   float4 v = float4(0.0f);
   while((x > v[0])) {
+    __asm__("");
     v = tint_symbol_3.sample(tint_symbol_4, float2(0.0f));
   }
 }

diff --git a/test/tint/expressions/binary/mul/mat3x3-vec3/f16.wgsl.expected.msl b/test/tint/expressions/binary/mul/mat3x3-vec3/f16.wgsl.expected.msl
index ff9f1c6..36c59c8 100644
--- a/test/tint/expressions/binary/mul/mat3x3-vec3/f16.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/mat3x3-vec3/f16.wgsl.expected.msl

@@ -28,6 +28,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/expressions/binary/mul/mat3x3-vec3/f32.wgsl.expected.msl b/test/tint/expressions/binary/mul/mat3x3-vec3/f32.wgsl.expected.msl
index 539e4ab..0afb867 100644
--- a/test/tint/expressions/binary/mul/mat3x3-vec3/f32.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/mat3x3-vec3/f32.wgsl.expected.msl

@@ -28,6 +28,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/expressions/binary/mul/vec3-mat3x3/f16.wgsl.expected.msl b/test/tint/expressions/binary/mul/vec3-mat3x3/f16.wgsl.expected.msl
index a1a9592..013c086 100644
--- a/test/tint/expressions/binary/mul/vec3-mat3x3/f16.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/vec3-mat3x3/f16.wgsl.expected.msl

@@ -28,6 +28,7 @@
 half3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 3> in) {
   half3x3 result = half3x3(0.0h);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/expressions/binary/mul/vec3-mat3x3/f32.wgsl.expected.msl b/test/tint/expressions/binary/mul/vec3-mat3x3/f32.wgsl.expected.msl
index a982edf..91e5eee 100644
--- a/test/tint/expressions/binary/mul/vec3-mat3x3/f32.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/vec3-mat3x3/f32.wgsl.expected.msl

@@ -28,6 +28,7 @@
 float3x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 3> in) {
   float3x3 result = float3x3(0.0f);
   for(uint i = 0u; (i < 3u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/expressions/binary/mul/vec3-mat4x3/f16.wgsl.expected.msl b/test/tint/expressions/binary/mul/vec3-mat4x3/f16.wgsl.expected.msl
index 571c316..5e12349 100644
--- a/test/tint/expressions/binary/mul/vec3-mat4x3/f16.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/vec3-mat4x3/f16.wgsl.expected.msl

@@ -28,6 +28,7 @@
 half4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f16_array_element, 4> in) {
   half4x3 result = half4x3(0.0h);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = half3(in[i].elements);
   }
   return result;

diff --git a/test/tint/expressions/binary/mul/vec3-mat4x3/f32.wgsl.expected.msl b/test/tint/expressions/binary/mul/vec3-mat4x3/f32.wgsl.expected.msl
index 420b7aa..e3ed262 100644
--- a/test/tint/expressions/binary/mul/vec3-mat4x3/f32.wgsl.expected.msl
+++ b/test/tint/expressions/binary/mul/vec3-mat4x3/f32.wgsl.expected.msl

@@ -28,6 +28,7 @@
 float4x3 tint_unpack_vec3_in_composite(tint_array<tint_packed_vec3_f32_array_element, 4> in) {
   float4x3 result = float4x3(0.0f);
   for(uint i = 0u; (i < 4u); i = (i + 1u)) {
+    __asm__("");
     result[i] = float3(in[i].elements);
   }
   return result;

diff --git a/test/tint/layout/storage/mat2x2/stride/16.spvasm.expected.msl b/test/tint/layout/storage/mat2x2/stride/16.spvasm.expected.msl
index 5a05bb7..03ea0b0 100644
--- a/test/tint/layout/storage/mat2x2/stride/16.spvasm.expected.msl
+++ b/test/tint/layout/storage/mat2x2/stride/16.spvasm.expected.msl

@@ -40,6 +40,7 @@
 
 void assign_and_preserve_padding(device tint_array<strided_arr, 2>* const dest, tint_array<strided_arr, 2> value) {
   for(uint i = 0u; (i < 2u); i = (i + 1u)) {
+    __asm__("");
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }

diff --git a/test/tint/loops/continue_in_switch.wgsl.expected.msl b/test/tint/loops/continue_in_switch.wgsl.expected.msl
index 494b64a..5d5f9c6 100644
--- a/test/tint/loops/continue_in_switch.wgsl.expected.msl
+++ b/test/tint/loops/continue_in_switch.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 kernel void f() {
   for(int i = 0; (i < 4); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         continue;

diff --git a/test/tint/loops/loop.wgsl.expected.msl b/test/tint/loops/loop.wgsl.expected.msl
index 49a3b1c..6eae6c8 100644
--- a/test/tint/loops/loop.wgsl.expected.msl
+++ b/test/tint/loops/loop.wgsl.expected.msl

@@ -4,6 +4,7 @@
 int f() {
   int i = 0;
   while (true) {
+    __asm__("");
     i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)));
     if ((i > 4)) {
       return i;

diff --git a/test/tint/loops/loop_with_break_if.wgsl.expected.msl b/test/tint/loops/loop_with_break_if.wgsl.expected.msl
index 97788fa..7dc8ba3 100644
--- a/test/tint/loops/loop_with_break_if.wgsl.expected.msl
+++ b/test/tint/loops/loop_with_break_if.wgsl.expected.msl

@@ -4,6 +4,7 @@
 int f() {
   int i = 0;
   while (true) {
+    __asm__("");
     if ((i > 4)) {
       return i;
     }

diff --git a/test/tint/loops/loop_with_continuing.wgsl.expected.msl b/test/tint/loops/loop_with_continuing.wgsl.expected.msl
index 52a015b..9a4f4d0 100644
--- a/test/tint/loops/loop_with_continuing.wgsl.expected.msl
+++ b/test/tint/loops/loop_with_continuing.wgsl.expected.msl

@@ -4,6 +4,7 @@
 int f() {
   int i = 0;
   while (true) {
+    __asm__("");
     if ((i > 4)) {
       return i;
     }

diff --git a/test/tint/loops/multiple_continues.wgsl.expected.msl b/test/tint/loops/multiple_continues.wgsl.expected.msl
index ed8fcc1..ed9d214 100644
--- a/test/tint/loops/multiple_continues.wgsl.expected.msl
+++ b/test/tint/loops/multiple_continues.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 kernel void tint_symbol() {
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         continue;

diff --git a/test/tint/loops/multiple_switch.wgsl.expected.msl b/test/tint/loops/multiple_switch.wgsl.expected.msl
index 342db50..adc1ac8 100644
--- a/test/tint/loops/multiple_switch.wgsl.expected.msl
+++ b/test/tint/loops/multiple_switch.wgsl.expected.msl

@@ -4,6 +4,7 @@
 kernel void tint_symbol() {
   int i = 0;
   for(int i_1 = 0; (i_1 < 2); i_1 = as_type<int>((as_type<uint>(i_1) + as_type<uint>(1)))) {
+    __asm__("");
     switch(i_1) {
       case 0: {
         continue;

diff --git a/test/tint/loops/nested_loop_loop_switch.wgsl.expected.msl b/test/tint/loops/nested_loop_loop_switch.wgsl.expected.msl
index a1f0843..6f58d1b 100644
--- a/test/tint/loops/nested_loop_loop_switch.wgsl.expected.msl
+++ b/test/tint/loops/nested_loop_loop_switch.wgsl.expected.msl

@@ -3,7 +3,9 @@
 using namespace metal;
 kernel void tint_symbol() {
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(2)))) {
+    __asm__("");
     for(int j = 0; (j < 2); j = as_type<int>((as_type<uint>(j) + as_type<uint>(2)))) {
+      __asm__("");
       switch(i) {
         case 0: {
           continue;

diff --git a/test/tint/loops/nested_loop_switch_loop_switch.wgsl.expected.msl b/test/tint/loops/nested_loop_switch_loop_switch.wgsl.expected.msl
index a5e768d..ce28e7d 100644
--- a/test/tint/loops/nested_loop_switch_loop_switch.wgsl.expected.msl
+++ b/test/tint/loops/nested_loop_switch_loop_switch.wgsl.expected.msl

@@ -3,9 +3,11 @@
 using namespace metal;
 kernel void tint_symbol() {
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(2)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         for(int j = 0; (j < 2); j = as_type<int>((as_type<uint>(j) + as_type<uint>(2)))) {
+          __asm__("");
           switch(j) {
             case 0: {
               continue;

diff --git a/test/tint/loops/nested_loop_switch_loop_switch_switch.wgsl.expected.msl b/test/tint/loops/nested_loop_switch_loop_switch_switch.wgsl.expected.msl
index b3a1fd6..2690672 100644
--- a/test/tint/loops/nested_loop_switch_loop_switch_switch.wgsl.expected.msl
+++ b/test/tint/loops/nested_loop_switch_loop_switch_switch.wgsl.expected.msl

@@ -4,9 +4,11 @@
 kernel void tint_symbol() {
   int k = 0;
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(2)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         for(int j = 0; (j < 2); j = as_type<int>((as_type<uint>(j) + as_type<uint>(2)))) {
+          __asm__("");
           switch(j) {
             case 0: {
               continue;

diff --git a/test/tint/loops/nested_loop_switch_switch.wgsl.expected.msl b/test/tint/loops/nested_loop_switch_switch.wgsl.expected.msl
index 55f4436..8402967 100644
--- a/test/tint/loops/nested_loop_switch_switch.wgsl.expected.msl
+++ b/test/tint/loops/nested_loop_switch_switch.wgsl.expected.msl

@@ -4,6 +4,7 @@
 kernel void tint_symbol() {
   int j = 0;
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(2)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         switch(j) {

diff --git a/test/tint/loops/nested_loops.wgsl.expected.msl b/test/tint/loops/nested_loops.wgsl.expected.msl
index 484b4ca..a84996f 100644
--- a/test/tint/loops/nested_loops.wgsl.expected.msl
+++ b/test/tint/loops/nested_loops.wgsl.expected.msl

@@ -5,11 +5,13 @@
   int i = 0;
   int j = 0;
   while (true) {
+    __asm__("");
     i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)));
     if ((i > 4)) {
       return 1;
     }
     while (true) {
+      __asm__("");
       j = as_type<int>((as_type<uint>(j) + as_type<uint>(1)));
       if ((j > 4)) {
         return 2;

diff --git a/test/tint/loops/nested_loops_with_continuing.wgsl.expected.msl b/test/tint/loops/nested_loops_with_continuing.wgsl.expected.msl
index f28c010..17720f6 100644
--- a/test/tint/loops/nested_loops_with_continuing.wgsl.expected.msl
+++ b/test/tint/loops/nested_loops_with_continuing.wgsl.expected.msl

@@ -5,10 +5,12 @@
   int i = 0;
   int j = 0;
   while (true) {
+    __asm__("");
     if ((i > 4)) {
       return 1;
     }
     while (true) {
+      __asm__("");
       if ((j > 4)) {
         return 2;
       }

diff --git a/test/tint/loops/single_continue.wgsl.expected.msl b/test/tint/loops/single_continue.wgsl.expected.msl
index 3ce3ba5..467b2e5 100644
--- a/test/tint/loops/single_continue.wgsl.expected.msl
+++ b/test/tint/loops/single_continue.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 kernel void tint_symbol() {
   for(int i = 0; (i < 2); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     switch(i) {
       case 0: {
         continue;

diff --git a/test/tint/loops/while.wgsl.expected.msl b/test/tint/loops/while.wgsl.expected.msl
index bd8f450..dd2c7ec 100644
--- a/test/tint/loops/while.wgsl.expected.msl
+++ b/test/tint/loops/while.wgsl.expected.msl

@@ -4,6 +4,7 @@
 int f() {
   int i = 0;
   while((i < 4)) {
+    __asm__("");
     i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)));
   }
   return i;

diff --git a/test/tint/loops/while_with_continue.wgsl.expected.msl b/test/tint/loops/while_with_continue.wgsl.expected.msl
index 7fee4b0..6c9936f 100644
--- a/test/tint/loops/while_with_continue.wgsl.expected.msl
+++ b/test/tint/loops/while_with_continue.wgsl.expected.msl

@@ -4,6 +4,7 @@
 int f() {
   int i = 0;
   while((i < 4)) {
+    __asm__("");
     i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)));
     continue;
   }

diff --git a/test/tint/ptr_ref/load/param/workgroup/array_in_struct.wgsl.expected.msl b/test/tint/ptr_ref/load/param/workgroup/array_in_struct.wgsl.expected.msl
index e8f0b1f..8b496c6 100644
--- a/test/tint/ptr_ref/load/param/workgroup/array_in_struct.wgsl.expected.msl
+++ b/test/tint/ptr_ref/load/param/workgroup/array_in_struct.wgsl.expected.msl

@@ -24,6 +24,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup str* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_1)).arr[i] = 0;
   }

diff --git a/test/tint/ptr_ref/load/param/workgroup/struct_in_array.wgsl.expected.msl b/test/tint/ptr_ref/load/param/workgroup/struct_in_array.wgsl.expected.msl
index 15c963d..b424dc8 100644
--- a/test/tint/ptr_ref/load/param/workgroup/struct_in_array.wgsl.expected.msl
+++ b/test/tint/ptr_ref/load/param/workgroup/struct_in_array.wgsl.expected.msl

@@ -24,6 +24,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup tint_array<str, 4>* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i_1 = idx;
     str const tint_symbol_1 = str{};
     (*(tint_symbol_2))[i_1] = tint_symbol_1;

diff --git a/test/tint/ptr_ref/store/param/workgroup/array_in_struct.wgsl.expected.msl b/test/tint/ptr_ref/store/param/workgroup/array_in_struct.wgsl.expected.msl
index 8652c09..b63f3c0 100644
--- a/test/tint/ptr_ref/store/param/workgroup/array_in_struct.wgsl.expected.msl
+++ b/test/tint/ptr_ref/store/param/workgroup/array_in_struct.wgsl.expected.msl

@@ -25,6 +25,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup str* const tint_symbol_2) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_2)).arr[i] = 0;
   }

diff --git a/test/tint/ptr_ref/store/param/workgroup/struct_in_array.wgsl.expected.msl b/test/tint/ptr_ref/store/param/workgroup/struct_in_array.wgsl.expected.msl
index bc419c0..01ad03d 100644
--- a/test/tint/ptr_ref/store/param/workgroup/struct_in_array.wgsl.expected.msl
+++ b/test/tint/ptr_ref/store/param/workgroup/struct_in_array.wgsl.expected.msl

@@ -25,6 +25,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup tint_array<str, 4>* const tint_symbol_3) {
   for(uint idx = local_invocation_index; (idx < 4u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i_1 = idx;
     str const tint_symbol_1 = str{};
     (*(tint_symbol_3))[i_1] = tint_symbol_1;

diff --git a/test/tint/samples/compute_boids.wgsl.expected.msl b/test/tint/samples/compute_boids.wgsl.expected.msl
index dd67382..e6b5e22 100644
--- a/test/tint/samples/compute_boids.wgsl.expected.msl
+++ b/test/tint/samples/compute_boids.wgsl.expected.msl

@@ -86,6 +86,7 @@
   float2 pos = 0.0f;
   float2 vel = 0.0f;
   for(uint i = 0u; (i < 5u); i = (i + 1u)) {
+    __asm__("");
     if ((i == index)) {
       continue;
     }

diff --git a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_body.wgsl.expected.msl b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_body.wgsl.expected.msl
index 939ff14..95e334c 100644
--- a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_body.wgsl.expected.msl
+++ b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_body.wgsl.expected.msl

@@ -30,6 +30,7 @@
   InnerS v = {};
   OuterS s1 = {};
   for(int i = 0; (i < 4); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     s1.a1[(*(tint_symbol_1)).i] = v;
   }
   return;

diff --git a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_continuing.wgsl.expected.msl b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_continuing.wgsl.expected.msl
index 76e27e9..33d1be5 100644
--- a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_continuing.wgsl.expected.msl
+++ b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_continuing.wgsl.expected.msl

@@ -30,6 +30,7 @@
   InnerS v = {};
   OuterS s1 = {};
   for(int i = 0; (i < 4); s1.a1[(*(tint_symbol_1)).i] = v) {
+    __asm__("");
     i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)));
   }
   return;

diff --git a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_init.wgsl.expected.msl b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_init.wgsl.expected.msl
index 82e2101..163f692 100644
--- a/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_init.wgsl.expected.msl
+++ b/test/tint/statements/assign/indexed_assign_to_array_in_struct/in_for_loop_init.wgsl.expected.msl

@@ -31,6 +31,7 @@
   OuterS s1 = {};
   int i = 0;
   for(s1.a1[(*(tint_symbol_1)).i] = v; (i < 4); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
   }
   return;
 }

diff --git a/test/tint/statements/compound_assign/for_loop.wgsl.expected.msl b/test/tint/statements/compound_assign/for_loop.wgsl.expected.msl
index cd34a12..f9d89b4 100644
--- a/test/tint/statements/compound_assign/for_loop.wgsl.expected.msl
+++ b/test/tint/statements/compound_assign/for_loop.wgsl.expected.msl

@@ -46,6 +46,7 @@
     int const tint_symbol_save = tint_symbol_2;
     a[tint_symbol_save] = (a[tint_symbol_save] * 2.0f);
     while (true) {
+      __asm__("");
       int const tint_symbol_3 = idx2(tint_private_vars);
       if (!((a[tint_symbol_3] < 10.0f))) {
         break;

diff --git a/test/tint/statements/decrement/complex.wgsl.expected.msl b/test/tint/statements/decrement/complex.wgsl.expected.msl
index 6bb4088..de484b1 100644
--- a/test/tint/statements/decrement/complex.wgsl.expected.msl
+++ b/test/tint/statements/decrement/complex.wgsl.expected.msl

@@ -61,6 +61,7 @@
     int const tint_symbol_3 = idx3(tint_private_vars);
     (*(tint_symbol_10))[tint_symbol_2_save].a[tint_symbol_2_save_1][tint_symbol_3] = as_type<int>((as_type<uint>((*(tint_symbol_10))[tint_symbol_2_save].a[tint_symbol_2_save_1][tint_symbol_3]) - as_type<uint>(1)));
     while (true) {
+      __asm__("");
       if (!(((*(tint_private_vars)).v < 10u))) {
         break;
       }

diff --git a/test/tint/statements/decrement/for_loop_continuing.wgsl.expected.msl b/test/tint/statements/decrement/for_loop_continuing.wgsl.expected.msl
index c91c6b2..faeb478 100644
--- a/test/tint/statements/decrement/for_loop_continuing.wgsl.expected.msl
+++ b/test/tint/statements/decrement/for_loop_continuing.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void tint_symbol(device uint* const tint_symbol_1) {
   for(; (*(tint_symbol_1) < 10u); *(tint_symbol_1) = (*(tint_symbol_1) - 1u)) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/decrement/for_loop_initializer.wgsl.expected.msl b/test/tint/statements/decrement/for_loop_initializer.wgsl.expected.msl
index 1f1ebdf..b700dca 100644
--- a/test/tint/statements/decrement/for_loop_initializer.wgsl.expected.msl
+++ b/test/tint/statements/decrement/for_loop_initializer.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void tint_symbol(device uint* const tint_symbol_1) {
   for(*(tint_symbol_1) = (*(tint_symbol_1) - 1u); (*(tint_symbol_1) < 10u); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/discard/atomic_in_for_loop_continuing.wgsl.expected.msl b/test/tint/statements/discard/atomic_in_for_loop_continuing.wgsl.expected.msl
index 36b7539..4931e68 100644
--- a/test/tint/statements/discard/atomic_in_for_loop_continuing.wgsl.expected.msl
+++ b/test/tint/statements/discard/atomic_in_for_loop_continuing.wgsl.expected.msl

@@ -26,6 +26,7 @@
   {
     int i = 0;
     while (true) {
+      __asm__("");
       if (!((i < 10))) {
         break;
       }

diff --git a/test/tint/statements/discard/multiple_returns.wgsl.expected.msl b/test/tint/statements/discard/multiple_returns.wgsl.expected.msl
index 7bbefc8..037a078 100644
--- a/test/tint/statements/discard/multiple_returns.wgsl.expected.msl
+++ b/test/tint/statements/discard/multiple_returns.wgsl.expected.msl

@@ -17,6 +17,7 @@
   if ((*(tint_symbol_2) < 0.0f)) {
     int i = 0;
     while (true) {
+      __asm__("");
       if ((*(tint_symbol_2) > float(i))) {
         if (!(tint_private_vars.tint_discarded)) {
           *(tint_symbol_2) = float(i);

diff --git a/test/tint/statements/for/basic.wgsl.expected.msl b/test/tint/statements/for/basic.wgsl.expected.msl
index 20c444b..d4ee785 100644
--- a/test/tint/statements/for/basic.wgsl.expected.msl
+++ b/test/tint/statements/for/basic.wgsl.expected.msl

@@ -6,6 +6,7 @@
 
 void f() {
   for(int i = 0; (i < 5); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     some_loop_body();
   }
 }

diff --git a/test/tint/statements/for/complex.wgsl.expected.msl b/test/tint/statements/for/complex.wgsl.expected.msl
index 534cd3a..297f77e 100644
--- a/test/tint/statements/for/complex.wgsl.expected.msl
+++ b/test/tint/statements/for/complex.wgsl.expected.msl

@@ -7,6 +7,7 @@
 void f() {
   int j = 0;
   for(int i = 0; ((i < 5) && (j < 10)); i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
     some_loop_body();
     j = as_type<int>((as_type<uint>(i) * as_type<uint>(30)));
   }

diff --git a/test/tint/statements/for/condition/array_ctor.wgsl.expected.msl b/test/tint/statements/for/condition/array_ctor.wgsl.expected.msl
index 31a0722..0bfd498 100644
--- a/test/tint/statements/for/condition/array_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/condition/array_ctor.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void f() {
   int i = 0;
   for(; (i < 1); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/condition/basic.wgsl.expected.msl b/test/tint/statements/for/condition/basic.wgsl.expected.msl
index 839541b..c29c578 100644
--- a/test/tint/statements/for/condition/basic.wgsl.expected.msl
+++ b/test/tint/statements/for/condition/basic.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void f() {
   int i = 0;
   for(; (i < 4); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/condition/struct_ctor.wgsl.expected.msl b/test/tint/statements/for/condition/struct_ctor.wgsl.expected.msl
index a0a3bf0..4b4bd74 100644
--- a/test/tint/statements/for/condition/struct_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/condition/struct_ctor.wgsl.expected.msl

@@ -8,6 +8,7 @@
 void f() {
   int i = 0;
   for(; (i < 1); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/continuing/array_ctor.wgsl.expected.msl b/test/tint/statements/for/continuing/array_ctor.wgsl.expected.msl
index 3130ddd..6c7e914 100644
--- a/test/tint/statements/for/continuing/array_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/continuing/array_ctor.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void f() {
   int i = 0;
   for(; false; i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/continuing/basic.wgsl.expected.msl b/test/tint/statements/for/continuing/basic.wgsl.expected.msl
index 3130ddd..6c7e914 100644
--- a/test/tint/statements/for/continuing/basic.wgsl.expected.msl
+++ b/test/tint/statements/for/continuing/basic.wgsl.expected.msl

@@ -4,6 +4,7 @@
 void f() {
   int i = 0;
   for(; false; i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/continuing/struct_ctor.wgsl.expected.msl b/test/tint/statements/for/continuing/struct_ctor.wgsl.expected.msl
index fd9b90b..45ae3b6 100644
--- a/test/tint/statements/for/continuing/struct_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/continuing/struct_ctor.wgsl.expected.msl

@@ -7,6 +7,7 @@
 
 void f() {
   for(int i = 0; false; i = as_type<int>((as_type<uint>(i) + as_type<uint>(1)))) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/empty.wgsl.expected.msl b/test/tint/statements/for/empty.wgsl.expected.msl
index a8d3f51..0032ddd 100644
--- a/test/tint/statements/for/empty.wgsl.expected.msl
+++ b/test/tint/statements/for/empty.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void f() {
   for(; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/initializer/array_ctor.wgsl.expected.msl b/test/tint/statements/for/initializer/array_ctor.wgsl.expected.msl
index ee30daf..2baee65 100644
--- a/test/tint/statements/for/initializer/array_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/initializer/array_ctor.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void f() {
   for(int i = 1; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/initializer/basic.wgsl.expected.msl b/test/tint/statements/for/initializer/basic.wgsl.expected.msl
index caf48e3..505277d 100644
--- a/test/tint/statements/for/initializer/basic.wgsl.expected.msl
+++ b/test/tint/statements/for/initializer/basic.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void f() {
   for(int i = 0; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/initializer/struct_ctor.wgsl.expected.msl b/test/tint/statements/for/initializer/struct_ctor.wgsl.expected.msl
index a497720..6718a1f 100644
--- a/test/tint/statements/for/initializer/struct_ctor.wgsl.expected.msl
+++ b/test/tint/statements/for/initializer/struct_ctor.wgsl.expected.msl

@@ -7,6 +7,7 @@
 
 void f() {
   for(int i = 1; false; ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/for/scoping.wgsl.expected.msl b/test/tint/statements/for/scoping.wgsl.expected.msl
index c6d9118..4a7d54e 100644
--- a/test/tint/statements/for/scoping.wgsl.expected.msl
+++ b/test/tint/statements/for/scoping.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void f() {
   for(int must_not_collide = 0; ; ) {
+    __asm__("");
     break;
   }
   int must_not_collide = 0;

diff --git a/test/tint/statements/increment/complex.wgsl.expected.msl b/test/tint/statements/increment/complex.wgsl.expected.msl
index ab1347f..ec39dfd 100644
--- a/test/tint/statements/increment/complex.wgsl.expected.msl
+++ b/test/tint/statements/increment/complex.wgsl.expected.msl

@@ -61,6 +61,7 @@
     int const tint_symbol_3 = idx3(tint_private_vars);
     (*(tint_symbol_10))[tint_symbol_2_save].a[tint_symbol_2_save_1][tint_symbol_3] = as_type<int>((as_type<uint>((*(tint_symbol_10))[tint_symbol_2_save].a[tint_symbol_2_save_1][tint_symbol_3]) + as_type<uint>(1)));
     while (true) {
+      __asm__("");
       if (!(((*(tint_private_vars)).v < 10u))) {
         break;
       }

diff --git a/test/tint/statements/increment/for_loop_continuing.wgsl.expected.msl b/test/tint/statements/increment/for_loop_continuing.wgsl.expected.msl
index ee23499..239618b 100644
--- a/test/tint/statements/increment/for_loop_continuing.wgsl.expected.msl
+++ b/test/tint/statements/increment/for_loop_continuing.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void tint_symbol(device uint* const tint_symbol_1) {
   for(; (*(tint_symbol_1) < 10u); *(tint_symbol_1) = (*(tint_symbol_1) + 1u)) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/statements/increment/for_loop_initializer.wgsl.expected.msl b/test/tint/statements/increment/for_loop_initializer.wgsl.expected.msl
index b29a247..37b9419 100644
--- a/test/tint/statements/increment/for_loop_initializer.wgsl.expected.msl
+++ b/test/tint/statements/increment/for_loop_initializer.wgsl.expected.msl

@@ -3,6 +3,7 @@
 using namespace metal;
 void tint_symbol(device uint* const tint_symbol_1) {
   for(*(tint_symbol_1) = (*(tint_symbol_1) + 1u); (*(tint_symbol_1) < 10u); ) {
+    __asm__("");
   }
 }
 

diff --git a/test/tint/var/initialization/workgroup/array/array_i32.wgsl.expected.msl b/test/tint/var/initialization/workgroup/array/array_i32.wgsl.expected.msl
index 81c7619..cc02af2 100644
--- a/test/tint/var/initialization/workgroup/array/array_i32.wgsl.expected.msl
+++ b/test/tint/var/initialization/workgroup/array/array_i32.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup tint_array<tint_array<int, 3>, 2>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 6u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = (idx / 3u);
     uint const i_1 = (idx % 3u);
     (*(tint_symbol_1))[i][i_1] = 0;

diff --git a/test/tint/var/initialization/workgroup/array/i32.wgsl.expected.msl b/test/tint/var/initialization/workgroup/array/i32.wgsl.expected.msl
index 5b98f4a..00dd95f 100644
--- a/test/tint/var/initialization/workgroup/array/i32.wgsl.expected.msl
+++ b/test/tint/var/initialization/workgroup/array/i32.wgsl.expected.msl

@@ -16,6 +16,7 @@
 
 void tint_symbol_inner(uint local_invocation_index, threadgroup tint_array<int, 3>* const tint_symbol_1) {
   for(uint idx = local_invocation_index; (idx < 3u); idx = (idx + 1u)) {
+    __asm__("");
     uint const i = idx;
     (*(tint_symbol_1))[i] = 0;
   }