tint/msl: Preserve trailing vec3 padding

In order to preserve padding properly for MSL, we need to use its
packed_vec type for all vec3 types in storage buffers, not just struct
members. This commit includes a complete rewrite of the PackedVec3
transform to achieve this. The key details are:

* An internal `__packed_vec3<>` type was added, which corresponds to a
  `type::Vector` with an additional flag to indicate that it will be
  emitted as packed vector.

* The `PackedVec3` transform replaces all vec3 types used in
  host-shareable address spaces with the internal `__packed_vec3`
  type. This includes vec3 types that appear as the store type of a
  pointer.

* When used as an array element, these `__packed_vec3` types are
  wrapped in a struct that contains a single `__packed_vec3`
  member. This allows us to add an `@align()` attribute that ensures
  that `array<vec3<T>>` still has the correct array element stride.

* When the `vec3<T>` appears as a struct member in the input program,
  we apply the `@align()` to that member to ensure that we do not
  change its offset.

* Matrix types with three rows that are used in memory are replaced
  with an array of columns, where each column uses a `__packed_vec3`
  inside an aligned wrapper structure as above.

* Accesses to host-shareable memory that involve any of these types
  invoke a "pack" or "unpack" helper function to convert them to the
  equivalent type that uses `__packed_vec3` or a regular `vec3` as
  required.

* The `chromium_internal_relaxed_uniform_layout` extension is used to
  avoid issues where modifying a type in the uniform address space
  triggers stricter layout validation rules.

Bug: tint:1571
Fixed: tint:1837
Change-Id: Idaf2da2f5bcb2be00c85ec657edfb614186476bb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121200
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>

test/tint/buffer/uniform/std140/array/mat4x3_f16/dynamic_index_via_ptr.wgsl.expected.msl

@@ -14,20 +14,41 @@ struct tint_array {
     T elements[N];
 };
 
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
 int i() {
   thread int tint_symbol_2 = 0;
   tint_symbol_2 = as_type<int>((as_type<uint>(tint_symbol_2) + as_type<uint>(1)));
   return tint_symbol_2;
 }
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol_3 [[buffer(0)]]) {
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol_3 [[buffer(0)]]) {
   int const tint_symbol = i();
   int const p_a_i_save = tint_symbol;
   int const tint_symbol_1 = i();
   int const p_a_i_i_save = tint_symbol_1;
-  tint_array<half4x3, 4> const l_a = *(tint_symbol_3);
-  half4x3 const l_a_i = (*(tint_symbol_3))[p_a_i_save];
-  half3 const l_a_i_i = (*(tint_symbol_3))[p_a_i_save][p_a_i_i_save];
+  tint_array<half4x3, 4> const l_a = tint_unpack_vec3_in_composite_1(*(tint_symbol_3));
+  half4x3 const l_a_i = tint_unpack_vec3_in_composite((*(tint_symbol_3))[p_a_i_save]);
+  half3 const l_a_i_i = half3((*(tint_symbol_3))[p_a_i_save][p_a_i_i_save].elements);
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/static_index_via_ptr.wgsl.expected.msl

@@ -14,10 +14,31 @@ struct tint_array {
     T elements[N];
 };
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol [[buffer(0)]]) {
-  tint_array<half4x3, 4> const l_a = *(tint_symbol);
-  half4x3 const l_a_i = (*(tint_symbol))[2];
-  half3 const l_a_i_i = (*(tint_symbol))[2][1];
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol [[buffer(0)]]) {
+  tint_array<half4x3, 4> const l_a = tint_unpack_vec3_in_composite_1(*(tint_symbol));
+  half4x3 const l_a_i = tint_unpack_vec3_in_composite((*(tint_symbol))[2]);
+  half3 const l_a_i_i = half3((*(tint_symbol))[2][1].elements);
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/to_builtin.wgsl.expected.msl

@@ -14,10 +14,23 @@ struct tint_array {
     T elements[N];
 };
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol [[buffer(0)]]) {
-  half3x4 const t = transpose((*(tint_symbol))[2]);
-  half const l = length(half3((*(tint_symbol))[0][1]).zxy);
-  half const a = fabs(half3((*(tint_symbol))[0][1]).zxy[0]);
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol [[buffer(0)]]) {
+  half3x4 const t = transpose(tint_unpack_vec3_in_composite((*(tint_symbol))[2]));
+  half const l = length(half3((*(tint_symbol))[0][1].elements).zxy);
+  half const a = fabs(half3((*(tint_symbol))[0][1].elements).zxy[0]);
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/to_fn.wgsl.expected.msl

@@ -14,6 +14,27 @@ struct tint_array {
     T elements[N];
 };
 
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
 void a(tint_array<half4x3, 4> a_1) {
 }
 
@@ -26,11 +47,11 @@ void c(half3 v) {
 void d(half f_1) {
 }
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol [[buffer(0)]]) {
-  a(*(tint_symbol));
-  b((*(tint_symbol))[1]);
-  c(half3((*(tint_symbol))[1][0]).zxy);
-  d(half3((*(tint_symbol))[1][0]).zxy[0]);
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol [[buffer(0)]]) {
+  a(tint_unpack_vec3_in_composite_1(*(tint_symbol)));
+  b(tint_unpack_vec3_in_composite((*(tint_symbol))[1]));
+  c(half3((*(tint_symbol))[1][0].elements).zxy);
+  d(half3((*(tint_symbol))[1][0].elements).zxy[0]);
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/to_private.wgsl.expected.msl

@@ -14,12 +14,33 @@ struct tint_array {
     T elements[N];
 };
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol_1 [[buffer(0)]]) {
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol_1 [[buffer(0)]]) {
   thread tint_array<half4x3, 4> tint_symbol = {};
-  tint_symbol = *(tint_symbol_1);
-  tint_symbol[1] = (*(tint_symbol_1))[2];
-  tint_symbol[1][0] = half3((*(tint_symbol_1))[0][1]).zxy;
-  tint_symbol[1][0][0] = (*(tint_symbol_1))[0][1][0];
+  tint_symbol = tint_unpack_vec3_in_composite_1(*(tint_symbol_1));
+  tint_symbol[1] = tint_unpack_vec3_in_composite((*(tint_symbol_1))[2]);
+  tint_symbol[1][0] = half3((*(tint_symbol_1))[0][1].elements).zxy;
+  tint_symbol[1][0][0] = (*(tint_symbol_1))[0][1].elements[0];
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/to_storage.wgsl.expected.msl

@@ -14,24 +14,45 @@ struct tint_array {
     T elements[N];
 };
 
-void assign_and_preserve_padding_1(device half4x3* const dest, half4x3 value) {
-  (*(dest))[0u] = value[0u];
-  (*(dest))[1u] = value[1u];
-  (*(dest))[2u] = value[2u];
-  (*(dest))[3u] = value[3u];
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
 }
 
-void assign_and_preserve_padding(device tint_array<half4x3, 4>* const dest, tint_array<half4x3, 4> value) {
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
+void assign_and_preserve_padding_1(device tint_array<tint_packed_vec3_f16_array_element, 4>* const dest, half4x3 value) {
+  (*(dest))[0u].elements = packed_half3(value[0u]);
+  (*(dest))[1u].elements = packed_half3(value[1u]);
+  (*(dest))[2u].elements = packed_half3(value[2u]);
+  (*(dest))[3u].elements = packed_half3(value[3u]);
+}
+
+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)) {
     assign_and_preserve_padding_1(&((*(dest))[i]), value[i]);
   }
 }
 
-kernel void f(device tint_array<half4x3, 4>* tint_symbol [[buffer(1)]], const constant tint_array<half4x3, 4>* tint_symbol_1 [[buffer(0)]]) {
-  assign_and_preserve_padding(tint_symbol, *(tint_symbol_1));
-  assign_and_preserve_padding_1(&((*(tint_symbol))[1]), (*(tint_symbol_1))[2]);
-  (*(tint_symbol))[1][0] = half3((*(tint_symbol_1))[0][1]).zxy;
-  (*(tint_symbol))[1][0][0] = (*(tint_symbol_1))[0][1][0];
+kernel void f(device tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol [[buffer(1)]], const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol_1 [[buffer(0)]]) {
+  assign_and_preserve_padding(tint_symbol, tint_unpack_vec3_in_composite_1(*(tint_symbol_1)));
+  assign_and_preserve_padding_1(&((*(tint_symbol))[1]), tint_unpack_vec3_in_composite((*(tint_symbol_1))[2]));
+  (*(tint_symbol))[1][0].elements = packed_half3(half3((*(tint_symbol_1))[0][1].elements).zxy);
+  (*(tint_symbol))[1][0].elements[0] = (*(tint_symbol_1))[0][1].elements[0];
   return;
 }
 

test/tint/buffer/uniform/std140/array/mat4x3_f16/to_workgroup.wgsl.expected.msl

@@ -18,19 +18,40 @@ struct tint_symbol_5 {
   tint_array<half4x3, 4> w;
 };
 
-void f_inner(uint local_invocation_index, threadgroup tint_array<half4x3, 4>* const tint_symbol, const constant tint_array<half4x3, 4>* const tint_symbol_1) {
+struct tint_packed_vec3_f16_array_element {
+  /* 0x0000 */ packed_half3 elements;
+  /* 0x0006 */ tint_array<int8_t, 2> tint_pad;
+};
+
+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)) {
+    result[i] = half3(in[i].elements);
+  }
+  return result;
+}
+
+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)) {
+    result[i] = tint_unpack_vec3_in_composite(in[i]);
+  }
+  return result;
+}
+
+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)) {
     uint const i = idx;
     (*(tint_symbol))[i] = half4x3(half3(0.0h), half3(0.0h), half3(0.0h), half3(0.0h));
   }
   threadgroup_barrier(mem_flags::mem_threadgroup);
-  *(tint_symbol) = *(tint_symbol_1);
-  (*(tint_symbol))[1] = (*(tint_symbol_1))[2];
-  (*(tint_symbol))[1][0] = half3((*(tint_symbol_1))[0][1]).zxy;
-  (*(tint_symbol))[1][0][0] = (*(tint_symbol_1))[0][1][0];
+  *(tint_symbol) = tint_unpack_vec3_in_composite_1(*(tint_symbol_1));
+  (*(tint_symbol))[1] = tint_unpack_vec3_in_composite((*(tint_symbol_1))[2]);
+  (*(tint_symbol))[1][0] = half3((*(tint_symbol_1))[0][1].elements).zxy;
+  (*(tint_symbol))[1][0][0] = (*(tint_symbol_1))[0][1].elements[0];
 }
 
-kernel void f(const constant tint_array<half4x3, 4>* tint_symbol_4 [[buffer(0)]], threadgroup tint_symbol_5* tint_symbol_3 [[threadgroup(0)]], uint local_invocation_index [[thread_index_in_threadgroup]]) {
+kernel void f(const constant tint_array<tint_array<tint_packed_vec3_f16_array_element, 4>, 4>* tint_symbol_4 [[buffer(0)]], threadgroup tint_symbol_5* tint_symbol_3 [[threadgroup(0)]], uint local_invocation_index [[thread_index_in_threadgroup]]) {
   threadgroup tint_array<half4x3, 4>* const tint_symbol_2 = &((*(tint_symbol_3)).w);
   f_inner(local_invocation_index, tint_symbol_2, tint_symbol_4);
   return;