blob: 5a5e16b31d2dbbe085dee272c587c4837438ad1c [file] [log] [blame]
enable f16;
struct Inner {
scalar_f32 : f32,
scalar_i32 : i32,
scalar_u32 : u32,
scalar_f16 : f16,
vec2_f32 : vec2<f32>,
vec2_i32 : vec2<i32>,
vec2_u32 : vec2<u32>,
vec2_f16 : vec2<f16>,
vec3_f32 : vec3<f32>,
vec3_i32 : vec3<i32>,
vec3_u32 : vec3<u32>,
vec3_f16 : vec3<f16>,
vec4_f32 : vec4<f32>,
vec4_i32 : vec4<i32>,
vec4_u32 : vec4<u32>,
vec4_f16 : vec4<f16>,
mat2x2_f32 : mat2x2<f32>,
mat2x3_f32 : mat2x3<f32>,
mat2x4_f32 : mat2x4<f32>,
mat3x2_f32 : mat3x2<f32>,
mat3x3_f32 : mat3x3<f32>,
mat3x4_f32 : mat3x4<f32>,
mat4x2_f32 : mat4x2<f32>,
mat4x3_f32 : mat4x3<f32>,
mat4x4_f32 : mat4x4<f32>,
mat2x2_f16 : mat2x2<f16>,
mat2x3_f16 : mat2x3<f16>,
mat2x4_f16 : mat2x4<f16>,
mat3x2_f16 : mat3x2<f16>,
mat3x3_f16 : mat3x3<f16>,
mat3x4_f16 : mat3x4<f16>,
mat4x2_f16 : mat4x2<f16>,
mat4x3_f16 : mat4x3<f16>,
mat4x4_f16 : mat4x4<f16>,
arr2_vec3_f32 : array<vec3<f32>, 2>,
arr2_mat4x2_f16 : array<mat4x2<f16>, 2>,
}
struct S {
arr : array<Inner>,
}
@binding(0) @group(0) var<storage, read> sb : S;
@compute @workgroup_size(1)
fn main(@builtin(local_invocation_index) idx : u32) {
let scalar_f32 : f32 = sb.arr[idx].scalar_f32;
let scalar_i32 : i32 = sb.arr[idx].scalar_i32;
let scalar_u32 : u32 = sb.arr[idx].scalar_u32;
let scalar_f16 : f16 = sb.arr[idx].scalar_f16;
let vec2_f32 : vec2<f32> = sb.arr[idx].vec2_f32;
let vec2_i32 : vec2<i32> = sb.arr[idx].vec2_i32;
let vec2_u32 : vec2<u32> = sb.arr[idx].vec2_u32;
let vec2_f16 : vec2<f16> = sb.arr[idx].vec2_f16;
let vec3_f32 : vec3<f32> = sb.arr[idx].vec3_f32;
let vec3_i32 : vec3<i32> = sb.arr[idx].vec3_i32;
let vec3_u32 : vec3<u32> = sb.arr[idx].vec3_u32;
let vec3_f16 : vec3<f16> = sb.arr[idx].vec3_f16;
let vec4_f32 : vec4<f32> = sb.arr[idx].vec4_f32;
let vec4_i32 : vec4<i32> = sb.arr[idx].vec4_i32;
let vec4_u32 : vec4<u32> = sb.arr[idx].vec4_u32;
let vec4_f16 : vec4<f16> = sb.arr[idx].vec4_f16;
let mat2x2_f32 : mat2x2<f32> = sb.arr[idx].mat2x2_f32;
let mat2x3_f32 : mat2x3<f32> = sb.arr[idx].mat2x3_f32;
let mat2x4_f32 : mat2x4<f32> = sb.arr[idx].mat2x4_f32;
let mat3x2_f32 : mat3x2<f32> = sb.arr[idx].mat3x2_f32;
let mat3x3_f32 : mat3x3<f32> = sb.arr[idx].mat3x3_f32;
let mat3x4_f32 : mat3x4<f32> = sb.arr[idx].mat3x4_f32;
let mat4x2_f32 : mat4x2<f32> = sb.arr[idx].mat4x2_f32;
let mat4x3_f32 : mat4x3<f32> = sb.arr[idx].mat4x3_f32;
let mat4x4_f32 : mat4x4<f32> = sb.arr[idx].mat4x4_f32;
let mat2x2_f16 : mat2x2<f16> = sb.arr[idx].mat2x2_f16;
let mat2x3_f16 : mat2x3<f16> = sb.arr[idx].mat2x3_f16;
let mat2x4_f16 : mat2x4<f16> = sb.arr[idx].mat2x4_f16;
let mat3x2_f16 : mat3x2<f16> = sb.arr[idx].mat3x2_f16;
let mat3x3_f16 : mat3x3<f16> = sb.arr[idx].mat3x3_f16;
let mat3x4_f16 : mat3x4<f16> = sb.arr[idx].mat3x4_f16;
let mat4x2_f16 : mat4x2<f16> = sb.arr[idx].mat4x2_f16;
let mat4x3_f16 : mat4x3<f16> = sb.arr[idx].mat4x3_f16;
let mat4x4_f16 : mat4x4<f16> = sb.arr[idx].mat4x4_f16;
let arr2_vec3_f32 : array<vec3<f32>, 2> = sb.arr[idx].arr2_vec3_f32;
let arr2_mat4x2_f16 : array<mat4x2<f16>, 2> = sb.arr[idx].arr2_mat4x2_f16;
}