Using inverse in GLSL appears to cause an issue when parsing resulting WGSL (wgpu v0.8.1)

[mitchmindtree]

While updating one of nannou's camera/teapot examples from 0.7 to 0.8.1, I ran into the following parser error:

wgpu error: Validation Error

Caused by:
    In Device::create_shader_module
      note: label = `nannou_shader_module`
    Failed to parse WGSL


thread 'main' panicked at 'Handling wgpu errors as fatal by default', /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/backend/direct.rs:1955:5
stack backtrace:
   0: std::panicking::begin_panic
             at /rustc/2fd73fabe469357a12c2c974c140f67e7cdd76d0/library/std/src/panicking.rs:519:12
   1: wgpu::backend::direct::default_error_handler
             at /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/backend/direct.rs:1955:5
   2: core::ops::function::Fn::call
             at /rustc/2fd73fabe469357a12c2c974c140f67e7cdd76d0/library/core/src/ops/function.rs:70:5
   3: <alloc::boxed::Box<F,A> as core::ops::function::Fn<Args>>::call
             at /rustc/2fd73fabe469357a12c2c974c140f67e7cdd76d0/library/alloc/src/boxed.rs:1535:9
   4: wgpu::backend::direct::ErrorSinkRaw::handle_error
             at /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/backend/direct.rs:1942:9
   5: wgpu::backend::direct::Context::handle_error
             at /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/backend/direct.rs:93:9
   6: <wgpu::backend::direct::Context as wgpu::Context>::device_create_shader_module
             at /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/backend/direct.rs:819:13
   7: wgpu::Device::create_shader_module
             at /home/mindtree/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.8.1/src/lib.rs:1537:17
   8: nannou::wgpu::shader_from_spirv_bytes
             at /home/mindtree/programming/rust/nannou/nannou/src/wgpu/mod.rs:95:5
   9: wgpu_teapot_camera::model
             at /home/mindtree/programming/rust/nannou/examples/wgpu/wgpu_teapot_camera/wgpu_teapot_camera.rs:102:18
  10: nannou::app::Builder<M,E>::run
             at /home/mindtree/programming/rust/nannou/nannou/src/app.rs:484:21
  11: wgpu_teapot_camera::main
             at /home/mindtree/programming/rust/nannou/examples/wgpu/wgpu_teapot_camera/wgpu_teapot_camera.rs:81:5
  12: core::ops::function::FnOnce::call_once
             at /rustc/2fd73fabe469357a12c2c974c140f67e7cdd76d0/library/core/src/ops/function.rs:227:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

The issue seems to be related to the use of the GLSL inverse function within the example's vertex shader:

#version 450

layout(location = 0) in vec3 position;
layout(location = 1) in vec3 normal;

layout(location = 0) out vec3 v_normal;

layout(set = 0, binding = 0) uniform Data {
    mat4 world;
    mat4 view;
    mat4 proj;
} uniforms;

void main() {
    mat4 worldview = uniforms.view * uniforms.world;
    v_normal = transpose(inverse(mat3(worldview))) * normal;
    gl_Position = uniforms.proj * worldview * vec4(position, 1.0);
}

Here's the SPIR-V generated via glslangValidator:

; SPIR-V
; Version: 1.0
; Generator: Khronos Glslang Reference Front End; 10
; Bound: 62
; Schema: 0
               OpCapability Shader
          %1 = OpExtInstImport "GLSL.std.450"
               OpMemoryModel Logical GLSL450
               OpEntryPoint Vertex %main "main" %v_normal %normal %_ %position
               OpSource GLSL 450
               OpName %main "main"
               OpName %worldview "worldview"
               OpName %Data "Data"
               OpMemberName %Data 0 "world"
               OpMemberName %Data 1 "view"
               OpMemberName %Data 2 "proj"
               OpName %uniforms "uniforms"
               OpName %v_normal "v_normal"
               OpName %normal "normal"
               OpName %gl_PerVertex "gl_PerVertex"
               OpMemberName %gl_PerVertex 0 "gl_Position"
               OpMemberName %gl_PerVertex 1 "gl_PointSize"
               OpMemberName %gl_PerVertex 2 "gl_ClipDistance"
               OpMemberName %gl_PerVertex 3 "gl_CullDistance"
               OpName %_ ""
               OpName %position "position"
               OpMemberDecorate %Data 0 ColMajor
               OpMemberDecorate %Data 0 Offset 0
               OpMemberDecorate %Data 0 MatrixStride 16
               OpMemberDecorate %Data 1 ColMajor
               OpMemberDecorate %Data 1 Offset 64
               OpMemberDecorate %Data 1 MatrixStride 16
               OpMemberDecorate %Data 2 ColMajor
               OpMemberDecorate %Data 2 Offset 128
               OpMemberDecorate %Data 2 MatrixStride 16
               OpDecorate %Data Block
               OpDecorate %uniforms DescriptorSet 0
               OpDecorate %uniforms Binding 0
               OpDecorate %v_normal Location 0
               OpDecorate %normal Location 1
               OpMemberDecorate %gl_PerVertex 0 BuiltIn Position
               OpMemberDecorate %gl_PerVertex 1 BuiltIn PointSize
               OpMemberDecorate %gl_PerVertex 2 BuiltIn ClipDistance
               OpMemberDecorate %gl_PerVertex 3 BuiltIn CullDistance
               OpDecorate %gl_PerVertex Block
               OpDecorate %position Location 0
       %void = OpTypeVoid
          %3 = OpTypeFunction %void
      %float = OpTypeFloat 32
    %v4float = OpTypeVector %float 4
%mat4v4float = OpTypeMatrix %v4float 4
%_ptr_Function_mat4v4float = OpTypePointer Function %mat4v4float
       %Data = OpTypeStruct %mat4v4float %mat4v4float %mat4v4float
%_ptr_Uniform_Data = OpTypePointer Uniform %Data
   %uniforms = OpVariable %_ptr_Uniform_Data Uniform
        %int = OpTypeInt 32 1
      %int_1 = OpConstant %int 1
%_ptr_Uniform_mat4v4float = OpTypePointer Uniform %mat4v4float
      %int_0 = OpConstant %int 0
    %v3float = OpTypeVector %float 3
%_ptr_Output_v3float = OpTypePointer Output %v3float
   %v_normal = OpVariable %_ptr_Output_v3float Output
%mat3v3float = OpTypeMatrix %v3float 3
%_ptr_Input_v3float = OpTypePointer Input %v3float
     %normal = OpVariable %_ptr_Input_v3float Input
       %uint = OpTypeInt 32 0
     %uint_1 = OpConstant %uint 1
%_arr_float_uint_1 = OpTypeArray %float %uint_1
%gl_PerVertex = OpTypeStruct %v4float %float %_arr_float_uint_1 %_arr_float_uint_1
%_ptr_Output_gl_PerVertex = OpTypePointer Output %gl_PerVertex
          %_ = OpVariable %_ptr_Output_gl_PerVertex Output
      %int_2 = OpConstant %int 2
   %position = OpVariable %_ptr_Input_v3float Input
    %float_1 = OpConstant %float 1
%_ptr_Output_v4float = OpTypePointer Output %v4float
       %main = OpFunction %void None %3
          %5 = OpLabel
  %worldview = OpVariable %_ptr_Function_mat4v4float Function
         %17 = OpAccessChain %_ptr_Uniform_mat4v4float %uniforms %int_1
         %18 = OpLoad %mat4v4float %17
         %20 = OpAccessChain %_ptr_Uniform_mat4v4float %uniforms %int_0
         %21 = OpLoad %mat4v4float %20
         %22 = OpMatrixTimesMatrix %mat4v4float %18 %21
               OpStore %worldview %22
         %26 = OpLoad %mat4v4float %worldview
         %28 = OpCompositeExtract %v4float %26 0
         %29 = OpVectorShuffle %v3float %28 %28 0 1 2
         %30 = OpCompositeExtract %v4float %26 1
         %31 = OpVectorShuffle %v3float %30 %30 0 1 2
         %32 = OpCompositeExtract %v4float %26 2
         %33 = OpVectorShuffle %v3float %32 %32 0 1 2
         %34 = OpCompositeConstruct %mat3v3float %29 %31 %33
         %35 = OpExtInst %mat3v3float %1 MatrixInverse %34
         %36 = OpTranspose %mat3v3float %35
         %39 = OpLoad %v3float %normal
         %40 = OpMatrixTimesVector %v3float %36 %39
               OpStore %v_normal %40
         %48 = OpAccessChain %_ptr_Uniform_mat4v4float %uniforms %int_2
         %49 = OpLoad %mat4v4float %48
         %50 = OpLoad %mat4v4float %worldview
         %51 = OpMatrixTimesMatrix %mat4v4float %49 %50
         %53 = OpLoad %v3float %position
         %55 = OpCompositeExtract %float %53 0
         %56 = OpCompositeExtract %float %53 1
         %57 = OpCompositeExtract %float %53 2
         %58 = OpCompositeConstruct %v4float %55 %56 %57 %float_1
         %59 = OpMatrixTimesVector %v4float %51 %58
         %61 = OpAccessChain %_ptr_Output_v4float %_ %int_0
               OpStore %61 %59
               OpReturn
               OpFunctionEnd

If I remove the inverse call the shader appears to pass validation successfully.

Tested on NixOS, GNOME Wayland.

Possibly related:

gfx-rs/wgpu#4330
#469
gfx-rs/wgpu#4336

[JCapucho]

Tested on e1552fd, the wgsl backend seems to no longer generate shaders which use inverse (since wgsl doesn't support it) and spirv consumes the input well.