ByteAddressBuffer sb : register(t0); struct tint_symbol_1 { uint idx : SV_GroupIndex; }; float2x2 sb_load_12(uint offset) { return float2x2(asfloat(sb.Load2((offset + 0u))), asfloat(sb.Load2((offset + 8u)))); } float2x3 sb_load_13(uint offset) { return float2x3(asfloat(sb.Load3((offset + 0u))), asfloat(sb.Load3((offset + 16u)))); } float2x4 sb_load_14(uint offset) { return float2x4(asfloat(sb.Load4((offset + 0u))), asfloat(sb.Load4((offset + 16u)))); } float3x2 sb_load_15(uint offset) { return float3x2(asfloat(sb.Load2((offset + 0u))), asfloat(sb.Load2((offset + 8u))), asfloat(sb.Load2((offset + 16u)))); } float3x3 sb_load_16(uint offset) { return float3x3(asfloat(sb.Load3((offset + 0u))), asfloat(sb.Load3((offset + 16u))), asfloat(sb.Load3((offset + 32u)))); } float3x4 sb_load_17(uint offset) { return float3x4(asfloat(sb.Load4((offset + 0u))), asfloat(sb.Load4((offset + 16u))), asfloat(sb.Load4((offset + 32u)))); } float4x2 sb_load_18(uint offset) { return float4x2(asfloat(sb.Load2((offset + 0u))), asfloat(sb.Load2((offset + 8u))), asfloat(sb.Load2((offset + 16u))), asfloat(sb.Load2((offset + 24u)))); } float4x3 sb_load_19(uint offset) { return float4x3(asfloat(sb.Load3((offset + 0u))), asfloat(sb.Load3((offset + 16u))), asfloat(sb.Load3((offset + 32u))), asfloat(sb.Load3((offset + 48u)))); } float4x4 sb_load_20(uint offset) { return float4x4(asfloat(sb.Load4((offset + 0u))), asfloat(sb.Load4((offset + 16u))), asfloat(sb.Load4((offset + 32u))), asfloat(sb.Load4((offset + 48u)))); } typedef float3 sb_load_21_ret[2]; sb_load_21_ret sb_load_21(uint offset) { float3 arr_1[2] = (float3[2])0; { for(uint i = 0u; (i < 2u); i = (i + 1u)) { arr_1[i] = asfloat(sb.Load3((offset + (i * 16u)))); } } return arr_1; } void main_inner(uint idx) { const float scalar_f32 = asfloat(sb.Load((544u * idx))); const int scalar_i32 = asint(sb.Load(((544u * idx) + 4u))); const uint scalar_u32 = sb.Load(((544u * idx) + 8u)); const float2 vec2_f32 = asfloat(sb.Load2(((544u * idx) + 16u))); const int2 vec2_i32 = asint(sb.Load2(((544u * idx) + 24u))); const uint2 vec2_u32 = sb.Load2(((544u * idx) + 32u)); const float3 vec3_f32 = asfloat(sb.Load3(((544u * idx) + 48u))); const int3 vec3_i32 = asint(sb.Load3(((544u * idx) + 64u))); const uint3 vec3_u32 = sb.Load3(((544u * idx) + 80u)); const float4 vec4_f32 = asfloat(sb.Load4(((544u * idx) + 96u))); const int4 vec4_i32 = asint(sb.Load4(((544u * idx) + 112u))); const uint4 vec4_u32 = sb.Load4(((544u * idx) + 128u)); const float2x2 mat2x2_f32 = sb_load_12(((544u * idx) + 144u)); const float2x3 mat2x3_f32 = sb_load_13(((544u * idx) + 160u)); const float2x4 mat2x4_f32 = sb_load_14(((544u * idx) + 192u)); const float3x2 mat3x2_f32 = sb_load_15(((544u * idx) + 224u)); const float3x3 mat3x3_f32 = sb_load_16(((544u * idx) + 256u)); const float3x4 mat3x4_f32 = sb_load_17(((544u * idx) + 304u)); const float4x2 mat4x2_f32 = sb_load_18(((544u * idx) + 352u)); const float4x3 mat4x3_f32 = sb_load_19(((544u * idx) + 384u)); const float4x4 mat4x4_f32 = sb_load_20(((544u * idx) + 448u)); const float3 arr2_vec3_f32[2] = sb_load_21(((544u * idx) + 512u)); } [numthreads(1, 1, 1)] void main(tint_symbol_1 tint_symbol) { main_inner(tint_symbol.idx); return; }