//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-vk -compute -output-using-type
//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-hlsl -compute -output-using-type
//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-metal -compute -output-using-type
//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-wgpu -compute -output-using-type

// CHECK: 1210.0

//TEST_INPUT:ubuffer(data=[0 0 0], stride=4):out,name=outputBuffer
RWStructuredBuffer<float> outputBuffer;

// Use of IFloat.scale used to cause an internal compiler error
func double_it<T: IFloat>(value: T) -> T {
    return value.scale(2.0);
}

[numthreads(1, 1, 1)]
[shader("compute")]
void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
{
    let mat = matrix<float,4,4>(1,0,0,1, 0,1,1,0, 0,0,1,0, 0,0,0,1);
    let vi = vector<float,4>(100,100,100,100);
    let dm = double_it(mat);
    let dm_sum = dot(dm[0], vi) + dot(dm[1], vi) + dot(dm[2], vi) + dot(dm[3], vi);
    outputBuffer[dispatchThreadID.x] = double_it(float2(1.0, 2.0)).x + double_it(4.0) + dm_sum;
}