#version 450

layout(local_size_x = 256) in;

layout(set = 0, binding = 0) readonly buffer InputBuffer {
    float input_data[];
};

layout(set = 0, binding = 1) buffer OutputBuffer {
    float output_data[];
};

layout(push_constant) uniform PushConstants {
    uint input_length;
} push;

shared float shared_data[256];

void main() {
    uint gid = gl_GlobalInvocationID.x;
    uint lid = gl_LocalInvocationID.x;
    
    // Load data into shared memory
    shared_data[lid] = 0.0;
    if (gid < push.input_length) {
        shared_data[lid] = input_data[gid];
    }
    
    barrier();
    memoryBarrierShared();
    
    // Reduction in shared memory
    for (uint stride = 128; stride > 0; stride >>= 1) {
        if (lid < stride && lid + stride < 256) {
            shared_data[lid] += shared_data[lid + stride];
        }
        barrier();
        memoryBarrierShared();
    }
    
    // Write result
    if (lid == 0) {
        output_data[gl_WorkGroupID.x] = shared_data[0];
    }
}