ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp

#version 450

#extension GL_EXT_control_flow_attributes : enable

#include "types.comp"
#include "generic_head.comp"

layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;

layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
layout (binding = 1) readonly buffer Y {B_TYPE data_b[];};
layout (binding = 2) buffer D {D_TYPE data_d[];};

const uint CHUNK_SIZE = 512;

void main() {
    const uint base = gl_WorkGroupID.x * CHUNK_SIZE;
    const uint col = gl_LocalInvocationID.x;

    uint count = 0;
    [[unroll]]
    for (uint i = 0; i < CHUNK_SIZE; i += gl_WorkGroupSize.x) {
        const uint idx = base + i + col;
        if (idx >= p.KX) {
            break;
        }
        count += uint(data_a[idx] == data_b[idx]);
    }

    atomicAdd(data_d[0], D_TYPE(count));
}