d1479f871a
This avoids a lot of huge code changes. Arrays generally cannot be copied in and out of buffers, at least no compiler frontend seems to do it. Also avoids a lot of issues surrounding packed vectors and matrices.
21 lines
776 B
Plaintext
21 lines
776 B
Plaintext
#include <metal_stdlib>
|
|
#include <simd/simd.h>
|
|
|
|
using namespace metal;
|
|
|
|
struct SSBO
|
|
{
|
|
float out_data[1];
|
|
};
|
|
|
|
constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(4u, 4u, 1u);
|
|
|
|
kernel void main0(device SSBO& _67 [[buffer(0)]], uint3 gl_LocalInvocationID [[thread_position_in_threadgroup]], uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]])
|
|
{
|
|
threadgroup float foo[4][4];
|
|
foo[gl_LocalInvocationID.x][gl_LocalInvocationID.y] = float(gl_LocalInvocationIndex);
|
|
threadgroup_barrier(mem_flags::mem_threadgroup);
|
|
_67.out_data[gl_GlobalInvocationID.x] = ((foo[gl_LocalInvocationID.x][0] + foo[gl_LocalInvocationID.x][1]) + foo[gl_LocalInvocationID.x][2]) + foo[gl_LocalInvocationID.x][3];
|
|
}
|
|
|