If spilling to L1 cache was acceptable, then don't use this feature. The compiler automatically spills your data to memory / the stack. This exists because the

If spilling to L1 cache was acceptable, then don't use this feature. The compiler automatically spills your data to memory / the stack. This exists because the GPU assembly programmer apparently wishes to provide a performance guarantee of some kind.