gru_flush_cache
gru_flush_cache(ins);
gru_flush_cache(tfm);
gru_flush_cache(cbe);
gru_flush_cache(tfh);
gru_flush_cache(tfh);
gru_flush_cache(tfh);
gru_flush_cache(tfh);
gru_flush_cache(cbk);
gru_flush_cache(tfh);
gru_flush_cache(cbk);
gru_flush_cache(tfh);
gru_flush_cache(cbe); /* CBE not coherent */
gru_flush_cache(h);
gru_flush_cache(cb);
gru_flush_cache(cbe); /* CBE not coherent */
gru_flush_cache(cbe);
gru_flush_cache(gen);
gru_flush_cache(cbe + i * GRU_HANDLE_STRIDE);
gru_flush_cache(cbe + i * GRU_HANDLE_STRIDE);