jax.experimental.pallas.mosaic_gpu.commit_tmem#

jax.experimental.pallas.mosaic_gpu.commit_tmem()[source]#

Commits all writes to TMEM issued by the current thread.

Once this function returns, the effects of calling async_store_tmem from the current thread are visible to TMEM loads, MMA and barrier operations of Barrier``s with ``orders_tensor_core=True.