[CUDA] Support multiple TIR-level dynamic shared memory allocations#8571

Merged

vinx13 merged 21 commits intoapache:mainfrom masahi:dyn_shmem_merge

Jul 31, 2021

+430-1

Commits on Jul 27, 2021

Commits on Jul 28, 2021

handle vector alloc
masa
committed
add vectorized test
masa
committed
drop multi-lane dtype allocation support, vectorized store working
masa
committed
doc update
masa
committed
dtype fix in the test
masa
committed
lint fix, do not run merging when number of alloc is 1
masa
committed

Commits on Jul 29, 2021

Commits on Jul 30, 2021

add test on combined buffer reuse and merge
masahi
committed