tree 1cb3a38a4cce1a0514cfafa80727c5dadfee4efb
parent 74b04b7099d41e6cf71ee5b54ab6cc0a45e425a3
author Thomas <thomasraoux@google.com> 1646256216 -0800
committer GitHub <noreply@github.com> 1646256216 -0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJiH+BYCRBK7hj4Ov3rIwAAfb0IAHWRd1EdhsEz7yOINvDustst
 nJxoidnS9ZyfSRx6wlIW1BSFGgkydHNa85B4J5IlmYEFsxOeHvtZDbBUtYdi0PAZ
 n1L08HU8vbLMC0Blqxuc4KNbGt5Zs7n47TMfb7T5k1bHqNRvMIOUUjTcmijUnsks
 xjfgdG2pD9FIT7GmG6SZLFCCae3u/o6nOVw9evNHPheuD+DVLXDNbz09BtDPa8MW
 E9lEUFwEfGF3OB5iRCE471UqHp+P1i9k8WQM8h+Kzbyn0ouXehdAkjaR30vqBRsd
 n1GDQbkJgvODGYSJ1Se/6cfZm1UnXYnBaNKzCdfg/ZFTzxvhEzYx2gPdGvi2Gso=
 =WmZ9
 -----END PGP SIGNATURE-----
 

[CUDA] Fix thread tile size calculation for unaligned cases (#8438)

This prevent skipping tiling for cases where the workgroup tile size is
smaller than workgroup size.