tree e7e7663fe1d976ae736929c052a90a0050e32a59
parent c88427b14d167b4559920bc817429e8e17a0b4be
author Kunwar Grover <groverkss@gmail.com> 1706559479 +0530
committer GitHub <noreply@github.com> 1706559479 +0000
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJluAf3CRC1aQ7uu5UhlAAAl64QACAUC+5RgNV9F/ETd69Ha0Ot
 dDvoGZ+cf0Ma65DHX6D+iZa4UbzaZJVIDVv732MgPb7RtWLzIp5anJu0oXI5eKGk
 SYYAr6FEm5Y3wCW4tClJZXGlqUDOODeu1JRIqpoU8Qj/aU2uyksoZX2QtHcRof7v
 m96CDIwxrjvXhEbKRBZa29x0Wa3zrOKd9jwEGhiyU77c7+W98MoS0YGmTeYRa9/1
 eeEVFmdFKBUbs8kRBBWZ1MbWaaSmUm5WzrAWh5iXHXV6Ro++5W5FsYuRIbJjQyKa
 9rBdR3YdbVcXFiPyA4cTXrJhEV8PZ2tiRTziHRI/8RBKvSD6bfgZqV3ZPnC+8mZk
 iYXX7HWnGYNCLsqvsUcbILicP+ucWTtRsg6ulDBBBZaIj5lrS+fq1aMhVl0IaXNs
 oS8KPQX8i1L6gca/9LAZ6gGM4v90BLh6SluFY+HKM/ebi26gl+A9VtiPNRYd+3ST
 v6Yl402B7Q8NrIxS87hFvsa/SYgoBiom/TTxAjvd60vJQhnAVK1OVKmlv/iJCM+X
 R1H6M9w0PO/EI2CK2i/Q090BC0/aYSsZi5v78RNSOVTJ7zgI0EBfAWbn5BNv0Sml
 jyfDmZV/gpmxNbA5CKl4DB/smVM4KuJ6VCsSHLw/HOnypxSYhofWlMYvzSiNBh1k
 PW5kSZk3AmeuLOUS09c3
 =V1is
 -----END PGP SIGNATURE-----
 

[Codegen][LLVMGPU] Add option to control sync after map_nested_for_all_to_gpu_threads (#16247)

map_nested_for_all_to_gpu_threads needs to synchronize because the the
number of threads being distributed may not be equal to workgroup size.
This patch adds an option to disable these barriers, which is useful in
cases when the number of threads are equal to workgroup size.