tree 38638875e35d78429b735af5a85ff6c943a993cd
parent 7e5fa972c57378fc35cbc8892444a6d7ce554378
author Thomas <thomasraoux@google.com> 1629238288 -0700
committer GitHub <noreply@github.com> 1629238288 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJhHDQQCRBK7hj4Ov3rIwAAdnQIACsVXByYTSMfT/CJg7gHaXHK
 auHUB6Ru9nPR+4UyyfU7HFDnDfR2B4BtSgvXlmK0OFsy7mwaS0akAZsdJrKTJld+
 xumMosMmIKHEizq4/PMu4Y0stZA+xQ4JCgADpdfrMiXjzl6TEpGRUhm/6e0+bMRz
 x360tTNaPHuMcyPraEZXRieXNL2Fqh8r71viAOhvJEIXqnwtW+hsgZHO8xiJh1Rd
 2V1jO25GfJjjo6ziZ4leJUizkxLzTHZV+RUt1IqlK32GsVFOFXSv0TIkIAyu86EF
 fUD2xyotezM/+ZeT1l0ezjoSbFg/9kbkr3GIHivKea6NxfWu0elJEgH/FEmpd2A=
 =YTU4
 -----END PGP SIGNATURE-----
 

[CUDA] Add some basic logic to pick better configuration (#6754)

Pick configuration based on the problem size and skip using shared
memory for workgroup of a single warp.