Add launch bounds needed for ROCm 2.0 (#15400)
authorJohannes M Dieterich <johannes.dieterich@amd.com>
Thu, 20 Dec 2018 22:26:14 +0000 (14:26 -0800)
committerFacebook Github Bot <facebook-github-bot@users.noreply.github.com>
Thu, 20 Dec 2018 22:39:13 +0000 (14:39 -0800)
commitc24a124fa0df4199d9180cc009c5b5e8ff88bba3
tree708324f0fcddc467fc80e6465e2d8cf971875f19
parent1a2ec10bd4a73f4ebeefeb7a7cc781a80b14794c
Add launch bounds needed for ROCm 2.0 (#15400)

Summary:
ROCm 2.0's compiler requires launch_bounds annotations if flat work group sizes are larger than the default of 256.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15400

Differential Revision: D13531239

Pulled By: ezyang

fbshipit-source-id: c0b40600a8c332823da6c7113c644d8dba424a9c
aten/src/ATen/native/cuda/LossCTC.cu
aten/src/THC/THCReduceAll.cuh
aten/src/THCUNN/MultiLabelMarginCriterion.cu
aten/src/THCUNN/SpatialClassNLLCriterion.cu