Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

More explict sharding methods for device memory (#4396)

Browse Source 
* Rename the Reshard method to Shard

* Add a new Reshard method for sharding a vector that's already sharded
This commit is contained in:
Rong Ou
2019-04-30 16:47:23 -07:00
committed by Rory Mitchell
parent 797ba8e72d
commit eaab364a63
12 changed files with 154 additions and 77 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/cpp/common/test_transform_range.cu
View File

@@ -22,10 +22,10 @@ TEST(Transform, MGPU_Basic) {
GPUDistribution::Block(GPUSet::Empty())};
out_vec.Fill(0);
in_vec.Reshard(GPUDistribution::Granular(devices, 8));
out_vec.Reshard(GPUDistribution::Block(devices));
in_vec.Shard(GPUDistribution::Granular(devices, 8));
out_vec.Shard(GPUDistribution::Block(devices));
// Granularity is different, resharding will throw.
// Granularity is different, sharding will throw.
EXPECT_ANY_THROW(
Transform<>::Init(TestTransformRange<bst_float>{}, Range{0, size}, devices)
.Eval(&out_vec, &in_vec));