Dmitry Razdoburdin

Optimization/apply split/column matrix

7

This is a part of https://github.com/dmlc/xgboost/pull/7208 . I tried to separate changes into the independent pull-requests for simplification of the reviewing. In the current one only the changes in ColumnMatrix...

Dev/unified dispatching prototype

1

In continuous of #5659 and #6212. Here I present a way for dispatching the various devices (cpu/cuda device/oneapi device). This request contains only the changes being related to all the...

Optimization/applysplit/common row partitioner

1

Hello all, I continue to work on decomposition of https://github.com/dmlc/xgboost/pull/7208 into smaller and clearer parts. I tried to take into account the feedback from https://github.com/dmlc/xgboost/pull/7991 In the original PR https://github.com/dmlc/xgboost/pull/7208...

Hi, I continue adding of sycl support (https://github.com/dmlc/xgboost/pull/9691, https://github.com/dmlc/xgboost/pull/9846, https://github.com/dmlc/xgboost/pull/9800, https://github.com/dmlc/xgboost/pull/10011, https://github.com/dmlc/xgboost/pull/10045). By this PR I add the next part of the training-related code for the sycl devices. Here I...

SYCL. Add data initialization for training

3

Hi, I continue adding of sycl support (https://github.com/dmlc/xgboost/pull/10216, https://github.com/dmlc/xgboost/pull/10174 etc). By this PR I add the next part of the training-related code for the sycl devices. Here I add data...

SYCL. Add nodes initialisation

1

Hi, I continue adding of sycl support (https://github.com/dmlc/xgboost/pull/10251, https://github.com/dmlc/xgboost/pull/10222 etc). By this PR I add the next part of the training-related code for the sycl devices. Here I add functional...

This PR optimizes histogram building on sycl devices by more suitable parameters choosing.

Hi, this PR make optimize gradients calculations on sycl devices by removing some unnecessary data copying and memory reallocations.