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

Improve operation efficiency for single predict (#5016)

Browse Source 
* Improve operation efficiency for single predict
This commit is contained in:
KaiJin Ji 2019-11-10 02:01:28 +08:00 committed by Jiaming Yuan
parent 374648c21a
commit 1733c9e8f7
2 changed files with 39 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/gbm/gbtree.cc
View File

@ -408,6 +408,7 @@ class Dart : public GBTree {
constexpr int kUnroll = 8; constexpr int kUnroll = 8;
const auto nsize = static_cast<bst_omp_uint>(batch.Size()); const auto nsize = static_cast<bst_omp_uint>(batch.Size());
const bst_omp_uint rest = nsize % kUnroll; const bst_omp_uint rest = nsize % kUnroll;
if (nsize >= kUnroll) {
#pragma omp parallel for schedule(static) #pragma omp parallel for schedule(static)
for (bst_omp_uint i = 0; i < nsize - rest; i += kUnroll) { for (bst_omp_uint i = 0; i < nsize - rest; i += kUnroll) {
const int tid = omp_get_thread_num(); const int tid = omp_get_thread_num();
@ -429,6 +430,7 @@ class Dart : public GBTree {
} }
} }
} }
}
for (bst_omp_uint i = nsize - rest; i < nsize; ++i) { for (bst_omp_uint i = nsize - rest; i < nsize; ++i) {
RegTree::FVec& feats = thread_temp_[0]; RegTree::FVec& feats = thread_temp_[0];
const auto ridx = static_cast<int64_t>(batch.base_rowid + i); const auto ridx = static_cast<int64_t>(batch.base_rowid + i);

2
src/predictor/cpu_predictor.cc
View File

@ -63,6 +63,7 @@ class CPUPredictor : public Predictor {
// Pull to host before entering omp block, as this is not thread safe. // Pull to host before entering omp block, as this is not thread safe.
batch.data.HostVector(); batch.data.HostVector();
batch.offset.HostVector(); batch.offset.HostVector();
if (nsize >= kUnroll) {
#pragma omp parallel for schedule(static) #pragma omp parallel for schedule(static)
for (bst_omp_uint i = 0; i < nsize - rest; i += kUnroll) { for (bst_omp_uint i = 0; i < nsize - rest; i += kUnroll) {
const int tid = omp_get_thread_num(); const int tid = omp_get_thread_num();
@ -84,6 +85,7 @@ class CPUPredictor : public Predictor {
} }
} }
} }
}
for (bst_omp_uint i = nsize - rest; i < nsize; ++i) { for (bst_omp_uint i = nsize - rest; i < nsize; ++i) {
RegTree::FVec& feats = thread_temp[0]; RegTree::FVec& feats = thread_temp[0];
const auto ridx = static_cast<int64_t>(batch.base_rowid + i); const auto ridx = static_cast<int64_t>(batch.base_rowid + i);