Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

INTEL MKL: Fixing a unit test failure, for direct session memory allocation #18240

Merged
merged 9 commits into from
May 18, 2018
Merged

INTEL MKL: Fixing a unit test failure, for direct session memory allocation #18240

Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
e7f3ed2
Fixing a unit test failure for INTEL MKL where memeory allocation ch…
ashraf-bhuiyan Apr 4, 2018
a2cba4a
if MKL is used allocation id is set to 9 and 10
ashraf-bhuiyan May 4, 2018
4a6ca8f
adding MKLDNN switch only for parameters
ashraf-bhuiyan May 9, 2018
d7596f5
Fixing a unit test failure for INTEL MKL where memeory allocation ch…
ashraf-bhuiyan Apr 4, 2018
ee78a3b
if MKL is used allocation id is set to 9 and 10
ashraf-bhuiyan May 4, 2018
d27e562
rebasing with master and removing the conflict
ashraf-bhuiyan May 10, 2018
460dfbd
merging with master and fixing the conflict
ashraf-bhuiyan May 10, 2018
7a667f6
updated the allocation ID in the direct sesssion file
ashraf-bhuiyan May 16, 2018
2acf231
resolving the conflict while merging master
ashraf-bhuiyan May 16, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
16 changes: 16 additions & 0 deletions tensorflow/core/common_runtime/direct_session_with_tracking_alloc_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,9 +102,25 @@ TEST(DirectSessionWithTrackingAllocTest, CostModelTest) {
EXPECT_EQ(2, shape.dim(0).size());
EXPECT_EQ(1, shape.dim(1).size());
if (node->name() == y->name()) {
#ifdef INTEL_MKL
// if MKL is used, it goes through various additional
// graph rewrite pass. In TF, everytime a graph pass
// happens, "constant" nodes are allocated
// and deallocated. Each allocation calls the
// (FindChunkPtr of BFCAllocator),
// which increments the value of AllocationId.
// Thus AllocationId becomes more than 3 and 4 if
// MKL is used. Now they are 9 and 10 for MKL.
EXPECT_EQ(19, cm->AllocationId(node, 0));
#else
EXPECT_EQ(13, cm->AllocationId(node, 0));
#endif
} else {
#ifdef INTEL_MKL
EXPECT_EQ(20, cm->AllocationId(node, 0));
#else
EXPECT_EQ(14, cm->AllocationId(node, 0));
#endif
}
}
EXPECT_LE(0, cm->MaxExecutionTime(node));
Expand Down