diff options
author | Cedric Nugteren <web@cedricnugteren.nl> | 2019-12-15 09:33:53 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-12-15 09:33:53 +0100 |
commit | 8a19667e75017a255aeb6addc32815a0ec46b30f (patch) | |
tree | a75de219a5393a848fa083f59f7add5fcc22aa42 /src/routines/levelx/xgemmbatched.cpp | |
parent | 6ac74008b6c676ab6834f1c758670c938a0529c3 (diff) | |
parent | 21b66ca76140be9ac30811e7648abe3837e19177 (diff) |
Merge pull request #372 from trantila/master
Reduced number of TestMatrix calls for the batched xgemm routines.
Diffstat (limited to 'src/routines/levelx/xgemmbatched.cpp')
-rw-r--r-- | src/routines/levelx/xgemmbatched.cpp | 8 |
1 files changed, 3 insertions, 5 deletions
diff --git a/src/routines/levelx/xgemmbatched.cpp b/src/routines/levelx/xgemmbatched.cpp index b12b8734..0d7ae5ab 100644 --- a/src/routines/levelx/xgemmbatched.cpp +++ b/src/routines/levelx/xgemmbatched.cpp @@ -79,11 +79,9 @@ void XgemmBatched<T>::DoGemmBatched(const Layout layout, const Transpose a_trans gemm_kernel_id); // Tests the matrices for validity - for (auto batch = size_t{0}; batch < batch_count; ++batch) { - TestMatrixA(a_one, a_two, a_buffer, a_offsets[batch], a_ld, false); // don't test for invalid LD - TestMatrixB(b_one, b_two, b_buffer, b_offsets[batch], b_ld, false); // don't test for invalid LD - TestMatrixC(c_one, c_two, c_buffer, c_offsets[batch], c_ld); - } + TestBatchedMatrixA(a_one, a_two, a_buffer, a_offsets, a_ld, false); // don't test for invalid LD + TestBatchedMatrixB(b_one, b_two, b_buffer, b_offsets, b_ld, false); // don't test for invalid LD + TestBatchedMatrixC(c_one, c_two, c_buffer, c_offsets, c_ld); // Upload the scalar arguments to the device auto alphas_device = Buffer<T>(context_, BufferAccess::kReadWrite, batch_count); |