summaryrefslogtreecommitdiff
path: root/src/tuning
diff options
context:
space:
mode:
authorCedric Nugteren <web@cedricnugteren.nl>2017-10-03 21:55:21 +0200
committerCedric Nugteren <web@cedricnugteren.nl>2017-10-03 21:55:21 +0200
commit375193fe4e72b320eb63fbc6f98c24714f6970c2 (patch)
treee01cf66d16e0a0e7d024e6ce3ba24fd85803f1fd /src/tuning
parent74fd6767b93b03fc62462f44854215c4c320babe (diff)
Gemm in-direct implementation now uses only 1 larger instead of max 3 optional temporary buffers
Diffstat (limited to 'src/tuning')
-rw-r--r--src/tuning/kernels/xgemm.cpp2
1 files changed, 2 insertions, 0 deletions
diff --git a/src/tuning/kernels/xgemm.cpp b/src/tuning/kernels/xgemm.cpp
index 7d0f3ed4..6dcdf68b 100644
--- a/src/tuning/kernels/xgemm.cpp
+++ b/src/tuning/kernels/xgemm.cpp
@@ -180,6 +180,8 @@ class TuneXgemm {
tuner.AddArgumentInput(a_mat);
tuner.AddArgumentInput(b_mat);
tuner.AddArgumentOutput(c_mat);
+ tuner.AddArgumentScalar(0);
+ tuner.AddArgumentScalar(0);
}
};