aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/assembly
diff options
context:
space:
mode:
authorMichele Di Giorgio <michele.digiorgio@arm.com>2018-09-10 18:22:03 +0100
committerAnthony Barbier <anthony.barbier@arm.com>2018-11-02 16:54:54 +0000
commita53a7bb46cdd5cdb6ef14703e0be00555c710a45 (patch)
tree7aeea9e3db0b38ab252fd976438965259940b972 /src/core/NEON/kernels/assembly
parent000d33a0e4bfc129a8f2968d4e5ee0793df70a1e (diff)
downloadComputeLibrary-a53a7bb46cdd5cdb6ef14703e0be00555c710a45.tar.gz
[MLCE-54] graph_alexnet example throw an exception
The 32bit version of AlexNet seems to require ~1.4 GB of memory to run (got this result via valgrind's massif tool). The problem is a miscalculation of the shape of the _transformed_b tensor, which should not be multiplied by the size of the data type, but rather represent only the size of the dimensions. Change-Id: I55dfb538b31772afc4ce0ce0ccd31fdeb1585cfb Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/147666 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com> Tested-by: Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/assembly')
-rw-r--r--src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp1
1 files changed, 0 insertions, 1 deletions
diff --git a/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp b/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp
index f33a14f2af..6c201cedb3 100644
--- a/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp
+++ b/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp
@@ -89,7 +89,6 @@ unsigned int get_B_pretransposed_array_size(unsigned int N, unsigned int K, cons
// Calculate the total size of the buffer:
size_t total = num_full_k * normal_k_size * (num_full_x * normal_x_size + left_over_x_size);
total += left_over_k_size * (left_over_x_size + num_full_x * normal_x_size);
- total *= sizeof(To);
return total;
}