aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp
diff options
context:
space:
mode:
authorGeorgios Pinitas <georgios.pinitas@arm.com>2019-01-09 18:35:17 +0000
committerGeorgios Pinitas <georgios.pinitas@arm.com>2019-01-18 13:41:40 +0000
commit7cd26d4a1b14bc4bf7c61496803416ab3d84791f (patch)
tree12cc4a27d7ecebc69a43e96b1f46c7eb05437978 /src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp
parent3ac2f3a1d9297220d1b0ce920dd13fdd4edcc187 (diff)
downloadComputeLibrary-7cd26d4a1b14bc4bf7c61496803416ab3d84791f.tar.gz
COMPMID-1867: Add NEON/SVE GEMM Hybrid kernels.
Change-Id: Ib40a9921e7f9a6a8be6c38872d6b3a0f24ed0cd3 Reviewed-on: https://review.mlplatform.org/515 Reviewed-by: Anthony Barbier <Anthony.barbier@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp')
-rw-r--r--src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp6
1 files changed, 3 insertions, 3 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp b/src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp
index 418a375a61..4ad38cbf62 100644
--- a/src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp
+++ b/src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -32,9 +32,9 @@
// Kernel implementation.
//
// Assume that "Apanel" points to a chunk of A blocks (each size 8xK) in read-order.
-// Assume that "Bpanel" points to a chunk of B blocks (each size 12xK) in read-order.
+// Assume that "Bpanel" points to a chunk of B blocks (each size 24xK) in read-order.
// Assume that "Cpanel" points to a chunk of C output blocks (each size
-// 12x8), the chunks being arranged in a row major fashion.
+// 24x8), the chunks being arranged in a row major fashion.
//
// Note that the intent of this is that either ablocks or bblocks will be 1
// - this construction allows the output loop to proceed in either order.