diff options
author | Gian Marco <gianmarco.iodice@arm.com> | 2017-11-08 12:24:09 +0000 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
commit | e75a02b60736f37c34388c23c0ccee230f65da59 (patch) | |
tree | f8e9423e40589e99bd8be6c1e740b17792e2058e /arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h | |
parent | 6c0348f4cbf6e30a715780f50aebf6dd0a2a8fc3 (diff) | |
download | ComputeLibrary-e75a02b60736f37c34388c23c0ccee230f65da59.tar.gz |
COMPMID-675 - Reworked NEGEMMLowp interface/function
The new interface makes NEGEMMLowp able to work with ASYMM8 data types.
Implemented 2 new functions:
- NEGEMMLowpMatrixMultiplyCore
- NEGEMMLowpOutputStage
These functions should make the integration in android NN doable
For more information about GEMMLowp:
https://github.com/google/gemmlowp/blob/master/doc/low-precision.md
Change-Id: Ie2c775f45234f68ca53dba644b3a912b997fd890
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/95504
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Diffstat (limited to 'arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h')
-rw-r--r-- | arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h b/arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h index f145eb6ca3..e9bfe4ea07 100644 --- a/arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h +++ b/arm_compute/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.h @@ -58,8 +58,8 @@ public: * The input matrices @p input0 and @p input1 must be the output of the kernels: @ref NEGEMMInterleave4x4Kernel and @ref NEGEMMTranspose1xWKernel. These two * kernels change the layout of the original matrices to be more cache-friendly. * - * @param[in] input0 Input tensor containing the interleaved Matrix A. Data type supported: S8 - * @param[in] input1 Input tensor containing the transposed Matrix B. Data type supported: same as @p input0 + * @param[in] input0 Input tensor containing the interleaved Matrix A. Data type supported: ASYMM8 + * @param[in] input1 Input tensor containing the transposed1xW Matrix B. Data type supported: same as @p input0 * @param[out] output Output tensor to store the result of matrix multiplication. Data type supported: S32 */ void configure(const ITensor *input0, const ITensor *input1, ITensor *output); |