diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2017-07-04 12:47:17 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-09-17 14:15:39 +0100 |
| commit | 21efeb4491feab09dc246f4da0023d7ca79b1d32 (patch) | |
| tree | 2fbacd7676d13a3ec7024e517acf2e462355275a /src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp | |
| parent | 368da83fdd7406d629e8cca64f3eb0af05437419 (diff) | |
| download | ComputeLibrary-21efeb4491feab09dc246f4da0023d7ca79b1d32.tar.gz | |
COMPMID-417: DepthConvert NEON for QS8/QS16.
Change-Id: Ieb120bccf146045b3a0001ceb3893d4e67fd19df
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79763
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Steven Niu <steven.niu@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp')
| -rw-r--r-- | src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp b/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp index 91fbe6f962..f2cd18d827 100644 --- a/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp @@ -94,7 +94,7 @@ void matrix_addition_f16(const ITensor *input, ITensor *output, const Window &wi void matrix_addition_qs8(const ITensor *input, ITensor *output, const Window &window, float beta) { const int fixed_point_position = input->info()->fixed_point_position(); - const qint8x16_t beta_qs8 = vdupq_n_qs8(scvt_qs8_f32(beta, fixed_point_position)); + const qint8x16_t beta_qs8 = vdupq_n_qs8(sqcvt_qs8_f32(beta, fixed_point_position)); Iterator in(input, window); Iterator out(output, window); @@ -118,7 +118,7 @@ void matrix_addition_qs8(const ITensor *input, ITensor *output, const Window &wi void matrix_addition_qs16(const ITensor *input, ITensor *output, const Window &window, float beta) { const int fixed_point_position = input->info()->fixed_point_position(); - const qint16x8_t beta_qs16 = vdupq_n_qs16(scvt_qs16_f32(beta, fixed_point_position)); + const qint16x8_t beta_qs16 = vdupq_n_qs16(sqcvt_qs16_f32(beta, fixed_point_position)); Iterator in(input, window); Iterator out(output, window); |