diff options
Diffstat (limited to 'src/core/CL/cl_kernels/common/gemmlowp.cl')
-rw-r--r-- | src/core/CL/cl_kernels/common/gemmlowp.cl | 18 |
1 files changed, 9 insertions, 9 deletions
diff --git a/src/core/CL/cl_kernels/common/gemmlowp.cl b/src/core/CL/cl_kernels/common/gemmlowp.cl index f9d18ec976..9fad66df68 100644 --- a/src/core/CL/cl_kernels/common/gemmlowp.cl +++ b/src/core/CL/cl_kernels/common/gemmlowp.cl @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2021 Arm Limited. + * Copyright (c) 2017-2022 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -1096,17 +1096,17 @@ __kernel void gemmlowp_matrix_a_reduction_dot8(TENSOR3D_DECLARATION(src), VEC_DATA_TYPE(DATA_TYPE, 16) a0 = vload16(0, matrix_a + i); - sum_row += arm_dot(a0.s0123, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.s4567, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.s89AB, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.sCDEF, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s0123, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s4567, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s89AB, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.sCDEF, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); a0 = vload16(1, matrix_a + i); - sum_row += arm_dot(a0.s0123, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.s4567, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.s89AB, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); - sum_row += arm_dot(a0.sCDEF, (VEC_DATA_TYPE(DATA_TYPE, 4))(1)); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s0123, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s4567, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.s89AB, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); + DOT_PRODUCT4_INTEGER8(DATA_TYPE, DATA_TYPE, DATA_TYPE, a0.sCDEF, (VEC_DATA_TYPE(DATA_TYPE, 4))(1), sum_row); } // This for loop performs the leftover accumulations |