diff options
author | Pablo Tello <pablo.tello@arm.com> | 2017-09-29 16:43:25 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
commit | bf2fb95c99ebd215b3c0d93cb970461185ef9716 (patch) | |
tree | ef9ea161a5b4bf04d057681eb435605f3d1fa5ab /tests/validation/CPP/GEMMLowp.cpp | |
parent | dd715f2a88827241a3fb9e4a2d8be82455f649f7 (diff) | |
download | ComputeLibrary-bf2fb95c99ebd215b3c0d93cb970461185ef9716.tar.gz |
COMPMID-481: Add gemmlowp_aarch64_v8p4 kernel.
Change-Id: I15496b16ffd636f5bff76572e750df7e15c80830
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/90532
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Diffstat (limited to 'tests/validation/CPP/GEMMLowp.cpp')
-rw-r--r-- | tests/validation/CPP/GEMMLowp.cpp | 36 |
1 files changed, 36 insertions, 0 deletions
diff --git a/tests/validation/CPP/GEMMLowp.cpp b/tests/validation/CPP/GEMMLowp.cpp index d172a773b6..06926e631e 100644 --- a/tests/validation/CPP/GEMMLowp.cpp +++ b/tests/validation/CPP/GEMMLowp.cpp @@ -34,6 +34,42 @@ namespace validation { namespace reference { +SimpleTensor<uint32_t> gemmlowp(const SimpleTensor<uint8_t> &a, const SimpleTensor<uint8_t> &b, SimpleTensor<uint32_t> &c) +{ + ARM_COMPUTE_UNUSED(a); + ARM_COMPUTE_UNUSED(b); + ARM_COMPUTE_UNUSED(c); + const int K = a.shape().x(); + const int b_width = b.shape().x(); + const int rows = c.shape().y(); //M + const int cols = c.shape().x(); //N + std::vector<int32_t> acc; + acc.resize(cols); + for(int i = 0; i < rows; ++i) + { + for(int j = 0; j < cols; ++j) + { + acc[j] = 0; + } + for(int k = 0; k < K; ++k) + { + auto tmp_a = static_cast<int32_t>(a[k + i * K]); + for(int j = 0; j < b_width; ++j) + { + auto tmp_b = static_cast<int32_t>(b[j + k * b_width]); + const int32_t mult_as_int = tmp_a * tmp_b; + acc[j] += mult_as_int; + } + } + for(int j = 0; j < cols; ++j) + { + c[j + i * cols] = acc[j]; + } + } + + return c; +} + template <typename T> SimpleTensor<T> gemmlowp(const SimpleTensor<T> &a, const SimpleTensor<T> &b, SimpleTensor<T> &c, int32_t a_offset, int32_t b_offset, int32_t c_offset, int32_t c_mult_int, int32_t out_shift) |