From 181e65145d153210ec5587a42d2938e27e1d5b01 Mon Sep 17 00:00:00 2001
From: Pablo Tello <pablo.tello@arm.com>
Date: Wed, 15 Nov 2017 13:28:27 +0000
Subject: COMPMID-675: NEGEMMLowp Assembly Integration

Added support for S8 input in NEGEMMLowp Matrix Multiply Kernel.
Added a new function to run assembly kernels such that A*B=C (no offsets involved)
Added new tests for the assembly gemmlowp kernels (no offsets)
Integrated the assembly kernel for the A57

Change-Id: Ib3e39c1f3f7f1baa0d39be69485f61cd18e3c9b3
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/95864
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
---
 tests/validation/CPP/GEMMLowp.cpp | 6 ++++++
 tests/validation/CPP/GEMMLowp.h   | 3 +++
 2 files changed, 9 insertions(+)

(limited to 'tests/validation/CPP')

diff --git a/tests/validation/CPP/GEMMLowp.cpp b/tests/validation/CPP/GEMMLowp.cpp
index bac3a20c8e..8670a22a66 100644
--- a/tests/validation/CPP/GEMMLowp.cpp
+++ b/tests/validation/CPP/GEMMLowp.cpp
@@ -73,6 +73,12 @@ SimpleTensor<int32_t> gemmlowp_matrix_multiply_core(const SimpleTensor<T> &a, co
     return c;
 }
 
+// used to validate assembly kernels which don't know anything about offsets
+SimpleTensor<int32_t> gemmlowp(const SimpleTensor<int8_t> &a, const SimpleTensor<int8_t> &b)
+{
+    return gemmlowp_matrix_multiply_core(a, b, 0, 0);
+}
+
 template <typename T>
 SimpleTensor<uint8_t> gemmlowp_quantize_down_int32_to_uint8_scale(const SimpleTensor<T> &in, int32_t result_offset, int32_t result_mult_int, int32_t result_shift)
 {
diff --git a/tests/validation/CPP/GEMMLowp.h b/tests/validation/CPP/GEMMLowp.h
index c09d8f6176..cbed2206e3 100644
--- a/tests/validation/CPP/GEMMLowp.h
+++ b/tests/validation/CPP/GEMMLowp.h
@@ -40,6 +40,9 @@ SimpleTensor<int32_t> gemmlowp_matrix_multiply_core(const SimpleTensor<T> &a, co
 
 template <typename T>
 SimpleTensor<uint8_t> gemmlowp_quantize_down_int32_to_uint8_scale(const SimpleTensor<T> &in, int32_t result_offset, int32_t result_mult_int, int32_t result_shift);
+
+SimpleTensor<int32_t> gemmlowp(const SimpleTensor<int8_t> &a, const SimpleTensor<int8_t> &b);
+
 } // namespace reference
 } // namespace validation
 } // namespace test
-- 
cgit v1.2.1