COMPMID-2805: Add QASYMM8_SIGNED support in NEGEMMLowpOutputStage

Add support from requantizing down from S32 to Int8 with fixed point
requantization. This involves the following:
- Compute fixed point multiplication between each entry of input by
  result_fixedpoint_multiplier
- Add bias to final result if bias tensor is not a nullptr
- Round to nearest division by a power-of-two using result_shift
- Add offset to each result
- Clamp the value between the specified min and max bounds
- Cast to int8 data type

Change-Id: I641b3fac0833c568d8565ccb859bbc561a24c17d
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2340
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 14 insertions, 0 deletions

diff --git a/src/runtime/NEON/functions/NEGEMMLowpOutputStage.cpp b/src/runtime/NEON/functions/NEGEMMLowpOutputStage.cpp
index b89e7a168e..3ef9351b78 100644
--- a/src/runtime/NEON/functions/NEGEMMLowpOutputStage.cpp
+++ b/src/runtime/NEON/functions/NEGEMMLowpOutputStage.cpp
@@ -25,6 +25,7 @@
 
 #include "arm_compute/core/ITensor.h"
 #include "arm_compute/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.h"
+#include "arm_compute/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel.h"
 #include "arm_compute/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.h"
 #include "arm_compute/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.h"
 #include "support/ToolchainSupport.h"
@@ -56,6 +57,19 @@ Status NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint::validate(const ITens
     return NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel::validate(input, bias, output, min, max);
 }
 
+void NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPoint::configure(const ITensor *input, const ITensor *bias, ITensor *output, int result_fixedpoint_multiplier, int result_shift,
+                                                                   int result_offset_after_shift, int min, int max)
+{
+    auto k = arm_compute::support::cpp14::make_unique<NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel>();
+    k->configure(input, bias, output, result_fixedpoint_multiplier, result_shift, result_offset_after_shift, min, max);
+    _kernel = std::move(k);
+}
+
+Status NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPoint::validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min, int max)
+{
+    return NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel::validate(input, bias, output, min, max);
+}
+
 void NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPoint::configure(const ITensor *input, const ITensor *bias, ITensor *output, int result_fixedpoint_multiplier, int result_shift, int min, int max)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel>();