From a668f9f8a4eab405df0fe8dd58e7d9425bcf9640 Mon Sep 17 00:00:00 2001
From: Jonathan Deakin <jonathan.deakin@arm.com>
Date: Wed, 24 Jan 2024 09:15:38 +0000
Subject: Add s8f32 kernels and dynamic QuantizationInfo

- Add support for QASYMM_SIGNED*QASYMM8_SIGNED->F32 in
  CpuGemmLowpMatrixMultiplyCore
- Add s8f32 kernel using existing s8->s32 kernels with a new
  DequantizeFloat OutputStage, the structure is similar to Requantize32
  but the opposite way around.
- Add SME s8f32 kernels with integrated support for DequantizeFloat.
- Add scale to CpuGemmLowpOffsetContributionKernel.
- Add virtual dequantize scale to gemm_common, only implemented for
  gemm_interleaved.
- Update year to 2024 in generate_build_files.
- Add dynamic flag to QuantizationInfo which signals to operators that
  it can change after configuration
- Add support for dynamic quantization in NEGEMMLowpMatrixMultiplyCore
- Add dynamic quantization fixture by extending
  GEMMLowpGenericMatrixMultiplyCoreValidationFixture
- Add GEMMLowpDequantizedMatrixMultiplyValidationFixture
- Store k (number of cols of A) rather than k_offset in the offset
  contribution kernels so that we can recompute it when the other
  offsets change

relates to: ONCPUML-1444 MLINFSW-439

Co-authored-by: Milos Puzovic <Milos.Puzovic@arm.com>
Co-authored-by: David Mansell <David.Mansell@arm.com>
Change-Id: I58a3acf2c09289a303e52eea6b336a696a5bc8da
Signed-off-by: Jonathan Deakin <jonathan.deakin@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11022
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
---
 src/cpu/kernels/assembly/arm_gemm.hpp    | 13 +++++++++++++
 src/cpu/kernels/assembly/gemm_common.hpp |  6 ++++++
 2 files changed, 19 insertions(+)

(limited to 'src/cpu/kernels/assembly')

diff --git a/src/cpu/kernels/assembly/arm_gemm.hpp b/src/cpu/kernels/assembly/arm_gemm.hpp
index 5d7cf79857..941fed0ba8 100644
--- a/src/cpu/kernels/assembly/arm_gemm.hpp
+++ b/src/cpu/kernels/assembly/arm_gemm.hpp
@@ -260,6 +260,19 @@ public:
     }
 };
 
+struct DequantizeFloat
+{
+public:
+    float scale = 0;
+
+    DequantizeFloat() = default;
+
+    // Constructor
+    DequantizeFloat(const float scale) : scale(scale)
+    {
+    }
+};
+
 struct Nothing
 {
 };
diff --git a/src/cpu/kernels/assembly/gemm_common.hpp b/src/cpu/kernels/assembly/gemm_common.hpp
index 4825814e31..45d1e43274 100644
--- a/src/cpu/kernels/assembly/gemm_common.hpp
+++ b/src/cpu/kernels/assembly/gemm_common.hpp
@@ -166,6 +166,12 @@ public:
     {
     }
 
+    /*** Dequanize scale interface (optional) ***/
+    /* Set the dequantize scale for GEMMs when converting from int to float (float out = scale * float(int out) ) */
+    virtual void set_dequantize_scale(const float)
+    {
+    }
+
     /*** Introspection interface ***/
     /* Get the configuration of this GEMM */
     virtual GemmConfig get_config() = 0;
-- 
cgit v1.2.1