Integrate new pretranspose_b_array with extra fused transpose of B

This patch fuses the transposition taking place in Acl with the transformations done in arm_gemm (called pretranspose_b_array) if the underlying kernel and transform supports it. This should improve start-up time (as it's for constant Rhs matrices) and memory footprint. The transformations in arm_gemm are kernel specific. The Rhs matrix is transformed into certain layouts to improve the performance.

Resolves: COMPMID-6595

Change-Id: Id2932dd966e59f903c279417bebcea83d9a42464
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11144
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 7 insertions, 2 deletions

diff --git a/src/core/NEON/kernels/arm_gemm/std_transforms_fixed.hpp b/src/core/NEON/kernels/arm_gemm/std_transforms_fixed.hpp
index 4669be9993..a9cbf4ec8d 100644
--- a/src/core/NEON/kernels/arm_gemm/std_transforms_fixed.hpp
+++ b/src/core/NEON/kernels/arm_gemm/std_transforms_fixed.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 Arm Limited.
+ * Copyright (c) 2018-2020, 2024 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -63,9 +63,14 @@ public:
         ConvolutionInterleave<height, block, VLType::None>(out, ptr, stride, conv, rounded_stringlen, y0, ymax, k0, kmax, integrate_sums, row_sum_multiplier);
     }
 
+    bool PrepareB_supports_transpose() const {
+        return false;
+    }
+
     template<typename TIn>
     void PrepareB(TOperand *out, const TIn *in, const int stride, const int x0,
-                  const int xmax, const int k0, const int kmax) const {
+                  const int xmax, const int k0, const int kmax, bool transposed) const {
+        assert(!transposed);
         Transform<width, block,  true>(out, in, stride, x0, xmax, k0, kmax);
     }