COMPMID-759 - CLGEMM optimization for McVail benchmarks

This patch introduces an optimization for CLGEMM on Bifrost
architectures which can bring to 40% of FMA utilization on
config 3 of McVail. The new CLGEMM does not require any reshape of
matrix A and matrix B.

This patch also adds the auto-config in CLConvolutionLayer and CLGEMM
and extends the interface for NEGEMM and CLGEMM.

Change-Id: Ibb354eda45e9ca64b14a99700fb21dff5989dda9
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/113716
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

2 files changed, 12 insertions, 9 deletions

diff --git a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
index a8a04a0bbf..3fe6604db9 100644
--- a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
@@ -138,10 +138,9 @@ private:
     CLTensor _gemm_output;
     CLTensor _tmp_output;
 
-    bool _append_bias;
-    bool _is_fully_connected_convolution;
     bool _are_weights_reshaped;
     bool _is_quantized;
+    bool _is_interleaved_transposed;
 };
 }
 #endif /* __ARM_COMPUTE_CLCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/CL/functions/CLGEMM.h b/arm_compute/runtime/CL/functions/CLGEMM.h
index 2765b77b7d..bf41226bda 100644
--- a/arm_compute/runtime/CL/functions/CLGEMM.h
+++ b/arm_compute/runtime/CL/functions/CLGEMM.h
@@ -61,14 +61,16 @@ public:
      *
      * @note Whilst the first input tensor can be a vector, the second input tensor must be at least a matrix
      *
-     * @param[in]  a      First input tensor  (Matrix or Vector A). Data types supported: QS8/QS16/F16/F32
-     * @param[in]  b      Second input tensor (Matrix B). Data type supported: same as @p a.
-     * @param[in]  c      Third input tensor  (Matrix C). It can be a nullptr if just the multiplication between @p a and @p b is needed. Data type supported: same as @p a.
-     * @param[out] output Output tensor. Data type supported: same as @p a
-     * @param[in]  alpha  Weight of the matrix product
-     * @param[in]  beta   Weight of matrix C
+     * @param[in]  a         First input tensor  (Matrix or Vector A). Data types supported: QS8/QS16/F16/F32
+     * @param[in]  b         Second input tensor (Matrix B). Data type supported: same as @p a.
+     * @param[in]  c         Third input tensor  (Matrix C). It can be a nullptr if just the multiplication between @p a and @p b is needed. Data type supported: same as @p a.
+     * @param[out] output    Output tensor. Data type supported: same as @p a
+     * @param[in]  alpha     Weight of the matrix product
+     * @param[in]  beta      Weight of matrix C
+     * @param[in]  gemm_info (Optional) Specifies if the matrix A and/or matrix B have been reshaped and
+     *                       if the reshape of matrix B should happen only for the first run
      */
-    void configure(const ICLTensor *a, const ICLTensor *b, const ICLTensor *c, ICLTensor *output, float alpha, float beta);
+    void configure(const ICLTensor *a, const ICLTensor *b, const ICLTensor *c, ICLTensor *output, float alpha, float beta, const GEMMInfo &gemm_info = GEMMInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -83,6 +85,8 @@ private:
     CLTensor                   _tmp_b;
     bool                       _is_interleaved_transposed;
     bool                       _run_addition;
+    bool                       _is_first_run;
+    bool                       _reshape_b_only_on_first_run;
 };
 }