Add Fused Activation to OpenCL MatMul

- Added fused activation to MatMul function interface
- Added fused activation to CL backend
- Includes tests for supported Activation Functions in MatMul

Resolves: [COMPMID-6192]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: Ie103212b600b60699eaf6a6394d609e6e1f5aba6
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/522465
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9714
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 18 insertions, 16 deletions

diff --git a/src/gpu/cl/operators/ClMatMul.cpp b/src/gpu/cl/operators/ClMatMul.cpp
index 3822c16aa1..c453761a8e 100644
--- a/src/gpu/cl/operators/ClMatMul.cpp
+++ b/src/gpu/cl/operators/ClMatMul.cpp
@@ -47,7 +47,7 @@ ClMatMul::ClMatMul()
 {
 }
 
-Status ClMatMul::validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &matmul_info)
+Status ClMatMul::validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &matmul_info, const ActivationLayerInfo &act_info)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(lhs, rhs, dst);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(lhs, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32);
@@ -57,15 +57,15 @@ Status ClMatMul::validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const
 
     std::unique_ptr<IClMatMulNativeKernelConfig> t = ClMatMulNativeKernelConfigurationFactory::create(gpu_target);
 
-    MatMulKernelInfo kernel_info = t->configure(lhs, rhs, matmul_info);
+    const MatMulKernelInfo kernel_info = t->configure(lhs, rhs, matmul_info);
 
-    bool is_quantized = is_data_type_quantized_asymmetric(lhs->data_type());
+    const bool is_quantized = is_data_type_quantized_asymmetric(lhs->data_type());
 
-    return is_quantized ? ClMatMulLowpNativeKernel::validate(lhs, rhs, dst, kernel_info) :
-                          ClMatMulNativeKernel::validate(lhs, rhs, dst, kernel_info);
+    return is_quantized ? ClMatMulLowpNativeKernel::validate(lhs, rhs, dst, kernel_info, act_info) :
+           ClMatMulNativeKernel::validate(lhs, rhs, dst, kernel_info, act_info);
 }
 
-void ClMatMul::configure(const CLCompileContext &compile_context, ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &matmul_info)
+void ClMatMul::configure(const CLCompileContext &compile_context, ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &matmul_info, const ActivationLayerInfo &act_info)
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(lhs, rhs, dst);
     ARM_COMPUTE_LOG_PARAMS(lhs, rhs, dst, matmul_info);
@@ -86,14 +86,14 @@ void ClMatMul::configure(const CLCompileContext &compile_context, ITensorInfo *l
         _matmul_lowp_native_kernel->set_target(gpu_target);
 
         // Configure the low-precision native matrix multiply kernel
-        _matmul_lowp_native_kernel->configure(compile_context, lhs, rhs, dst, kernel_info);
+        _matmul_lowp_native_kernel->configure(compile_context, lhs, rhs, dst, kernel_info, act_info);
     }
     else
     {
         _matmul_native_kernel->set_target(gpu_target);
 
         // Configure the native matrix multiply kernel
-        _matmul_native_kernel->configure(compile_context, lhs, rhs, dst, kernel_info);
+        _matmul_native_kernel->configure(compile_context, lhs, rhs, dst, kernel_info, act_info);
     }
 }
 
diff --git a/src/gpu/cl/operators/ClMatMul.h b/src/gpu/cl/operators/ClMatMul.h
index 6aba801301..9dce5288e6 100644
--- a/src/gpu/cl/operators/ClMatMul.h
+++ b/src/gpu/cl/operators/ClMatMul.h
@@ -21,14 +21,14 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef ACL_ARM_COMPUTE_SRC_GPU_CL_OPERATORS_CLMATMUL
-#define ACL_ARM_COMPUTE_SRC_GPU_CL_OPERATORS_CLMATMUL
+#ifndef ACL_SRC_GPU_CL_OPERATORS_CLMATMUL
+#define ACL_SRC_GPU_CL_OPERATORS_CLMATMUL
 
 #include "arm_compute/core/ActivationLayerInfo.h"
 #include "arm_compute/core/MatMulInfo.h"
 #include "src/gpu/cl/IClOperator.h"
-#include "src/gpu/cl/kernels/ClMatMulNativeKernel.h"
 #include "src/gpu/cl/kernels/ClMatMulLowpNativeKernel.h"
+#include "src/gpu/cl/kernels/ClMatMulNativeKernel.h"
 
 #include <memory>
 
@@ -71,24 +71,26 @@ public:
      * @param[in]  rhs             Right-hand side tensor info. Data types supported: same as @p lhs.
      * @param[out] dst             Output tensor to store the result of the batched matrix multiplication. Data types supported: same as @p lhs.
      * @param[in]  matmul_info     Contains MatMul operation information described in @ref MatMulInfo.
+     * @param[in]  act_info        Class containing information about fused activation function.
      */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &matmul_info);
+    void configure(const CLCompileContext &compile_context, ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &matmul_info,
+                   const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration
      *
      * Similar to @ref ClMatMul::configure()
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &matmul_info);
+    static Status validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &matmul_info, const ActivationLayerInfo &act_info = ActivationLayerInfo());
     // Inherited methods overridden:
     void run(ITensorPack &tensors) override;
 
 private:
-    std::unique_ptr<kernels::ClMatMulNativeKernel>     _matmul_native_kernel{nullptr};
-    std::unique_ptr<kernels::ClMatMulLowpNativeKernel> _matmul_lowp_native_kernel{nullptr};
+    std::unique_ptr<kernels::ClMatMulNativeKernel>     _matmul_native_kernel{ nullptr };
+    std::unique_ptr<kernels::ClMatMulLowpNativeKernel> _matmul_lowp_native_kernel{ nullptr };
 
     bool _is_quantized{ false };
 };
 } // namespace opencl
 } // namespace arm_compute
-#endif /* ACL_ARM_COMPUTE_SRC_GPU_CL_OPERATORS_CLMATMUL */
+#endif /* ACL_SRC_GPU_CL_OPERATORS_CLMATMUL */