Add Subtraction operator to Dynamic Fusion interface

Partially-Resolves: COMPMID-5518

Change-Id: I8358784815bcac461d50e384fa7bc96f476d3983
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9045
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Dynamic-Fusion: SiCong Li <sicong.li@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

3 files changed, 80 insertions, 0 deletions

diff --git a/src/dynamic_fusion/sketch/gpu/components/cl/ClComponentElementwiseBinary.cpp b/src/dynamic_fusion/sketch/gpu/components/cl/ClComponentElementwiseBinary.cpp
index 2611d6d575..b21c7c382f 100644
--- a/src/dynamic_fusion/sketch/gpu/components/cl/ClComponentElementwiseBinary.cpp
+++ b/src/dynamic_fusion/sketch/gpu/components/cl/ClComponentElementwiseBinary.cpp
@@ -38,6 +38,7 @@ namespace
 std::set<ElementwiseBinaryCommonAttributes::ElementwiseOp> supported_ops
 {
     ElementwiseBinaryCommonAttributes::ElementwiseOp::Add,
+    ElementwiseBinaryCommonAttributes::ElementwiseOp::Sub,
     ElementwiseBinaryCommonAttributes::ElementwiseOp::Mul
 };
 }
diff --git a/src/dynamic_fusion/sketch/gpu/operators/GpuSub.cpp b/src/dynamic_fusion/sketch/gpu/operators/GpuSub.cpp
new file mode 100644
index 0000000000..8240008f2a
--- /dev/null
+++ b/src/dynamic_fusion/sketch/gpu/operators/GpuSub.cpp
@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2023 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/dynamic_fusion/sketch/gpu/operators/GpuSub.h"
+#include "arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h"
+
+#include "src/dynamic_fusion/sketch/gpu/operators/internal/GpuElementwiseBinaryCommon.h"
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+Status GpuSub::validate_op(const GpuWorkloadSketch &sketch,
+                           const ITensorInfo       *lhs,
+                           const ITensorInfo       *rhs)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(lhs, rhs);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(lhs, 1, DataType::F16, DataType::F32, DataType::U8, DataType::S16, DataType::S32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs->data_type() != rhs->data_type(), "Input tensors must be the same data type");
+
+    // Set the elementwise operation to Sub then call the elementwise common validate_op
+    ElementwiseBinaryCommonAttributes common_attributes{};
+    common_attributes.operation(ElementwiseBinaryCommonAttributes::ElementwiseOp::Sub);
+    return GpuElementwiseBinaryCommon::validate_op(sketch, lhs, rhs, common_attributes);
+}
+
+Status GpuSub::is_supported_op(const GpuWorkloadContext &context,
+                               const ITensorInfo        *lhs,
+                               const ITensorInfo        *rhs)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(lhs, rhs);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(lhs, 1, DataType::F16, DataType::F32, DataType::U8, DataType::S16, DataType::S32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs->data_type() != rhs->data_type(), "Input tensors must be the same data type");
+
+    // Set the elementwise operation to Sub then call the elementwise common is_supported_op
+    ElementwiseBinaryCommonAttributes common_attributes{};
+    common_attributes.operation(ElementwiseBinaryCommonAttributes::ElementwiseOp::Sub);
+    return GpuElementwiseBinaryCommon::is_supported_op(context, lhs, rhs, common_attributes);
+}
+
+ITensorInfo *GpuSub::create_op(GpuWorkloadSketch &sketch,
+                               ITensorInfo       *lhs,
+                               ITensorInfo       *rhs)
+{
+    // No need to log or validate as they'll be handled inside GpuElementwiseBinaryCommon::create_op()
+    // Set the elementwise operation to Sub then call the elementwise common create_op
+    ElementwiseBinaryCommonAttributes common_attributes{};
+    common_attributes.operation(ElementwiseBinaryCommonAttributes::ElementwiseOp::Sub);
+    return GpuElementwiseBinaryCommon::create_op(sketch, lhs, rhs, common_attributes);
+}
+
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
diff --git a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateElementwiseBinary.cpp b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateElementwiseBinary.cpp
index 0dd7ca5e78..52164ba41d 100644
--- a/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateElementwiseBinary.cpp
+++ b/src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateElementwiseBinary.cpp
@@ -185,6 +185,9 @@ TagLUT ClTemplateElementwiseBinary::get_tag_lut(const GpuKernelVariableTable &vt
         case Attributes::ElementwiseOp::Add:
             lut["ELTWISE_OP"] = "ADD";
             break;
+        case Attributes::ElementwiseOp::Sub:
+            lut["ELTWISE_OP"] = "SUB";
+            break;
         case Attributes::ElementwiseOp::Mul:
             lut["ELTWISE_OP"] = "MUL";
             break;