arm_compute v17.06

8 files changed, 317 insertions, 17 deletions

diff --git a/examples/SConscript b/examples/SConscript
new file mode 100644
index 000000000..748f771ec
--- /dev/null
+++ b/examples/SConscript
@@ -0,0 +1,70 @@
+# Copyright (c) 2017 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.
+import SCons
+import os.path
+
+Import('env')
+Import('arm_compute_a')
+Import('arm_compute_so')
+
+if env['opencl']:
+    Import('opencl')
+
+examples_env = env.Clone()
+
+examples_env.Append(CPPPATH = ["#"])
+examples_env.Append(LIBPATH = ["#build/%s" % env['build_dir']])
+examples_env.Append(LIBPATH = ["#build/%s/opencl-1.2-stubs" % env['build_dir']])
+
+# Build examples
+utils = examples_env.Object("../utils/Utils.cpp")
+
+if env['os'] in ['android', 'bare_metal']:
+    arm_compute_lib = arm_compute_a
+    arm_compute_dependency = arm_compute_a
+else:
+    arm_compute_lib = "arm_compute"
+    arm_compute_dependency = arm_compute_so
+
+if env['opencl'] and env['neon']:
+    for file in Glob("./neoncl_*.cpp"):
+        example = os.path.basename(os.path.splitext(str(file))[0])
+        prog = examples_env.Program(example, ["{}.cpp".format(example), utils], LIBS = [arm_compute_lib, "OpenCL"])
+        Depends(prog, [arm_compute_dependency, opencl])
+        alias = examples_env.Alias(example, prog)
+        Default(alias)
+
+if env['opencl']:
+    for file in Glob("./cl_*.cpp"):
+        example = os.path.basename(os.path.splitext(str(file))[0])
+        prog = examples_env.Program(example, ["{}.cpp".format(example), utils], LIBS = [arm_compute_lib, "OpenCL"])
+        Depends(prog, [arm_compute_dependency, opencl])
+        alias = examples_env.Alias(example, prog)
+        Default(alias)
+
+if env['neon']:
+    for file in Glob("./neon_*.cpp"):
+        example = os.path.basename(os.path.splitext(str(file))[0])
+        prog = examples_env.Program(example, ["{}.cpp".format(example), utils], LIBS = [arm_compute_lib])
+        Depends(prog, arm_compute_dependency)
+        alias = examples_env.Alias(example, prog)
+        Default(alias)
diff --git a/examples/cl_convolution.cpp b/examples/cl_convolution.cpp
index a021cdbae..06f6f144e 100644
--- a/examples/cl_convolution.cpp
+++ b/examples/cl_convolution.cpp
@@ -25,10 +25,10 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLFunctions.h"
 #include "arm_compute/runtime/CL/CLScheduler.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 using namespace arm_compute;
-using namespace test_helpers;
+using namespace utils;
 
 /** Gaussian 3x3 matrix
  */
@@ -114,5 +114,5 @@ void main_cl_convolution(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_cl_convolution);
+    return utils::run_example(argc, argv, main_cl_convolution);
 }
diff --git a/examples/cl_events.cpp b/examples/cl_events.cpp
index 5c39788b4..768f62062 100644
--- a/examples/cl_events.cpp
+++ b/examples/cl_events.cpp
@@ -25,10 +25,10 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLFunctions.h"
 #include "arm_compute/runtime/CL/CLScheduler.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 using namespace arm_compute;
-using namespace test_helpers;
+using namespace utils;
 
 void main_cl_events(int argc, const char **argv)
 {
@@ -110,5 +110,5 @@ void main_cl_events(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_cl_events);
+    return utils::run_example(argc, argv, main_cl_events);
 }
diff --git a/examples/neon_cnn.cpp b/examples/neon_cnn.cpp
new file mode 100644
index 000000000..952ae4d48
--- /dev/null
+++ b/examples/neon_cnn.cpp
@@ -0,0 +1,230 @@
+/*
+ * Copyright (c) 2016, 2017 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/runtime/NEON/NEFunctions.h"
+
+#include "arm_compute/core/Types.h"
+#include "utils/Utils.h"
+
+using namespace arm_compute;
+using namespace utils;
+
+void main_cnn(int argc, const char **argv)
+{
+    ARM_COMPUTE_UNUSED(argc);
+    ARM_COMPUTE_UNUSED(argv);
+
+    // The src tensor should contain the input image
+    Tensor src;
+
+    // The weights and biases tensors should be initialized with the values inferred with the training
+    Tensor weights0;
+    Tensor weights1;
+    Tensor weights2;
+    Tensor biases0;
+    Tensor biases1;
+    Tensor biases2;
+
+    Tensor out_conv0;
+    Tensor out_conv1;
+    Tensor out_act0;
+    Tensor out_act1;
+    Tensor out_act2;
+    Tensor out_pool0;
+    Tensor out_pool1;
+    Tensor out_fc0;
+    Tensor out_softmax;
+
+    NEConvolutionLayer    conv0;
+    NEConvolutionLayer    conv1;
+    NEPoolingLayer        pool0;
+    NEPoolingLayer        pool1;
+    NEFullyConnectedLayer fc0;
+    NEActivationLayer     act0;
+    NEActivationLayer     act1;
+    NEActivationLayer     act2;
+    NESoftmaxLayer        softmax;
+
+    /* [Initialize tensors] */
+
+    // Initialize src tensor
+    constexpr unsigned int width_src_image  = 32;
+    constexpr unsigned int height_src_image = 32;
+    constexpr unsigned int ifm_src_img      = 1;
+
+    const TensorShape src_shape(width_src_image, height_src_image, ifm_src_img);
+    src.allocator()->init(TensorInfo(src_shape, 1, DataType::F32));
+
+    // Initialize tensors of conv0
+    constexpr unsigned int kernel_x_conv0 = 5;
+    constexpr unsigned int kernel_y_conv0 = 5;
+    constexpr unsigned int ofm_conv0      = 8;
+
+    const TensorShape weights_shape_conv0(kernel_x_conv0, kernel_y_conv0, src_shape.z(), ofm_conv0);
+    const TensorShape biases_shape_conv0(weights_shape_conv0[3]);
+    const TensorShape out_shape_conv0(src_shape.x(), src_shape.y(), weights_shape_conv0[3]);
+
+    weights0.allocator()->init(TensorInfo(weights_shape_conv0, 1, DataType::F32));
+    biases0.allocator()->init(TensorInfo(biases_shape_conv0, 1, DataType::F32));
+    out_conv0.allocator()->init(TensorInfo(out_shape_conv0, 1, DataType::F32));
+
+    // Initialize tensor of act0
+    out_act0.allocator()->init(TensorInfo(out_shape_conv0, 1, DataType::F32));
+
+    // Initialize tensor of pool0
+    TensorShape out_shape_pool0 = out_shape_conv0;
+    out_shape_pool0.set(0, out_shape_pool0.x() / 2);
+    out_shape_pool0.set(1, out_shape_pool0.y() / 2);
+    out_pool0.allocator()->init(TensorInfo(out_shape_pool0, 1, DataType::F32));
+
+    // Initialize tensors of conv1
+    constexpr unsigned int kernel_x_conv1 = 3;
+    constexpr unsigned int kernel_y_conv1 = 3;
+    constexpr unsigned int ofm_conv1      = 16;
+
+    const TensorShape weights_shape_conv1(kernel_x_conv1, kernel_y_conv1, out_shape_pool0.z(), ofm_conv1);
+
+    const TensorShape biases_shape_conv1(weights_shape_conv1[3]);
+    const TensorShape out_shape_conv1(out_shape_pool0.x(), out_shape_pool0.y(), weights_shape_conv1[3]);
+
+    weights1.allocator()->init(TensorInfo(weights_shape_conv1, 1, DataType::F32));
+    biases1.allocator()->init(TensorInfo(biases_shape_conv1, 1, DataType::F32));
+    out_conv1.allocator()->init(TensorInfo(out_shape_conv1, 1, DataType::F32));
+
+    // Initialize tensor of act1
+    out_act1.allocator()->init(TensorInfo(out_shape_conv1, 1, DataType::F32));
+
+    // Initialize tensor of pool1
+    TensorShape out_shape_pool1 = out_shape_conv1;
+    out_shape_pool1.set(0, out_shape_pool1.x() / 2);
+    out_shape_pool1.set(1, out_shape_pool1.y() / 2);
+    out_pool1.allocator()->init(TensorInfo(out_shape_pool1, 1, DataType::F32));
+
+    // Initialize tensor of fc0
+    constexpr unsigned int num_labels = 128;
+
+    const TensorShape weights_shape_fc0(out_shape_pool1.x() * out_shape_pool1.y() * out_shape_pool1.z(), num_labels);
+    const TensorShape biases_shape_fc0(num_labels);
+    const TensorShape out_shape_fc0(num_labels);
+
+    weights2.allocator()->init(TensorInfo(weights_shape_fc0, 1, DataType::F32));
+    biases2.allocator()->init(TensorInfo(biases_shape_fc0, 1, DataType::F32));
+    out_fc0.allocator()->init(TensorInfo(out_shape_fc0, 1, DataType::F32));
+
+    // Initialize tensor of act2
+    out_act2.allocator()->init(TensorInfo(out_shape_fc0, 1, DataType::F32));
+
+    // Initialize tensor of softmax
+    const TensorShape out_shape_softmax(out_shape_fc0.x());
+    out_softmax.allocator()->init(TensorInfo(out_shape_softmax, 1, DataType::F32));
+
+    /* -----------------------End: [Initialize tensors] */
+
+    /* [Configure functions] */
+
+    // in:32x32x1: 5x5 convolution, 8 output features maps (OFM)
+    conv0.configure(&src, &weights0, &biases0, &out_conv0, PadStrideInfo());
+
+    // in:32x32x8, out:32x32x8, Activation function: relu
+    act0.configure(&out_conv0, &out_act0, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
+
+    // in:32x32x8, out:16x16x8 (2x2 pooling), Pool type function: Max
+    pool0.configure(&out_act0, &out_pool0, PoolingLayerInfo(PoolingType::MAX, 2));
+
+    // in:16x16x8: 3x3 convolution, 16 output features maps (OFM)
+    conv1.configure(&out_pool0, &weights1, &biases1, &out_conv1, PadStrideInfo());
+
+    // in:16x16x16, out:16x16x16, Activation function: relu
+    act1.configure(&out_conv1, &out_act1, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
+
+    // in:16x16x16, out:8x8x16 (2x2 pooling), Pool type function: Average
+    pool1.configure(&out_act1, &out_pool1, PoolingLayerInfo(PoolingType::AVG, 2));
+
+    // in:8x8x16, out:128
+    fc0.configure(&out_pool1, &weights2, &biases2, &out_fc0);
+
+    // in:128, out:128, Activation function: relu
+    act2.configure(&out_fc0, &out_act2, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
+
+    // in:128, out:128
+    softmax.configure(&out_act2, &out_softmax);
+
+    /* -----------------------End: [Configure functions] */
+
+    /* [Allocate tensors] */
+
+    // Now that the padding requirements are known we can allocate the images:
+    src.allocator()->allocate();
+    weights0.allocator()->allocate();
+    weights1.allocator()->allocate();
+    weights2.allocator()->allocate();
+    biases0.allocator()->allocate();
+    biases1.allocator()->allocate();
+    biases2.allocator()->allocate();
+    out_conv0.allocator()->allocate();
+    out_conv1.allocator()->allocate();
+    out_act0.allocator()->allocate();
+    out_act1.allocator()->allocate();
+    out_act2.allocator()->allocate();
+    out_pool0.allocator()->allocate();
+    out_pool1.allocator()->allocate();
+    out_fc0.allocator()->allocate();
+    out_softmax.allocator()->allocate();
+
+    /* -----------------------End: [Allocate tensors] */
+
+    /* [Initialize weights and biases tensors] */
+
+    // Once the tensors have been allocated, the src, weights and biases tensors can be initialized
+    // ...
+
+    /* -----------------------[Initialize weights and biases tensors] */
+
+    /* [Execute the functions] */
+
+    conv0.run();
+    act0.run();
+    pool0.run();
+    conv1.run();
+    act1.run();
+    pool1.run();
+    fc0.run();
+    act2.run();
+    softmax.run();
+
+    /* -----------------------End: [Execute the functions] */
+}
+
+/** Main program for cnn test
+ *
+ * The example implements the following CNN architecture:
+ *
+ * Input -> conv0:5x5 -> act0:relu -> pool:2x2 -> conv1:3x3 -> act1:relu -> pool:2x2 -> fc0 -> act2:relu -> softmax
+ *
+ * @param[in] argc Number of arguments
+ * @param[in] argv Arguments
+ */
+int main(int argc, const char **argv)
+{
+    return utils::run_example(argc, argv, main_cnn);
+}
\ No newline at end of file
diff --git a/examples/neon_convolution.cpp b/examples/neon_convolution.cpp
index fc68aa24b..222c8f9a3 100644
--- a/examples/neon_convolution.cpp
+++ b/examples/neon_convolution.cpp
@@ -24,10 +24,10 @@
 #include "arm_compute/runtime/NEON/NEFunctions.h"
 
 #include "arm_compute/core/Types.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 using namespace arm_compute;
-using namespace test_helpers;
+using namespace utils;
 
 /** Gaussian 3x3 matrix
  */
@@ -113,5 +113,5 @@ void main_neon_convolution(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_neon_convolution);
+    return utils::run_example(argc, argv, main_neon_convolution);
 }
diff --git a/examples/neon_copy_objects.cpp b/examples/neon_copy_objects.cpp
index 3f53939c3..191f45555 100644
--- a/examples/neon_copy_objects.cpp
+++ b/examples/neon_copy_objects.cpp
@@ -25,7 +25,7 @@
 #include "arm_compute/runtime/NEON/NEFunctions.h"
 
 #include "arm_compute/core/Types.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 #include <cstring>
 #include <iostream>
@@ -148,5 +148,5 @@ void main_neon_copy_objects(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_neon_copy_objects);
+    return utils::run_example(argc, argv, main_neon_copy_objects);
 }
diff --git a/examples/neon_scale.cpp b/examples/neon_scale.cpp
index c1435af08..75780c9bd 100644
--- a/examples/neon_scale.cpp
+++ b/examples/neon_scale.cpp
@@ -24,10 +24,10 @@
 #include "arm_compute/runtime/NEON/NEFunctions.h"
 
 #include "arm_compute/core/Types.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 using namespace arm_compute;
-using namespace test_helpers;
+using namespace utils;
 
 void main_neon_scale(int argc, const char **argv)
 {
@@ -86,5 +86,5 @@ void main_neon_scale(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_neon_scale);
+    return utils::run_example(argc, argv, main_neon_scale);
 }
diff --git a/examples/neoncl_scale_median_gaussian.cpp b/examples/neoncl_scale_median_gaussian.cpp
index a4e44146e..a32ba6daf 100644
--- a/examples/neoncl_scale_median_gaussian.cpp
+++ b/examples/neoncl_scale_median_gaussian.cpp
@@ -26,10 +26,10 @@
 #include "arm_compute/runtime/CL/CLFunctions.h"
 #include "arm_compute/runtime/CL/CLScheduler.h"
 #include "arm_compute/runtime/NEON/NEFunctions.h"
-#include "test_helpers/Utils.h"
+#include "utils/Utils.h"
 
 using namespace arm_compute;
-using namespace test_helpers;
+using namespace utils;
 
 /** Example demonstrating how to use both CL and NEON functions in the same pipeline
  *
@@ -122,5 +122,5 @@ void main_neoncl_scale_median_gaussian(int argc, const char **argv)
  */
 int main(int argc, const char **argv)
 {
-    return test_helpers::run_example(argc, argv, main_neoncl_scale_median_gaussian);
+    return utils::run_example(argc, argv, main_neoncl_scale_median_gaussian);
 }