* [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group.
@ 2018-12-04 21:30 Qin Su
2018-12-04 21:35 ` Denys Dmytriyenko
0 siblings, 1 reply; 3+ messages in thread
From: Qin Su @ 2018-12-04 21:30 UTC (permalink / raw)
To: meta-arago
Updated to the latest version of 18.08.
Updated LIC_FILES_CHKSUM as Copyright time changed in LICENSE from 2017 to 2017-2018.
Updated to generate versioned library.
Added benchmark preset group for squeezenet and inceptionnet.
Cleaned up install built source package needed by ARMNN.
Fixed symlinks lost issue in do_install().
Signed-off-by: Qin Su <qsu@ti.com>
---
Changes in PACTH v4:
1. Fixed the issue -"ERROR:QA Issue: armnn rdepends on arm-compute-library-dev" in building armnn.
Changes in PATCH v3:
1. Fixed patch files permission issue
2. Cleaned up install source package required by armnn
Changes in PATCH v2:
1. Added Upstream-Status in the commit log in patch files
2. Added description of license change in commit log
3. Removed unnecessary modification
---
.../0001-add-ti-benchmark-test-group.patch | 61 +++++++++++++++++++
.../0002-add-ti-benchmark-test-group.patch | 71 ++++++++++++++++++++++
.../0003-add-ti-benchmark-test-group.patch | 39 ++++++++++++
.../0004-add-ti-benchmark-test-group.patch | 49 +++++++++++++++
.../0005-add-ti-benchmark-test-group.patch | 46 ++++++++++++++
.../0006-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
.../0007-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
.../0008-add-ti-benchmark-test-group.patch | 31 ++++++++++
.../0009-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
.../0010-add-ti-benchmark-test-group.patch | 50 +++++++++++++++
.../0011-add-ti-benchmark-test-group.patch | 37 +++++++++++
11 files changed, 516 insertions(+)
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..f00561e
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
@@ -0,0 +1,61 @@
+From 497a8abf760dbd9be715dfe6abf81c670fc85912 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:16:33 -0500
+Subject: [PATCH 01/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ .../squeezenet/SqueezeNetActivationLayerDataset.h | 35 ++++++++++++++++++++++
+ 1 file changed, 35 insertions(+)
+
+diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
+index 7f4bf4d..a66f473 100644
+--- a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
+@@ -71,6 +71,41 @@ public:
+ SqueezeNetActivationLayerDataset(SqueezeNetActivationLayerDataset &&) = default;
+ ~SqueezeNetActivationLayerDataset() = default;
+ };
++class TISqueezeNetActivationLayerDataset final : public
++ framework::dataset::CartesianProductDataset<framework::dataset::InitializerListDataset<TensorShape>, framework::dataset::SingletonDataset<ActivationLayerInfo>>
++{
++public:
++ TISqueezeNetActivationLayerDataset()
++ : CartesianProductDataset
++ {
++ framework::dataset::make("Shape", { // relu_conv1
++ TensorShape(114U, 114U, 64U),
++ // fire2/relu_squeeze1x1, fire3/relu_squeeze1x1
++ TensorShape(57U, 57U, 16U),
++ // fire2/relu_expand1x1, fire2/relu_expand3x3, fire3/relu_expand1x1, fire3/relu_expand3x3
++ TensorShape(57U, 57U, 64U),
++ // fire4/relu_squeeze1x1, fire5/relu_squeeze1x1
++ TensorShape(28U, 28U, 32U),
++ // fire4/relu_expand1x1, fire4/relu_expand3x3, fire5/relu_expand1x1, fire5/relu_expand3x3
++ TensorShape(28U, 28U, 128U),
++ // fire6/relu_squeeze1x1, fire7/relu_squeeze1x1
++ TensorShape(14U, 14U, 48U),
++ // fire6/relu_expand1x1, fire6/relu_expand3x3, fire7/relu_expand1x1, fire7/relu_expand3x3
++ TensorShape(14U, 14U, 192U),
++ // fire8/relu_squeeze1x1, fire9/relu_squeeze1x1
++ TensorShape(14U, 14U, 64U),
++ // fire8/relu_expand1x1, fire8/relu_expand3x3, fire9/relu_expand1x1, fire9/relu_expand3x3
++ TensorShape(14U, 14U, 256U),
++ // relu_conv10
++ TensorShape(14U, 14U, 1000U) }),
++ framework::dataset::make("Info", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
++ }
++ {
++ }
++ TISqueezeNetActivationLayerDataset(TISqueezeNetActivationLayerDataset &&) = default;
++ ~TISqueezeNetActivationLayerDataset() = default;
++};
++
+ } // namespace datasets
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..f891f74
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
@@ -0,0 +1,71 @@
+From 39ebc12e4de88a7839e33755c65cb9b9f0e8535f Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:17:19 -0500
+Subject: [PATCH 02/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ .../squeezenet/SqueezeNetConvolutionLayerDataset.h | 45 ++++++++++++++++++++++
+ 1 file changed, 45 insertions(+)
+
+diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
+index f98d90a..cba37eb 100644
+--- a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
+@@ -96,6 +96,51 @@ public:
+ add_config(TensorShape(13U, 13U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(13U, 13U, 1000U), PadStrideInfo(1, 1, 0, 0));
+ }
+ };
++
++class TISqueezeNetConvolutionLayerDataset final : public ConvolutionLayerDataset
++{
++public:
++ TISqueezeNetConvolutionLayerDataset()
++ {
++ // conv1
++ add_config(TensorShape(227U, 227U, 3U), TensorShape(3U, 3U, 3U, 64U), TensorShape(64U), TensorShape(114U, 114U, 64U), PadStrideInfo(2, 2, 0, 0));
++ // fire2/squeeze1x1
++ add_config(TensorShape(57U, 57U, 64U), TensorShape(1U, 1U, 64U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
++ // fire2/expand1x1, fire3/expand1x1
++ add_config(TensorShape(57U, 57U, 16U), TensorShape(1U, 1U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 0, 0));
++ // fire2/expand3x3, fire3/expand3x3
++ add_config(TensorShape(57U, 57U, 16U), TensorShape(3U, 3U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 1, 1));
++ // fire3/squeeze1x1
++ add_config(TensorShape(57U, 57U, 128U), TensorShape(1U, 1U, 128U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
++ // fire4/squeeze1x1
++ add_config(TensorShape(28U, 28U, 128U), TensorShape(1U, 1U, 128U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
++ // fire4/expand1x1, fire5/expand1x1
++ add_config(TensorShape(28U, 28U, 32U), TensorShape(1U, 1U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 0, 0));
++ // fire4/expand3x3, fire5/expand3x3
++ add_config(TensorShape(28U, 28U, 32U), TensorShape(3U, 3U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 1, 1));
++ // fire5/squeeze1x1
++ add_config(TensorShape(28U, 28U, 256U), TensorShape(1U, 1U, 256U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
++ // fire6/squeeze1x1
++ add_config(TensorShape(14U, 14U, 256U), TensorShape(1U, 1U, 256U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
++ // fire6/expand1x1, fire7/expand1x1
++ add_config(TensorShape(14U, 14U, 48U), TensorShape(1U, 1U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 0, 0));
++ // fire6/expand3x3, fire7/expand3x3
++ add_config(TensorShape(14U, 14U, 48U), TensorShape(3U, 3U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 1, 1));
++ // fire7/squeeze1x1
++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
++ // fire8/squeeze1x1
++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
++ // fire8/expand1x1, fire9/expand1x1
++ add_config(TensorShape(14U, 14U, 64U), TensorShape(1U, 1U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 0, 0));
++ // fire8/expand3x3, fire9/expand3x3
++ add_config(TensorShape(14U, 14U, 64U), TensorShape(3U, 3U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 1, 1));
++ // fire9/squeeze1x1
++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
++ // conv10
++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(14U, 14U, 1000U), PadStrideInfo(1, 1, 0, 0));
++ }
++};
++
+ } // namespace datasets
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..4a80548
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
@@ -0,0 +1,39 @@
+From 8c433526117e4aa5480efdbe5718ee54e9de2ff5 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:17:43 -0500
+Subject: [PATCH 03/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ .../system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h | 13 +++++++++++++
+ 1 file changed, 13 insertions(+)
+
+diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
+index 9518ce5..a168d29 100644
+--- a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
+@@ -50,6 +50,19 @@ public:
+ add_config(TensorShape(27U, 27U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
+ }
+ };
++class TISqueezeNetPoolingLayerDataset final : public PoolingLayerDataset
++{
++public:
++ TISqueezeNetPoolingLayerDataset()
++ {
++ // pool1
++ add_config(TensorShape(114U, 114U, 64U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
++ // pool3
++ add_config(TensorShape(57U, 57U, 128U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
++ // pool5
++ add_config(TensorShape(28U, 28U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
++ }
++};
+ } // namespace datasets
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..9693e6c
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
@@ -0,0 +1,49 @@
+From c2e6412d2aa9137b4b1b75d091ba9c8a53ede99b Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:18:05 -0500
+Subject: [PATCH 04/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/ActivationLayer.cpp | 13 +++++++++++++
+ 1 file changed, 13 insertions(+)
+
+diff --git a/tests/benchmark/NEON/ActivationLayer.cpp b/tests/benchmark/NEON/ActivationLayer.cpp
+index 1c4ea21..7b47c11 100644
+--- a/tests/benchmark/NEON/ActivationLayer.cpp
++++ b/tests/benchmark/NEON/ActivationLayer.cpp
+@@ -51,9 +51,11 @@ namespace
+ #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
+ const auto data_types_mobilenet = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
+ const auto data_types_mobilenet = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ } // namespace
+
+@@ -127,6 +129,17 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2ActivationLayer, NEActivationLayerFixture,
+ framework::dataset::make("Batches", { 4, 8 })));
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetActivationLayerDataset(), data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ActivationLayerDataset(), data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
++
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..f498ffa
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
@@ -0,0 +1,46 @@
+From 14ba479fe707af2575fac4d2780e31a10264d295 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:18:32 -0500
+Subject: [PATCH 05/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/ConvolutionLayer.cpp | 11 ++++++++++-
+ 1 file changed, 10 insertions(+), 1 deletion(-)
+
+diff --git a/tests/benchmark/NEON/ConvolutionLayer.cpp b/tests/benchmark/NEON/ConvolutionLayer.cpp
+index ac27e7a..6cfb00d 100644
+--- a/tests/benchmark/NEON/ConvolutionLayer.cpp
++++ b/tests/benchmark/NEON/ConvolutionLayer.cpp
+@@ -52,9 +52,10 @@ namespace
+ {
+ #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
+-
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ } // namespace
+
+@@ -197,6 +198,14 @@ REGISTER_FIXTURE_DATA_TEST_CASE(SqueezeNetWinogradLayer, NEWinogradConvolutionLa
+
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ConvolutionLayer, NEGEMMConvolutionLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ConvolutionLayerDataset(),
++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
++ data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..3dfc2e7
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
@@ -0,0 +1,44 @@
+From c0c2b79824a52dee291b275a0f0a66746a99dbf5 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:18:50 -0500
+Subject: [PATCH 06/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/DirectConvolutionLayer.cpp | 9 +++++++++
+ 1 file changed, 9 insertions(+)
+
+diff --git a/tests/benchmark/NEON/DirectConvolutionLayer.cpp b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
+index f94ef6b..afdcbad 100644
+--- a/tests/benchmark/NEON/DirectConvolutionLayer.cpp
++++ b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
+@@ -50,8 +50,10 @@ namespace
+ // Special data types for networks that need 5x5 direct convolution
+ #ifdef ARM_COMPUTE_ENABLE_F16
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* ARM_COMPUTE_ENABLE_F16 */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* ARM_COMPUTE_ENABLE_F16 */
+ } // namespace
+
+@@ -122,6 +124,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2DirectConvolutionLayer, NEDirectConvolutio
+
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetDirectConvolutionLayer, NEDirectConvolutionLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetConvolutionLayerDataset(),
++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
++ data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..a1ed073
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
@@ -0,0 +1,44 @@
+From e8c375920cd21e6e151aabaa88b68358296bbe87 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:19:10 -0500
+Subject: [PATCH 07/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/FullyConnectedLayer.cpp | 9 +++++++++
+ 1 file changed, 9 insertions(+)
+
+diff --git a/tests/benchmark/NEON/FullyConnectedLayer.cpp b/tests/benchmark/NEON/FullyConnectedLayer.cpp
+index d0fb04f..9ee852d 100644
+--- a/tests/benchmark/NEON/FullyConnectedLayer.cpp
++++ b/tests/benchmark/NEON/FullyConnectedLayer.cpp
+@@ -47,8 +47,10 @@ namespace
+ {
+ #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ } // namespace
+
+@@ -108,6 +110,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV4FullyConnectedLayer, NEFully
+ framework::dataset::make("Batches", { 4, 8 })));
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1FullyConnectedLayer, NEFullyConnectedLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1FullyConnectedLayerDataset(),
++ data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
++
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..dcf0989
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
@@ -0,0 +1,31 @@
+From d657c798ec32c6a0f08384f725f3b8601d035300 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:19:26 -0500
+Subject: [PATCH 08/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/GEMMLowp.cpp | 5 +++++
+ 1 file changed, 5 insertions(+)
+
+diff --git a/tests/benchmark/NEON/GEMMLowp.cpp b/tests/benchmark/NEON/GEMMLowp.cpp
+index 4b13f31..c2c6bdf 100644
+--- a/tests/benchmark/NEON/GEMMLowp.cpp
++++ b/tests/benchmark/NEON/GEMMLowp.cpp
+@@ -50,6 +50,11 @@ REGISTER_FIXTURE_DATA_TEST_CASE(MatrixMultiplyGEMMLowp, NEGEMMLowpFixture, frame
+ REGISTER_FIXTURE_DATA_TEST_CASE(GoogleNetGEMMLowp, NEGEMMLowpFixture, framework::DatasetMode::NIGHTLY, datasets::GoogleNetGEMMDataset());
+
+ TEST_SUITE_END()
++
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1GEMMLowp, NEGEMMLowpFixture, framework::DatasetMode::ALL, datasets::GoogLeNetInceptionV1GEMMDataset());
++TEST_SUITE_END()
++
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..8498e27
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
@@ -0,0 +1,44 @@
+From 5e6db070b3cc563095991eaaab5d4c1fe8f12fcf Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:19:42 -0500
+Subject: [PATCH 09/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/NormalizationLayer.cpp | 9 +++++++++
+ 1 file changed, 9 insertions(+)
+
+diff --git a/tests/benchmark/NEON/NormalizationLayer.cpp b/tests/benchmark/NEON/NormalizationLayer.cpp
+index 0c9a864..3654ef5 100644
+--- a/tests/benchmark/NEON/NormalizationLayer.cpp
++++ b/tests/benchmark/NEON/NormalizationLayer.cpp
+@@ -44,8 +44,10 @@ namespace
+ {
+ #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ } // namespace
+ using NENormalizationLayerFixture = NormalizationLayerFixture<Tensor, NENormalizationLayer, Accessor>;
+@@ -74,6 +76,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormal
+ framework::dataset::make("Batches", { 4, 8 })));
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormalizationLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1NormalizationLayerDataset(),
++ data_types_tibenchmark),
++ framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
++
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..4c562e7
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
@@ -0,0 +1,50 @@
+From 76a4ca1417987c6542dd76b625d001ae8f30657e Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:19:58 -0500
+Subject: [PATCH 10/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/benchmark/NEON/PoolingLayer.cpp | 13 +++++++++++++
+ 1 file changed, 13 insertions(+)
+
+diff --git a/tests/benchmark/NEON/PoolingLayer.cpp b/tests/benchmark/NEON/PoolingLayer.cpp
+index 8b7ee84..de93006 100644
+--- a/tests/benchmark/NEON/PoolingLayer.cpp
++++ b/tests/benchmark/NEON/PoolingLayer.cpp
+@@ -48,10 +48,13 @@ namespace benchmark
+ namespace
+ {
+ const auto data_layouts = framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC });
++const auto data_layouts_tibenchmark = framework::dataset::make("DataLayout", { DataLayout::NCHW});
+ #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+ const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ const auto data_types = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
+ #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+ } // namespace
+
+@@ -107,6 +110,16 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2PoolingLayer, NEPoolingLayerFixture, frame
+ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::YOLOV2PoolingLayerDataset(), data_types), data_layouts), framework::dataset::make("Batches", { 4, 8 })));
+ TEST_SUITE_END()
+ TEST_SUITE_END()
++
++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetPoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetPoolingLayerDataset(), data_types_tibenchmark), data_layouts_tibenchmark), framework::dataset::make("Batches", 1)));
++TEST_SUITE_END()
++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1PoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1PoolingLayerDataset(), data_types_tibenchmark), data_layouts_tibenchmark), framework::dataset::make("Batches",
++ 1)));
++TEST_SUITE_END()
+ } // namespace benchmark
+ } // namespace test
+ } // namespace arm_compute
+--
+1.9.1
+
diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
new file mode 100644
index 0000000..fd3a9d1
--- /dev/null
+++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
@@ -0,0 +1,37 @@
+From 012d21f475e1a004c2a7f688835a1b5416c3b8b6 Mon Sep 17 00:00:00 2001
+From: Qin Su <qsu@ti.com>
+Date: Tue, 27 Nov 2018 17:20:10 -0500
+Subject: [PATCH 11/11] add TI benchmark test group
+
+Upstream-Status: Inappropriate [TI only test code]
+
+Signed-off-by: Qin Su <qsu@ti.com>
+---
+ tests/framework/printers/PrettyPrinter.cpp | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/tests/framework/printers/PrettyPrinter.cpp b/tests/framework/printers/PrettyPrinter.cpp
+index 3181951..2222286 100644
+--- a/tests/framework/printers/PrettyPrinter.cpp
++++ b/tests/framework/printers/PrettyPrinter.cpp
+@@ -83,7 +83,7 @@ void PrettyPrinter::print_run_footer()
+
+ void PrettyPrinter::print_test_header(const TestInfo &info)
+ {
+- *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << "\n";
++ *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << " ";
+ }
+
+ void PrettyPrinter::print_test_footer()
+@@ -125,7 +125,7 @@ void PrettyPrinter::print_measurements(const Profiler::MeasurementsMap &measurem
+ InstrumentsStats stats(instrument.second);
+
+ *_stream << " ";
+- *_stream << "AVG=" << stats.mean() << " " << stats.max().unit();
++ *_stream << "AVG= " << stats.mean() << " " << stats.max().unit();
+ if(instrument.second.size() > 1)
+ {
+ *_stream << ", STDDEV=" << arithmetic_to_string(stats.relative_standard_deviation(), 2) << " %";
+--
+1.9.1
+
--
1.9.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group.
2018-12-04 21:30 [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group Qin Su
@ 2018-12-04 21:35 ` Denys Dmytriyenko
2018-12-04 21:36 ` Su, Qin
0 siblings, 1 reply; 3+ messages in thread
From: Denys Dmytriyenko @ 2018-12-04 21:35 UTC (permalink / raw)
To: Qin Su; +Cc: meta-arago
The recipe itseld is missing.
On Tue, Dec 04, 2018 at 04:30:51PM -0500, Qin Su wrote:
> Updated to the latest version of 18.08.
> Updated LIC_FILES_CHKSUM as Copyright time changed in LICENSE from 2017 to 2017-2018.
> Updated to generate versioned library.
> Added benchmark preset group for squeezenet and inceptionnet.
> Cleaned up install built source package needed by ARMNN.
> Fixed symlinks lost issue in do_install().
>
> Signed-off-by: Qin Su <qsu@ti.com>
> ---
> Changes in PACTH v4:
> 1. Fixed the issue -"ERROR:QA Issue: armnn rdepends on arm-compute-library-dev" in building armnn.
Why/how? Better say - use "cp" instead of "install" to preserve symlinks...
> Changes in PATCH v3:
> 1. Fixed patch files permission issue
> 2. Cleaned up install source package required by armnn
>
> Changes in PATCH v2:
> 1. Added Upstream-Status in the commit log in patch files
> 2. Added description of license change in commit log
> 3. Removed unnecessary modification
> ---
> .../0001-add-ti-benchmark-test-group.patch | 61 +++++++++++++++++++
> .../0002-add-ti-benchmark-test-group.patch | 71 ++++++++++++++++++++++
> .../0003-add-ti-benchmark-test-group.patch | 39 ++++++++++++
> .../0004-add-ti-benchmark-test-group.patch | 49 +++++++++++++++
> .../0005-add-ti-benchmark-test-group.patch | 46 ++++++++++++++
> .../0006-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0007-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0008-add-ti-benchmark-test-group.patch | 31 ++++++++++
> .../0009-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0010-add-ti-benchmark-test-group.patch | 50 +++++++++++++++
> .../0011-add-ti-benchmark-test-group.patch | 37 +++++++++++
> 11 files changed, 516 insertions(+)
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
> create mode 100644 meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
>
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f00561e
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0001-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,61 @@
> +From 497a8abf760dbd9be715dfe6abf81c670fc85912 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:16:33 -0500
> +Subject: [PATCH 01/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../squeezenet/SqueezeNetActivationLayerDataset.h | 35 ++++++++++++++++++++++
> + 1 file changed, 35 insertions(+)
> +
> +diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
> +index 7f4bf4d..a66f473 100644
> +--- a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDataset.h
> +@@ -71,6 +71,41 @@ public:
> + SqueezeNetActivationLayerDataset(SqueezeNetActivationLayerDataset &&) = default;
> + ~SqueezeNetActivationLayerDataset() = default;
> + };
> ++class TISqueezeNetActivationLayerDataset final : public
> ++ framework::dataset::CartesianProductDataset<framework::dataset::InitializerListDataset<TensorShape>, framework::dataset::SingletonDataset<ActivationLayerInfo>>
> ++{
> ++public:
> ++ TISqueezeNetActivationLayerDataset()
> ++ : CartesianProductDataset
> ++ {
> ++ framework::dataset::make("Shape", { // relu_conv1
> ++ TensorShape(114U, 114U, 64U),
> ++ // fire2/relu_squeeze1x1, fire3/relu_squeeze1x1
> ++ TensorShape(57U, 57U, 16U),
> ++ // fire2/relu_expand1x1, fire2/relu_expand3x3, fire3/relu_expand1x1, fire3/relu_expand3x3
> ++ TensorShape(57U, 57U, 64U),
> ++ // fire4/relu_squeeze1x1, fire5/relu_squeeze1x1
> ++ TensorShape(28U, 28U, 32U),
> ++ // fire4/relu_expand1x1, fire4/relu_expand3x3, fire5/relu_expand1x1, fire5/relu_expand3x3
> ++ TensorShape(28U, 28U, 128U),
> ++ // fire6/relu_squeeze1x1, fire7/relu_squeeze1x1
> ++ TensorShape(14U, 14U, 48U),
> ++ // fire6/relu_expand1x1, fire6/relu_expand3x3, fire7/relu_expand1x1, fire7/relu_expand3x3
> ++ TensorShape(14U, 14U, 192U),
> ++ // fire8/relu_squeeze1x1, fire9/relu_squeeze1x1
> ++ TensorShape(14U, 14U, 64U),
> ++ // fire8/relu_expand1x1, fire8/relu_expand3x3, fire9/relu_expand1x1, fire9/relu_expand3x3
> ++ TensorShape(14U, 14U, 256U),
> ++ // relu_conv10
> ++ TensorShape(14U, 14U, 1000U) }),
> ++ framework::dataset::make("Info", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
> ++ }
> ++ {
> ++ }
> ++ TISqueezeNetActivationLayerDataset(TISqueezeNetActivationLayerDataset &&) = default;
> ++ ~TISqueezeNetActivationLayerDataset() = default;
> ++};
> ++
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f891f74
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0002-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,71 @@
> +From 39ebc12e4de88a7839e33755c65cb9b9f0e8535f Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:17:19 -0500
> +Subject: [PATCH 02/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../squeezenet/SqueezeNetConvolutionLayerDataset.h | 45 ++++++++++++++++++++++
> + 1 file changed, 45 insertions(+)
> +
> +diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
> +index f98d90a..cba37eb 100644
> +--- a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDataset.h
> +@@ -96,6 +96,51 @@ public:
> + add_config(TensorShape(13U, 13U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(13U, 13U, 1000U), PadStrideInfo(1, 1, 0, 0));
> + }
> + };
> ++
> ++class TISqueezeNetConvolutionLayerDataset final : public ConvolutionLayerDataset
> ++{
> ++public:
> ++ TISqueezeNetConvolutionLayerDataset()
> ++ {
> ++ // conv1
> ++ add_config(TensorShape(227U, 227U, 3U), TensorShape(3U, 3U, 3U, 64U), TensorShape(64U), TensorShape(114U, 114U, 64U), PadStrideInfo(2, 2, 0, 0));
> ++ // fire2/squeeze1x1
> ++ add_config(TensorShape(57U, 57U, 64U), TensorShape(1U, 1U, 64U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire2/expand1x1, fire3/expand1x1
> ++ add_config(TensorShape(57U, 57U, 16U), TensorShape(1U, 1U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire2/expand3x3, fire3/expand3x3
> ++ add_config(TensorShape(57U, 57U, 16U), TensorShape(3U, 3U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire3/squeeze1x1
> ++ add_config(TensorShape(57U, 57U, 128U), TensorShape(1U, 1U, 128U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/squeeze1x1
> ++ add_config(TensorShape(28U, 28U, 128U), TensorShape(1U, 1U, 128U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/expand1x1, fire5/expand1x1
> ++ add_config(TensorShape(28U, 28U, 32U), TensorShape(1U, 1U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/expand3x3, fire5/expand3x3
> ++ add_config(TensorShape(28U, 28U, 32U), TensorShape(3U, 3U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire5/squeeze1x1
> ++ add_config(TensorShape(28U, 28U, 256U), TensorShape(1U, 1U, 256U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 256U), TensorShape(1U, 1U, 256U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/expand1x1, fire7/expand1x1
> ++ add_config(TensorShape(14U, 14U, 48U), TensorShape(1U, 1U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/expand3x3, fire7/expand3x3
> ++ add_config(TensorShape(14U, 14U, 48U), TensorShape(3U, 3U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire7/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/expand1x1, fire9/expand1x1
> ++ add_config(TensorShape(14U, 14U, 64U), TensorShape(1U, 1U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/expand3x3, fire9/expand3x3
> ++ add_config(TensorShape(14U, 14U, 64U), TensorShape(3U, 3U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire9/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // conv10
> ++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(14U, 14U, 1000U), PadStrideInfo(1, 1, 0, 0));
> ++ }
> ++};
> ++
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..4a80548
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0003-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,39 @@
> +From 8c433526117e4aa5480efdbe5718ee54e9de2ff5 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:17:43 -0500
> +Subject: [PATCH 03/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h | 13 +++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
> +index 9518ce5..a168d29 100644
> +--- a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h
> +@@ -50,6 +50,19 @@ public:
> + add_config(TensorShape(27U, 27U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> + }
> + };
> ++class TISqueezeNetPoolingLayerDataset final : public PoolingLayerDataset
> ++{
> ++public:
> ++ TISqueezeNetPoolingLayerDataset()
> ++ {
> ++ // pool1
> ++ add_config(TensorShape(114U, 114U, 64U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ // pool3
> ++ add_config(TensorShape(57U, 57U, 128U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ // pool5
> ++ add_config(TensorShape(28U, 28U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ }
> ++};
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..9693e6c
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0004-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,49 @@
> +From c2e6412d2aa9137b4b1b75d091ba9c8a53ede99b Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:05 -0500
> +Subject: [PATCH 04/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/ActivationLayer.cpp | 13 +++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/ActivationLayer.cpp b/tests/benchmark/NEON/ActivationLayer.cpp
> +index 1c4ea21..7b47c11 100644
> +--- a/tests/benchmark/NEON/ActivationLayer.cpp
> ++++ b/tests/benchmark/NEON/ActivationLayer.cpp
> +@@ -51,9 +51,11 @@ namespace
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
> + const auto data_types_mobilenet = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
> + const auto data_types_mobilenet = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + } // namespace
> +
> +@@ -127,6 +129,17 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2ActivationLayer, NEActivationLayerFixture,
> + framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetActivationLayerDataset(), data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ActivationLayerDataset(), data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f498ffa
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0005-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,46 @@
> +From 14ba479fe707af2575fac4d2780e31a10264d295 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:32 -0500
> +Subject: [PATCH 05/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/ConvolutionLayer.cpp | 11 ++++++++++-
> + 1 file changed, 10 insertions(+), 1 deletion(-)
> +
> +diff --git a/tests/benchmark/NEON/ConvolutionLayer.cpp b/tests/benchmark/NEON/ConvolutionLayer.cpp
> +index ac27e7a..6cfb00d 100644
> +--- a/tests/benchmark/NEON/ConvolutionLayer.cpp
> ++++ b/tests/benchmark/NEON/ConvolutionLayer.cpp
> +@@ -52,9 +52,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
> +-
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + } // namespace
> +
> +@@ -197,6 +198,14 @@ REGISTER_FIXTURE_DATA_TEST_CASE(SqueezeNetWinogradLayer, NEWinogradConvolutionLa
> +
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ConvolutionLayer, NEGEMMConvolutionLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ConvolutionLayerDataset(),
> ++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
> ++ data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..3dfc2e7
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0006-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From c0c2b79824a52dee291b275a0f0a66746a99dbf5 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:50 -0500
> +Subject: [PATCH 06/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/DirectConvolutionLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/DirectConvolutionLayer.cpp b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> +index f94ef6b..afdcbad 100644
> +--- a/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> ++++ b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> +@@ -50,8 +50,10 @@ namespace
> + // Special data types for networks that need 5x5 direct convolution
> + #ifdef ARM_COMPUTE_ENABLE_F16
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* ARM_COMPUTE_ENABLE_F16 */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* ARM_COMPUTE_ENABLE_F16 */
> + } // namespace
> +
> +@@ -122,6 +124,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2DirectConvolutionLayer, NEDirectConvolutio
> +
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetDirectConvolutionLayer, NEDirectConvolutionLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetConvolutionLayerDataset(),
> ++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
> ++ data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..a1ed073
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0007-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From e8c375920cd21e6e151aabaa88b68358296bbe87 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:10 -0500
> +Subject: [PATCH 07/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/FullyConnectedLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/FullyConnectedLayer.cpp b/tests/benchmark/NEON/FullyConnectedLayer.cpp
> +index d0fb04f..9ee852d 100644
> +--- a/tests/benchmark/NEON/FullyConnectedLayer.cpp
> ++++ b/tests/benchmark/NEON/FullyConnectedLayer.cpp
> +@@ -47,8 +47,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + } // namespace
> +
> +@@ -108,6 +110,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV4FullyConnectedLayer, NEFully
> + framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1FullyConnectedLayer, NEFullyConnectedLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1FullyConnectedLayerDataset(),
> ++ data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..dcf0989
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0008-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,31 @@
> +From d657c798ec32c6a0f08384f725f3b8601d035300 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:26 -0500
> +Subject: [PATCH 08/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/GEMMLowp.cpp | 5 +++++
> + 1 file changed, 5 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/GEMMLowp.cpp b/tests/benchmark/NEON/GEMMLowp.cpp
> +index 4b13f31..c2c6bdf 100644
> +--- a/tests/benchmark/NEON/GEMMLowp.cpp
> ++++ b/tests/benchmark/NEON/GEMMLowp.cpp
> +@@ -50,6 +50,11 @@ REGISTER_FIXTURE_DATA_TEST_CASE(MatrixMultiplyGEMMLowp, NEGEMMLowpFixture, frame
> + REGISTER_FIXTURE_DATA_TEST_CASE(GoogleNetGEMMLowp, NEGEMMLowpFixture, framework::DatasetMode::NIGHTLY, datasets::GoogleNetGEMMDataset());
> +
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1GEMMLowp, NEGEMMLowpFixture, framework::DatasetMode::ALL, datasets::GoogLeNetInceptionV1GEMMDataset());
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..8498e27
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0009-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From 5e6db070b3cc563095991eaaab5d4c1fe8f12fcf Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:42 -0500
> +Subject: [PATCH 09/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/NormalizationLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/NormalizationLayer.cpp b/tests/benchmark/NEON/NormalizationLayer.cpp
> +index 0c9a864..3654ef5 100644
> +--- a/tests/benchmark/NEON/NormalizationLayer.cpp
> ++++ b/tests/benchmark/NEON/NormalizationLayer.cpp
> +@@ -44,8 +44,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + } // namespace
> + using NENormalizationLayerFixture = NormalizationLayerFixture<Tensor, NENormalizationLayer, Accessor>;
> +@@ -74,6 +76,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormal
> + framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormalizationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1NormalizationLayerDataset(),
> ++ data_types_tibenchmark),
> ++ framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..4c562e7
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0010-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,50 @@
> +From 76a4ca1417987c6542dd76b625d001ae8f30657e Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:58 -0500
> +Subject: [PATCH 10/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/PoolingLayer.cpp | 13 +++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/PoolingLayer.cpp b/tests/benchmark/NEON/PoolingLayer.cpp
> +index 8b7ee84..de93006 100644
> +--- a/tests/benchmark/NEON/PoolingLayer.cpp
> ++++ b/tests/benchmark/NEON/PoolingLayer.cpp
> +@@ -48,10 +48,13 @@ namespace benchmark
> + namespace
> + {
> + const auto data_layouts = framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC });
> ++const auto data_layouts_tibenchmark = framework::dataset::make("DataLayout", { DataLayout::NCHW});
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark = framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + } // namespace
> +
> +@@ -107,6 +110,16 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2PoolingLayer, NEPoolingLayerFixture, frame
> + framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::YOLOV2PoolingLayerDataset(), data_types), data_layouts), framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetPoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetPoolingLayerDataset(), data_types_tibenchmark), data_layouts_tibenchmark), framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1PoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1PoolingLayerDataset(), data_types_tibenchmark), data_layouts_tibenchmark), framework::dataset::make("Batches",
> ++ 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..fd3a9d1
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-library/0011-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,37 @@
> +From 012d21f475e1a004c2a7f688835a1b5416c3b8b6 Mon Sep 17 00:00:00 2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:20:10 -0500
> +Subject: [PATCH 11/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/framework/printers/PrettyPrinter.cpp | 4 ++--
> + 1 file changed, 2 insertions(+), 2 deletions(-)
> +
> +diff --git a/tests/framework/printers/PrettyPrinter.cpp b/tests/framework/printers/PrettyPrinter.cpp
> +index 3181951..2222286 100644
> +--- a/tests/framework/printers/PrettyPrinter.cpp
> ++++ b/tests/framework/printers/PrettyPrinter.cpp
> +@@ -83,7 +83,7 @@ void PrettyPrinter::print_run_footer()
> +
> + void PrettyPrinter::print_test_header(const TestInfo &info)
> + {
> +- *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << "\n";
> ++ *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << " ";
> + }
> +
> + void PrettyPrinter::print_test_footer()
> +@@ -125,7 +125,7 @@ void PrettyPrinter::print_measurements(const Profiler::MeasurementsMap &measurem
> + InstrumentsStats stats(instrument.second);
> +
> + *_stream << " ";
> +- *_stream << "AVG=" << stats.mean() << " " << stats.max().unit();
> ++ *_stream << "AVG= " << stats.mean() << " " << stats.max().unit();
> + if(instrument.second.size() > 1)
> + {
> + *_stream << ", STDDEV=" << arithmetic_to_string(stats.relative_standard_deviation(), 2) << " %";
> +--
> +1.9.1
> +
> --
> 1.9.1
>
> _______________________________________________
> meta-arago mailing list
> meta-arago@arago-project.org
> http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group.
2018-12-04 21:35 ` Denys Dmytriyenko
@ 2018-12-04 21:36 ` Su, Qin
0 siblings, 0 replies; 3+ messages in thread
From: Su, Qin @ 2018-12-04 21:36 UTC (permalink / raw)
To: Dmytriyenko, Denys; +Cc: meta-arago
I am really sorry, I saw it, I am doing v5 now.
-----Original Message-----
From: Dmytriyenko, Denys
Sent: Tuesday, December 04, 2018 4:36 PM
To: Su, Qin
Cc: meta-arago@arago-project.org
Subject: Re: [meta-arago] [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group.
The recipe itseld is missing.
On Tue, Dec 04, 2018 at 04:30:51PM -0500, Qin Su wrote:
> Updated to the latest version of 18.08.
> Updated LIC_FILES_CHKSUM as Copyright time changed in LICENSE from 2017 to 2017-2018.
> Updated to generate versioned library.
> Added benchmark preset group for squeezenet and inceptionnet.
> Cleaned up install built source package needed by ARMNN.
> Fixed symlinks lost issue in do_install().
>
> Signed-off-by: Qin Su <qsu@ti.com>
> ---
> Changes in PACTH v4:
> 1. Fixed the issue -"ERROR:QA Issue: armnn rdepends on arm-compute-library-dev" in building armnn.
Why/how? Better say - use "cp" instead of "install" to preserve symlinks...
> Changes in PATCH v3:
> 1. Fixed patch files permission issue
> 2. Cleaned up install source package required by armnn
>
> Changes in PATCH v2:
> 1. Added Upstream-Status in the commit log in patch files 2. Added
> description of license change in commit log 3. Removed unnecessary
> modification
> ---
> .../0001-add-ti-benchmark-test-group.patch | 61 +++++++++++++++++++
> .../0002-add-ti-benchmark-test-group.patch | 71 ++++++++++++++++++++++
> .../0003-add-ti-benchmark-test-group.patch | 39 ++++++++++++
> .../0004-add-ti-benchmark-test-group.patch | 49 +++++++++++++++
> .../0005-add-ti-benchmark-test-group.patch | 46 ++++++++++++++
> .../0006-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0007-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0008-add-ti-benchmark-test-group.patch | 31 ++++++++++
> .../0009-add-ti-benchmark-test-group.patch | 44 ++++++++++++++
> .../0010-add-ti-benchmark-test-group.patch | 50 +++++++++++++++
> .../0011-add-ti-benchmark-test-group.patch | 37 +++++++++++
> 11 files changed, 516 insertions(+)
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0001-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0002-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0003-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0004-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0005-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0006-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0007-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0008-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0009-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0010-add-ti-benchmark-test-group.patch
> create mode 100644
> meta-arago-extras/recipes-support/arm-compute-library/arm-compute-libr
> ary/0011-add-ti-benchmark-test-group.patch
>
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0001-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0001-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f00561e
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0001-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,61 @@
> +From 497a8abf760dbd9be715dfe6abf81c670fc85912 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:16:33 -0500
> +Subject: [PATCH 01/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../squeezenet/SqueezeNetActivationLayerDataset.h | 35
> +++++++++++++++++++++++
> + 1 file changed, 35 insertions(+)
> +
> +diff --git
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDat
> +aset.h
> +b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDat
> +aset.h
> +index 7f4bf4d..a66f473 100644
> +---
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLayerDat
> +aset.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetActivationLaye
> ++++ rDataset.h
> +@@ -71,6 +71,41 @@ public:
> + SqueezeNetActivationLayerDataset(SqueezeNetActivationLayerDataset &&) = default;
> + ~SqueezeNetActivationLayerDataset() = default;
> + };
> ++class TISqueezeNetActivationLayerDataset final : public
> ++
> ++framework::dataset::CartesianProductDataset<framework::dataset::Init
> ++ializerListDataset<TensorShape>,
> ++framework::dataset::SingletonDataset<ActivationLayerInfo>>
> ++{
> ++public:
> ++ TISqueezeNetActivationLayerDataset()
> ++ : CartesianProductDataset
> ++ {
> ++ framework::dataset::make("Shape", { // relu_conv1
> ++ TensorShape(114U, 114U, 64U),
> ++ // fire2/relu_squeeze1x1, fire3/relu_squeeze1x1
> ++ TensorShape(57U, 57U, 16U),
> ++ // fire2/relu_expand1x1, fire2/relu_expand3x3, fire3/relu_expand1x1, fire3/relu_expand3x3
> ++ TensorShape(57U, 57U, 64U),
> ++ // fire4/relu_squeeze1x1, fire5/relu_squeeze1x1
> ++ TensorShape(28U, 28U, 32U),
> ++ // fire4/relu_expand1x1, fire4/relu_expand3x3, fire5/relu_expand1x1, fire5/relu_expand3x3
> ++ TensorShape(28U, 28U, 128U),
> ++ // fire6/relu_squeeze1x1, fire7/relu_squeeze1x1
> ++ TensorShape(14U, 14U, 48U),
> ++ // fire6/relu_expand1x1, fire6/relu_expand3x3, fire7/relu_expand1x1, fire7/relu_expand3x3
> ++ TensorShape(14U, 14U, 192U),
> ++ // fire8/relu_squeeze1x1, fire9/relu_squeeze1x1
> ++ TensorShape(14U, 14U, 64U),
> ++ // fire8/relu_expand1x1, fire8/relu_expand3x3, fire9/relu_expand1x1, fire9/relu_expand3x3
> ++ TensorShape(14U, 14U, 256U),
> ++ // relu_conv10
> ++ TensorShape(14U, 14U, 1000U) }),
> ++ framework::dataset::make("Info", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
> ++ }
> ++ {
> ++ }
> ++ TISqueezeNetActivationLayerDataset(TISqueezeNetActivationLayerDataset &&) = default;
> ++ ~TISqueezeNetActivationLayerDataset() = default;
> ++};
> ++
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0002-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0002-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f891f74
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0002-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,71 @@
> +From 39ebc12e4de88a7839e33755c65cb9b9f0e8535f Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:17:19 -0500
> +Subject: [PATCH 02/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../squeezenet/SqueezeNetConvolutionLayerDataset.h | 45
> +++++++++++++++++++++++
> + 1 file changed, 45 insertions(+)
> +
> +diff --git
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDa
> +taset.h
> +b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDa
> +taset.h
> +index f98d90a..cba37eb 100644
> +---
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLayerDa
> +taset.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetConvolutionLay
> ++++ erDataset.h
> +@@ -96,6 +96,51 @@ public:
> + add_config(TensorShape(13U, 13U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(13U, 13U, 1000U), PadStrideInfo(1, 1, 0, 0));
> + }
> + };
> ++
> ++class TISqueezeNetConvolutionLayerDataset final : public
> ++ConvolutionLayerDataset {
> ++public:
> ++ TISqueezeNetConvolutionLayerDataset()
> ++ {
> ++ // conv1
> ++ add_config(TensorShape(227U, 227U, 3U), TensorShape(3U, 3U, 3U, 64U), TensorShape(64U), TensorShape(114U, 114U, 64U), PadStrideInfo(2, 2, 0, 0));
> ++ // fire2/squeeze1x1
> ++ add_config(TensorShape(57U, 57U, 64U), TensorShape(1U, 1U, 64U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire2/expand1x1, fire3/expand1x1
> ++ add_config(TensorShape(57U, 57U, 16U), TensorShape(1U, 1U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire2/expand3x3, fire3/expand3x3
> ++ add_config(TensorShape(57U, 57U, 16U), TensorShape(3U, 3U, 16U, 64U), TensorShape(64U), TensorShape(57U, 57U, 64U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire3/squeeze1x1
> ++ add_config(TensorShape(57U, 57U, 128U), TensorShape(1U, 1U, 128U, 16U), TensorShape(16U), TensorShape(57U, 57U, 16U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/squeeze1x1
> ++ add_config(TensorShape(28U, 28U, 128U), TensorShape(1U, 1U, 128U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/expand1x1, fire5/expand1x1
> ++ add_config(TensorShape(28U, 28U, 32U), TensorShape(1U, 1U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire4/expand3x3, fire5/expand3x3
> ++ add_config(TensorShape(28U, 28U, 32U), TensorShape(3U, 3U, 32U, 128U), TensorShape(128U), TensorShape(28U, 28U, 128U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire5/squeeze1x1
> ++ add_config(TensorShape(28U, 28U, 256U), TensorShape(1U, 1U, 256U, 32U), TensorShape(32U), TensorShape(28U, 28U, 32U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 256U), TensorShape(1U, 1U, 256U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/expand1x1, fire7/expand1x1
> ++ add_config(TensorShape(14U, 14U, 48U), TensorShape(1U, 1U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire6/expand3x3, fire7/expand3x3
> ++ add_config(TensorShape(14U, 14U, 48U), TensorShape(3U, 3U, 48U, 192U), TensorShape(192U), TensorShape(14U, 14U, 192U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire7/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 48U), TensorShape(48U), TensorShape(14U, 14U, 48U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 384U), TensorShape(1U, 1U, 384U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/expand1x1, fire9/expand1x1
> ++ add_config(TensorShape(14U, 14U, 64U), TensorShape(1U, 1U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 0, 0));
> ++ // fire8/expand3x3, fire9/expand3x3
> ++ add_config(TensorShape(14U, 14U, 64U), TensorShape(3U, 3U, 64U, 256U), TensorShape(256U), TensorShape(14U, 14U, 256U), PadStrideInfo(1, 1, 1, 1));
> ++ // fire9/squeeze1x1
> ++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 64U), TensorShape(64U), TensorShape(14U, 14U, 64U), PadStrideInfo(1, 1, 0, 0));
> ++ // conv10
> ++ add_config(TensorShape(14U, 14U, 512U), TensorShape(1U, 1U, 512U, 1000U), TensorShape(1000U), TensorShape(14U, 14U, 1000U), PadStrideInfo(1, 1, 0, 0));
> ++ }
> ++};
> ++
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0003-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0003-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..4a80548
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0003-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,39 @@
> +From 8c433526117e4aa5480efdbe5718ee54e9de2ff5 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:17:43 -0500
> +Subject: [PATCH 03/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + .../system_tests/squeezenet/SqueezeNetPoolingLayerDataset.h | 13
> ++++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDatase
> +t.h
> +b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDatase
> +t.h
> +index 9518ce5..a168d29 100644
> +---
> +a/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDatase
> +t.h
> ++++ b/tests/datasets/system_tests/squeezenet/SqueezeNetPoolingLayerDa
> ++++ taset.h
> +@@ -50,6 +50,19 @@ public:
> + add_config(TensorShape(27U, 27U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> + }
> + };
> ++class TISqueezeNetPoolingLayerDataset final : public
> ++PoolingLayerDataset {
> ++public:
> ++ TISqueezeNetPoolingLayerDataset()
> ++ {
> ++ // pool1
> ++ add_config(TensorShape(114U, 114U, 64U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ // pool3
> ++ add_config(TensorShape(57U, 57U, 128U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ // pool5
> ++ add_config(TensorShape(28U, 28U, 256U), PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)));
> ++ }
> ++};
> + } // namespace datasets
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0004-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0004-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..9693e6c
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0004-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,49 @@
> +From c2e6412d2aa9137b4b1b75d091ba9c8a53ede99b Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:05 -0500
> +Subject: [PATCH 04/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/ActivationLayer.cpp | 13 +++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/ActivationLayer.cpp
> +b/tests/benchmark/NEON/ActivationLayer.cpp
> +index 1c4ea21..7b47c11 100644
> +--- a/tests/benchmark/NEON/ActivationLayer.cpp
> ++++ b/tests/benchmark/NEON/ActivationLayer.cpp
> +@@ -51,9 +51,11 @@ namespace
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", { DataType::F16, DataType::F32 });
> + const auto data_types_mobilenet =
> +framework::dataset::make("DataType", { DataType::F16, DataType::F32,
> +DataType::QASYMM8 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
> + const auto data_types = framework::dataset::make("DataType", { DataType::F32 });
> + const auto data_types_mobilenet =
> + framework::dataset::make("DataType", { DataType::F32,
> + DataType::QASYMM8 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } // namespace
> +
> +@@ -127,6 +129,17 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2ActivationLayer, NEActivationLayerFixture,
> +
> +framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetActivationLayerDataset(), data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ActivationLayer, NEActivationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ActivationLayerDataset(), data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0005-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0005-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..f498ffa
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0005-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,46 @@
> +From 14ba479fe707af2575fac4d2780e31a10264d295 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:32 -0500
> +Subject: [PATCH 05/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/ConvolutionLayer.cpp | 11 ++++++++++-
> + 1 file changed, 10 insertions(+), 1 deletion(-)
> +
> +diff --git a/tests/benchmark/NEON/ConvolutionLayer.cpp
> +b/tests/benchmark/NEON/ConvolutionLayer.cpp
> +index ac27e7a..6cfb00d 100644
> +--- a/tests/benchmark/NEON/ConvolutionLayer.cpp
> ++++ b/tests/benchmark/NEON/ConvolutionLayer.cpp
> +@@ -52,9 +52,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", {
> +DataType::F16, DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ const auto
> +data_types = framework::dataset::make("DataType", { DataType::F32,
> +DataType::QASYMM8 });
> +-
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } // namespace
> +
> +@@ -197,6 +198,14 @@
> +REGISTER_FIXTURE_DATA_TEST_CASE(SqueezeNetWinogradLayer,
> +NEWinogradConvolutionLa
> +
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1ConvolutionLayer, NEGEMMConvolutionLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1ConvolutionLayerDataset(),
> ++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
> ++ data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0006-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0006-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..3dfc2e7
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0006-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From c0c2b79824a52dee291b275a0f0a66746a99dbf5 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:18:50 -0500
> +Subject: [PATCH 06/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/DirectConvolutionLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> +b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> +index f94ef6b..afdcbad 100644
> +--- a/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> ++++ b/tests/benchmark/NEON/DirectConvolutionLayer.cpp
> +@@ -50,8 +50,10 @@ namespace
> + // Special data types for networks that need 5x5 direct convolution
> +#ifdef ARM_COMPUTE_ENABLE_F16 const auto data_types =
> +framework::dataset::make("DataType", { DataType::F16, DataType::F32
> +});
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* ARM_COMPUTE_ENABLE_F16 */
> + const auto data_types = framework::dataset::make("DataType", {
> + DataType::F32 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* ARM_COMPUTE_ENABLE_F16 */
> + } // namespace
> +
> +@@ -122,6 +124,13 @@
> +REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2DirectConvolutionLayer,
> +NEDirectConvolutio
> +
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetDirectConvolutionLayer, NEDirectConvolutionLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::TISqueezeNetConvolutionLayerDataset(),
> ++ framework::dataset::make("ActivationInfo", ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))),
> ++ data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0007-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0007-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..a1ed073
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0007-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From e8c375920cd21e6e151aabaa88b68358296bbe87 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:10 -0500
> +Subject: [PATCH 07/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/FullyConnectedLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/FullyConnectedLayer.cpp
> +b/tests/benchmark/NEON/FullyConnectedLayer.cpp
> +index d0fb04f..9ee852d 100644
> +--- a/tests/benchmark/NEON/FullyConnectedLayer.cpp
> ++++ b/tests/benchmark/NEON/FullyConnectedLayer.cpp
> +@@ -47,8 +47,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", {
> +DataType::F16, DataType::F32 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ const auto
> + data_types = framework::dataset::make("DataType", { DataType::F32
> + });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } // namespace
> +
> +@@ -108,6 +110,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV4FullyConnectedLayer, NEFully
> +
> +framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1FullyConnectedLayer, NEFullyConnectedLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1FullyConnectedLayerDataset(),
> ++ data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0008-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0008-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..dcf0989
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0008-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,31 @@
> +From d657c798ec32c6a0f08384f725f3b8601d035300 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:26 -0500
> +Subject: [PATCH 08/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/GEMMLowp.cpp | 5 +++++
> + 1 file changed, 5 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/GEMMLowp.cpp
> +b/tests/benchmark/NEON/GEMMLowp.cpp
> +index 4b13f31..c2c6bdf 100644
> +--- a/tests/benchmark/NEON/GEMMLowp.cpp
> ++++ b/tests/benchmark/NEON/GEMMLowp.cpp
> +@@ -50,6 +50,11 @@
> +REGISTER_FIXTURE_DATA_TEST_CASE(MatrixMultiplyGEMMLowp,
> +NEGEMMLowpFixture, frame
> +REGISTER_FIXTURE_DATA_TEST_CASE(GoogleNetGEMMLowp, NEGEMMLowpFixture,
> +framework::DatasetMode::NIGHTLY, datasets::GoogleNetGEMMDataset());
> +
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1GEMMLowp,
> ++NEGEMMLowpFixture, framework::DatasetMode::ALL,
> ++datasets::GoogLeNetInceptionV1GEMMDataset());
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0009-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0009-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..8498e27
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0009-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,44 @@
> +From 5e6db070b3cc563095991eaaab5d4c1fe8f12fcf Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:42 -0500
> +Subject: [PATCH 09/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/NormalizationLayer.cpp | 9 +++++++++
> + 1 file changed, 9 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/NormalizationLayer.cpp
> +b/tests/benchmark/NEON/NormalizationLayer.cpp
> +index 0c9a864..3654ef5 100644
> +--- a/tests/benchmark/NEON/NormalizationLayer.cpp
> ++++ b/tests/benchmark/NEON/NormalizationLayer.cpp
> +@@ -44,8 +44,10 @@ namespace
> + {
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", {
> +DataType::F16, DataType::F32 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ const auto
> + data_types = framework::dataset::make("DataType", { DataType::F32
> + });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } // namespace
> +using NENormalizationLayerFixture = NormalizationLayerFixture<Tensor,
> +NENormalizationLayer, Accessor>; @@ -74,6 +76,13 @@ REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormal
> +
> +framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1NormalizationLayer, NENormalizationLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1NormalizationLayerDataset(),
> ++ data_types_tibenchmark),
> ++
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0010-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0010-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..4c562e7
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0010-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,50 @@
> +From 76a4ca1417987c6542dd76b625d001ae8f30657e Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:19:58 -0500
> +Subject: [PATCH 10/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/benchmark/NEON/PoolingLayer.cpp | 13 +++++++++++++
> + 1 file changed, 13 insertions(+)
> +
> +diff --git a/tests/benchmark/NEON/PoolingLayer.cpp
> +b/tests/benchmark/NEON/PoolingLayer.cpp
> +index 8b7ee84..de93006 100644
> +--- a/tests/benchmark/NEON/PoolingLayer.cpp
> ++++ b/tests/benchmark/NEON/PoolingLayer.cpp
> +@@ -48,10 +48,13 @@ namespace benchmark namespace { const auto
> +data_layouts = framework::dataset::make("DataLayout", {
> +DataLayout::NCHW, DataLayout::NHWC });
> ++const auto data_layouts_tibenchmark =
> ++framework::dataset::make("DataLayout", { DataLayout::NCHW});
> + #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
> + const auto data_types = framework::dataset::make("DataType", {
> + DataType::F16, DataType::F32, DataType::QASYMM8 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ const auto
> + data_types = framework::dataset::make("DataType", { DataType::F32,
> + DataType::QASYMM8 });
> ++const auto data_types_tibenchmark =
> ++framework::dataset::make("DataType", { DataType::F32 });
> + #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } // namespace
> +
> +@@ -107,6 +110,16 @@ REGISTER_FIXTURE_DATA_TEST_CASE(YOLOV2PoolingLayer, NEPoolingLayerFixture, frame
> +
> +framework::dataset::combine(framework::dataset::combine(framework::da
> +taset::combine(datasets::YOLOV2PoolingLayerDataset(), data_types),
> +data_layouts), framework::dataset::make("Batches", { 4, 8 })));
> + TEST_SUITE_END()
> + TEST_SUITE_END()
> ++
> ++TEST_SUITE(TIBENCHMARK_SQUEEZENET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(TISqueezeNetPoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
> ++
> ++framework::dataset::combine(framework::dataset::combine(framework::d
> ++ataset::combine(datasets::TISqueezeNetPoolingLayerDataset(),
> ++data_types_tibenchmark), data_layouts_tibenchmark),
> ++framework::dataset::make("Batches", 1)));
> ++TEST_SUITE_END()
> ++TEST_SUITE(TIBENCHMARK_INCEPTIONNET)
> ++REGISTER_FIXTURE_DATA_TEST_CASE(GoogLeNetInceptionV1PoolingLayer, NEPoolingLayerFixture, framework::DatasetMode::ALL,
> ++ framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(datasets::GoogLeNetInceptionV1PoolingLayerDataset(), data_types_tibenchmark), data_layouts_tibenchmark), framework::dataset::make("Batches",
> ++ 1)));
> ++TEST_SUITE_END()
> + } // namespace benchmark
> + } // namespace test
> + } // namespace arm_compute
> +--
> +1.9.1
> +
> diff --git
> a/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0011-add-ti-benchmark-test-group.patch
> b/meta-arago-extras/recipes-support/arm-compute-library/arm-compute-li
> brary/0011-add-ti-benchmark-test-group.patch
> new file mode 100644
> index 0000000..fd3a9d1
> --- /dev/null
> +++ b/meta-arago-extras/recipes-support/arm-compute-library/arm-comput
> +++ e-library/0011-add-ti-benchmark-test-group.patch
> @@ -0,0 +1,37 @@
> +From 012d21f475e1a004c2a7f688835a1b5416c3b8b6 Mon Sep 17 00:00:00
> +2001
> +From: Qin Su <qsu@ti.com>
> +Date: Tue, 27 Nov 2018 17:20:10 -0500
> +Subject: [PATCH 11/11] add TI benchmark test group
> +
> +Upstream-Status: Inappropriate [TI only test code]
> +
> +Signed-off-by: Qin Su <qsu@ti.com>
> +---
> + tests/framework/printers/PrettyPrinter.cpp | 4 ++--
> + 1 file changed, 2 insertions(+), 2 deletions(-)
> +
> +diff --git a/tests/framework/printers/PrettyPrinter.cpp
> +b/tests/framework/printers/PrettyPrinter.cpp
> +index 3181951..2222286 100644
> +--- a/tests/framework/printers/PrettyPrinter.cpp
> ++++ b/tests/framework/printers/PrettyPrinter.cpp
> +@@ -83,7 +83,7 @@ void PrettyPrinter::print_run_footer()
> +
> + void PrettyPrinter::print_test_header(const TestInfo &info) {
> +- *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << "\n";
> ++ *_stream << begin_color("2") << "Running [" << info.id << "] '" << info.name << "'" << end_color() << " ";
> + }
> +
> + void PrettyPrinter::print_test_footer()
> +@@ -125,7 +125,7 @@ void PrettyPrinter::print_measurements(const Profiler::MeasurementsMap &measurem
> + InstrumentsStats stats(instrument.second);
> +
> + *_stream << " ";
> +- *_stream << "AVG=" << stats.mean() << " " << stats.max().unit();
> ++ *_stream << "AVG= " << stats.mean() << " " << stats.max().unit();
> + if(instrument.second.size() > 1)
> + {
> + *_stream << ", STDDEV=" <<
> +arithmetic_to_string(stats.relative_standard_deviation(), 2) << " %";
> +--
> +1.9.1
> +
> --
> 1.9.1
>
> _______________________________________________
> meta-arago mailing list
> meta-arago@arago-project.org
> http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2018-12-04 21:36 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-12-04 21:30 [PATCH v4] arm-compute-library: bump SRCREV and add TI benchmark test group Qin Su
2018-12-04 21:35 ` Denys Dmytriyenko
2018-12-04 21:36 ` Su, Qin
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.