Add support OpenCL 3.0 non-uniform workgroup

* Add OpenCL version 3 detection.
* Use -cl-std=CL3.0 build option to support non-uniform workgroup
  when OpenCL 3 is detected and the feature is supported.

Resolves: COMPMID-5208
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: Ifd8cbae6b34228c07e761bcb94ee8f35bdf1bace
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7655
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

7 files changed, 63 insertions, 15 deletions

diff --git a/arm_compute/AclOpenClExt.h b/arm_compute/AclOpenClExt.h
index c349f76d86..ef80fd2443 100644
--- a/arm_compute/AclOpenClExt.h
+++ b/arm_compute/AclOpenClExt.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -27,7 +27,7 @@
 #include "arm_compute/AclTypes.h"
 
 #ifndef CL_TARGET_OPENCL_VERSION
-#define CL_TARGET_OPENCL_VERSION 200
+#define CL_TARGET_OPENCL_VERSION 300
 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
 #define CL_USE_DEPRECATED_OPENCL_1_2_APIS
 #endif /* CL_TARGET_OPENCL_VERSION */
diff --git a/arm_compute/core/CL/CLDevice.h b/arm_compute/core/CL/CLDevice.h
index 06aaac88f4..5e0f86e6d9 100644
--- a/arm_compute/core/CL/CLDevice.h
+++ b/arm_compute/core/CL/CLDevice.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2021 Arm Limited.
+ * Copyright (c) 2020-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -143,6 +143,32 @@ public:
         return _options.extensions.count(extension) != 0;
     }
 
+    /** Returns whether non-uniform workgroup is supported and the build options.
+     *
+     * If the feature is supported, the appropriate build options will be
+     * appended to the specified string.
+     *
+     * @return A tuple (supported, build_options) indicating whether the feature
+     *         is supported and the corresponding build options to enable it.
+     */
+    std::tuple<bool, std::string> is_non_uniform_workgroup_supported() const
+    {
+        if(version() == CLVersion::CL30 && get_cl_non_uniform_work_group_supported(_device))
+        {
+            return {true, " -cl-std=CL3.0 "};
+        }
+        else if(version() == CLVersion::CL20)
+        {
+            return {true, " -cl-std=CL2.0 "};
+        }
+        else if(supported("cl_arm_non_uniform_work_group_size"))
+        {
+            return {true, " -cl-arm-non-uniform-work-group-size "};
+        }
+
+        return {false, ""};
+    }
+
 private:
     cl::Device             _device;  /**< OpenCL device. */
     struct CLDeviceOptions _options; /**< OpenCL device options */
diff --git a/arm_compute/core/CL/CLHelpers.h b/arm_compute/core/CL/CLHelpers.h
index 729eb13398..94ec5d781b 100644
--- a/arm_compute/core/CL/CLHelpers.h
+++ b/arm_compute/core/CL/CLHelpers.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021 Arm Limited.
+ * Copyright (c) 2016-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -124,6 +124,14 @@ CLVersion get_cl_version(const cl::Device &device);
  */
 size_t get_cl_image_pitch_alignment(const cl::Device &device);
 
+/** Helper function to check whether non-uniform work group is supported
+ *
+ * @param[in] device A CL device
+ *
+ * @return True if the feature is supported
+ */
+bool get_cl_non_uniform_work_group_supported(const cl::Device &device);
+
 /** Helper function to check whether a given extension is supported
  *
  * @param[in] device         A CL device
diff --git a/arm_compute/core/CL/CLTypes.h b/arm_compute/core/CL/CLTypes.h
index 86335dae67..00b7cda2e1 100644
--- a/arm_compute/core/CL/CLTypes.h
+++ b/arm_compute/core/CL/CLTypes.h
@@ -41,7 +41,8 @@ enum class CLVersion
     CL10,   /* the OpenCL 1.0 */
     CL11,   /* the OpenCL 1.1 */
     CL12,   /* the OpenCL 1.2 */
-    CL20,   /* the OpenCL 2.0 and above */
+    CL20,   /* the OpenCL 2.x */
+    CL30,   /* the OpenCL 3.x */
     UNKNOWN /* unkown version */
 };
 
diff --git a/arm_compute/core/CL/OpenCL.h b/arm_compute/core/CL/OpenCL.h
index 4ff42c6b8a..93f7c318dd 100644
--- a/arm_compute/core/CL/OpenCL.h
+++ b/arm_compute/core/CL/OpenCL.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021 Arm Limited.
+ * Copyright (c) 2016-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -31,7 +31,7 @@
 #ifndef ARM_COMPUTE_NO_EXCEPTIONS
 #define CL_HPP_ENABLE_EXCEPTIONS
 #endif // ARM_COMPUTE_NO_EXCEPTIONS
-#define CL_TARGET_OPENCL_VERSION 200
+#define CL_TARGET_OPENCL_VERSION 300
 #define CL_HPP_TARGET_OPENCL_VERSION 110
 #define CL_HPP_MINIMUM_OPENCL_VERSION 110
 #pragma GCC diagnostic push
diff --git a/src/core/CL/CLCompileContext.cpp b/src/core/CL/CLCompileContext.cpp
index 2fed76555b..81eb748ab8 100644
--- a/src/core/CL/CLCompileContext.cpp
+++ b/src/core/CL/CLCompileContext.cpp
@@ -232,6 +232,8 @@ void CLCompileContext::set_context(cl::Context context)
 std::string CLCompileContext::generate_build_options(const StringSet &build_options_set, const std::string &kernel_path) const
 {
     std::string concat_str;
+    bool ext_supported = false;
+    std::string ext_buildopts;
 
 #if defined(ARM_COMPUTE_DEBUG_ENABLED)
     // Enable debug properties in CL kernels
@@ -257,13 +259,11 @@ std::string CLCompileContext::generate_build_options(const StringSet &build_opti
         concat_str += " -DARM_COMPUTE_OPENCL_DOT8_ACC_ENABLED=1 ";
     }
 
-    if(_device.version() == CLVersion::CL20)
-    {
-        concat_str += " -cl-std=CL2.0 ";
-    }
-    else if(_device.supported("cl_arm_non_uniform_work_group_size"))
+    std::tie(ext_supported, ext_buildopts) = _device.is_non_uniform_workgroup_supported();
+
+    if(ext_supported)
     {
-        concat_str += " -cl-arm-non-uniform-work-group-size ";
+        concat_str += ext_buildopts;
     }
     else
     {
diff --git a/src/core/CL/CLHelpers.cpp b/src/core/CL/CLHelpers.cpp
index 8685180b7f..5172a7730a 100644
--- a/src/core/CL/CLHelpers.cpp
+++ b/src/core/CL/CLHelpers.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021 Arm Limited.
+ * Copyright (c) 2016-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -256,7 +256,11 @@ bool dot8_acc_supported(const cl::Device &device)
 CLVersion get_cl_version(const cl::Device &device)
 {
     std::string version_str = device.getInfo<CL_DEVICE_VERSION>();
-    if(version_str.find("OpenCL 2") != std::string::npos)
+    if(version_str.find("OpenCL 3") != std::string::npos)
+    {
+        return CLVersion::CL30;
+    }
+    else if(version_str.find("OpenCL 2") != std::string::npos)
     {
         return CLVersion::CL20;
     }
@@ -388,6 +392,15 @@ size_t get_cl_image_pitch_alignment(const cl::Device &device)
     }
 }
 
+bool get_cl_non_uniform_work_group_supported(const cl::Device &device)
+{
+    cl_bool supported = CL_FALSE;
+
+    cl_int err = clGetDeviceInfo(device(), CL_DEVICE_NON_UNIFORM_WORK_GROUP_SUPPORT, sizeof(cl_bool), &supported, nullptr);
+
+    return (err == CL_SUCCESS && supported == CL_TRUE);
+}
+
 cl::Kernel create_kernel(const CLCompileContext &ctx, const std::string &kernel_name, const std::set<std::string> &build_opts)
 {
     opencl::ClKernelLibrary &klib = opencl::ClKernelLibrary::get();