Merge branch 'develop' into mlir-c

.clang-tidy

@@ -4,7 +4,7 @@ CheckOptions:
   - key:             bugprone-unused-return-value.CheckedFunctions
     value:           '::std::async;::std::launder;::std::remove;::std::remove_if;::std::unique;::std::unique_ptr::release;::std::basic_string::empty;::std::vector::empty;::std::find;::std::find_if;::std::find_if_not;::std::all_of;::std::any_of;::std::none_of;::std::count;::std::count_if;::std::mismatch;::std::find_end;::std::find_first_of;::std::adjacent_find;::std::search;::std::search_n;::std::nth_element;::std::lower_bound;::std::upper_bound;::std::binary_search;::std::equal_range;::std::max;::std::max_element;::std::min;::std::min_element;::std::minmax;::std::minmax_element;::std::equal;::std::lexicographical_compare;::std::accumulate;::std::inner_product'
   - key:             cppcoreguidelines-macro-usage.AllowedRegexp
-    value:           'DEBUG|FALLTHROUGH|STRINGIZE|_HAS_|_THROW|_REQUIRES|_DECLARE_|_VISIT_|_REGISTER_|_GENERATE_|_DETAIL_|_TIDY_|_MANAGE_PTR|_MATCHER|DEVICE_SHARED|_WORKAROUND_'
+    value:           'DEBUG|ASSERT|ASSUME|UNREACHABLE|FALLTHROUGH|DEPRECATED|STRINGIZE|_HAS_|_THROW|_REQUIRES|_DECLARE_|_VISIT_|_REGISTER_|_GENERATE_|_DETAIL_|_TIDY_|_MANAGE_PTR|_MATCHER|DEVICE_SHARED|_WORKAROUND_'
   - key:             modernize-loop-convert.MinConfidence
     value:           risky   
   - key:             modernize-loop-convert.NamingStyle

.github/workflows/ci.yaml

@@ -15,7 +15,8 @@ jobs:
 
     steps:
     - name: Free space
-      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia*
+      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia* /usr/local/lib/android
+
     - uses: actions/checkout@v2
 
     # In this step, this action saves a list of existing images,
@@ -63,7 +64,7 @@ jobs:
 
     steps:
     - name: Free space
-      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia*
+      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia* /usr/local/lib/android
     - uses: actions/checkout@v2
 
     # In this step, this action saves a list of existing images,
@@ -108,7 +109,7 @@ jobs:
 
     steps:
     - name: Free space
-      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia*
+      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia* /usr/local/lib/android
     - uses: actions/checkout@v2
 
     # In this step, this action saves a list of existing images,
@@ -143,7 +144,7 @@ jobs:
 
     steps:
     - name: Free space
-      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia*
+      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia* /usr/local/lib/android
     - uses: actions/checkout@v2
     - name: Set up Python
       uses: actions/setup-python@v2
@@ -182,7 +183,7 @@ jobs:
 
     steps:
     - name: Free space
-      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia*
+      run: sudo rm -rf /usr/local/android /usr/share/dotnet /usr/local/share/boost /opt/ghc /usr/local/share/chrom* /usr/share/swift /usr/local/julia* /usr/local/lib/android
     - uses: actions/checkout@v2
     - name: Set up Python
       uses: actions/setup-python@v2

CMakeLists.txt

@@ -42,7 +42,7 @@ find_package(nlohmann_json 3.8.0 REQUIRED)
 
 include(ROCMSetupVersion)
 
-rocm_setup_version(VERSION 2.2)
+rocm_setup_version(VERSION 2.3)
 set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR})
 
 option( BUILD_SHARED_LIBS "Build as a shared library" ON )
@@ -93,11 +93,14 @@ rocm_enable_clang_tidy(
         modernize-*
         performance-*
         readability-*
-        -bugprone-signed-char-misuse
+        -bugprone-easily-swappable-parameters
+        -bugprone-implicit-widening-of-multiplication-result
         -bugprone-macro-parentheses
+        -bugprone-signed-char-misuse
         # Disable the aliased reserved identifiers
         -cert-dcl37-c
         -cert-dcl51-cpp
+        -cert-err33-c
         -cert-str34-c
         # Disable all alpha checks by default
         -clang-analyzer-alpha*
@@ -127,6 +130,7 @@ rocm_enable_clang_tidy(
         -cppcoreguidelines-pro-type-union-access
         -cppcoreguidelines-pro-type-vararg
         -cppcoreguidelines-special-member-functions
+        -cppcoreguidelines-virtual-class-destructor
         -google-readability-*
         -google-runtime-int
         -google-runtime-references
@@ -144,8 +148,10 @@ rocm_enable_clang_tidy(
         -readability-convert-member-functions-to-static
         -readability-else-after-return
         -readability-function-cognitive-complexity
+        -readability-identifier-length
         -readability-named-parameter
         -readability-redundant-string-init
+        -readability-suspicious-call-argument
         -readability-uppercase-literal-suffix
         -*-avoid-c-arrays
         -*-explicit-constructor
@@ -178,7 +184,7 @@ rocm_enable_cppcheck(
         style
         performance
         portability
-    SUPPRESS 
+    SUPPRESS
         ConfigurationNotChecked
         unmatchedSuppression
         unusedFunction
@@ -197,6 +203,8 @@ rocm_enable_cppcheck(
         useSmartPointer:*make_shared_array.hpp
         constParameter:*src/targets/gpu/*.cpp
         constParameter:*src/targets/gpu/*.hpp
+        # Suppress mlir_conv.cpp since this file will be deleted
+        *:*src/targets/gpu/mlir_conv.cpp
     FORCE
     INCONCLUSIVE
     RULE_FILE
@@ -216,6 +224,7 @@ rocm_enable_cppcheck(
         CPPCHECK=1
         __device__=
         __host__=
+        __global__=
 )
 
 enable_testing()

Dockerfile

-FROM ubuntu:18.04
+FROM ubuntu:20.04
 
 ARG PREFIX=/usr/local
 
@@ -6,7 +6,7 @@ ARG PREFIX=/usr/local
 RUN dpkg --add-architecture i386
 
 # Add rocm repository
-RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.5/ ubuntu main > /etc/apt/sources.list.d/rocm.list'
+RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/5.0.2/ ubuntu main > /etc/apt/sources.list.d/rocm.list'
 
 # Install dependencies
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
@@ -16,16 +16,12 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-
     cmake \
     curl \
     doxygen \
-    g++-5 \
     g++-7 \
     gdb \
     git \
     lcov \
     locales \
     pkg-config \
-    python \
-    python-dev \
-    python-pip \
     python3 \
     python3-dev \
     python3-pip \

cmake/Embed.cmake

@@ -39,6 +39,8 @@ function(generate_embed_source EMBED_NAME)
 
     file(WRITE "${PARSE_HEADER}" "
 #include <unordered_map>
+#include <string>
+#include <utility>
 const std::unordered_map<std::string, std::pair<const char*,const char*>>& ${EMBED_NAME}();
 ")
 
@@ -94,5 +96,6 @@ function(add_embed_library EMBED_NAME)
     generate_embed_source(${EMBED_NAME} SRC ${SRC_FILE} HEADER ${HEADER_FILE} OBJECTS ${OUTPUT_FILES} SYMBOLS ${SYMBOLS})
     add_library(${EMBED_NAME} STATIC ${OUTPUT_FILES} "${SRC_FILE}")
     target_include_directories(${EMBED_NAME} PUBLIC "${EMBED_DIR}/include")
+    target_compile_options(${EMBED_NAME} PRIVATE -Wno-reserved-identifier)
     set_target_properties(${EMBED_NAME} PROPERTIES POSITION_INDEPENDENT_CODE On)
 endfunction()

cmake/EnableCompilerWarnings.cmake

@@ -96,6 +96,7 @@ else()
                 -Wno-gnu-zero-variadic-macro-arguments
                 -Wno-missing-prototypes
                 -Wno-nested-anon-types
+                -Wno-option-ignored
                 -Wno-padded
                 -Wno-shorten-64-to-32
                 -Wno-sign-conversion

cmake/PythonModules.cmake

@@ -54,6 +54,10 @@ function(py_add_module NAME)
 
 endfunction()
 set(PYTHON_SEARCH_VERSIONS 2.7 3.5 3.6 3.7 3.8 3.9)
+set(PYTHON_DISABLE_VERSIONS "" CACHE STRING "")
+foreach(PYTHON_DISABLE_VERSION ${PYTHON_DISABLE_VERSIONS})
+    list(REMOVE_ITEM PYTHON_SEARCH_VERSIONS ${PYTHON_DISABLE_VERSION})
+endforeach()
 
 set(_PYTHON_VERSIONS)
 foreach(PYTHON_VERSION ${PYTHON_SEARCH_VERSIONS})

dev-requirements.txt

@@ -2,6 +2,6 @@ pfultz2/rocm-recipes
 facebook/zstd@v1.4.5 -X subdir -DCMAKE_DIR=build/cmake
 ccache@v4.1
 pcre,pfultz2/pcre@8.45 -H sha256:d6f7182602a775a7d500a0cedca6449af0400c6493951513046d17615ed0bf11
-danmar/cppcheck@2.6 -DHAVE_RULES=1
+danmar/cppcheck@2.8 -DHAVE_RULES=1
 RadeonOpenCompute/rocm-cmake@1ebf7e7bc61bb5e949c171562b421264065230a7 --build
 -f requirements.txt

doc/src/reference/py.rst

@@ -146,6 +146,13 @@ module
     :param list[module] mod_args: optional list of module arguments to the operator.
     :rtype instruction
 
+.. py:method:: add_literal(data)
+
+    Adds constant or literal data of provided shape into the module from python buffer which includes numpy array.    
+
+    :param py::buffer data: Python buffer or numpy array 
+    :rtype instruction 
+
 .. py:method:: add_parameter(name, shape)
     
     Adds a parameter to the module with provided name and shape.

examples/nlp/python_bert_squad/BERT-Squad.ipynb

@@ -62,7 +62,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!wget -nc https://github.com/onnx/models/raw/master/text/machine_comprehension/bert-squad/model/bertsquad-10.onnx"
+    "!wget -nc https://github.com/onnx/models/raw/main/text/machine_comprehension/bert-squad/model/bertsquad-10.onnx"
    ]
   },
   {

examples/nlp/python_bert_squad/README.md

@@ -23,7 +23,7 @@ unzip uncased_L-12_H-768_A-12.zip
 ```
 5) Get BERT ONNX model (bertsquad-10.onnx):
 ```
-wget https://github.com/onnx/models/raw/master/text/machine_comprehension/bert-squad/model/bertsquad-10.onnx
+wget https://github.com/onnx/models/raw/main/text/machine_comprehension/bert-squad/model/bertsquad-10.onnx
 ```
 6) Run the inference, it will compile and run the model on three questions and small data provided in `inputs.json`:
 ```

examples/nlp/python_bert_squad/requirements_bertsquad.txt

-tensorflow==2.5.3
+tensorflow==2.7.2
 onnxruntime
 tokenizers
\ No newline at end of file

examples/vision/python_resnet50/resnet50_inference.ipynb

@@ -106,8 +106,8 @@
    "outputs": [],
    "source": [
     "if not path.exists(\"./resnet50.onnx\"):\n",
-    "    !wget https://github.com/onnx/models/blob/master/vision/classification/resnet/model/resnet50-v2-7.onnx?raw=true\n",
-    "    !mv 'resnet50-v2-7.onnx?raw=true' resnet50.onnx"
+    "    !wget https://github.com/onnx/models/raw/main/vision/classification/resnet/model/resnet50-v2-7.onnx",
+    "    !mv resnet50-v2-7.onnx resnet50.onnx"
    ]
   },
   {

examples/vision/python_yolov4/yolov4_inference.ipynb

@@ -24,16 +24,16 @@
     "import os.path\n",
     "\n",
     "if not os.path.exists(\"./utilities/coco.names\"):\n",
-    "    !wget https://github.com/onnx/models/raw/master/vision/object_detection_segmentation/yolov4/dependencies/coco.names -P ./utilities/\n",
+    "    !wget https://github.com/onnx/models/raw/main/vision/object_detection_segmentation/yolov4/dependencies/coco.names -P ./utilities/\n",
     "if not os.path.exists(\"./utilities/yolov4_anchors.txt\"):\n",
-    "    !wget https://github.com/onnx/models/raw/master/vision/object_detection_segmentation/yolov4/dependencies/yolov4_anchors.txt -P ./utilities/\n",
+    "    !wget https://github.com/onnx/models/raw/main/vision/object_detection_segmentation/yolov4/dependencies/yolov4_anchors.txt -P ./utilities/\n",
     "if not os.path.exists(\"./utilities/input.jpg\"):\n",
     "    # The image used is from the COCO dataset (https://cocodataset.org/#explore)\n",
     "    # Other images can be tested by replacing the link below\n",
     "    image_link = \"https://farm3.staticflickr.com/2009/2306189268_88cc86b30f_z.jpg\"\n",
     "    !wget -O ./utilities/input.jpg $image_link\n",
     "if not os.path.exists(\"./utilities/yolov4.onnx\"):\n",
-    "    !wget https://github.com/onnx/models/raw/master/vision/object_detection_segmentation/yolov4/model/yolov4.onnx -P ./utilities/"
+    "    !wget https://github.com/onnx/models/raw/main/vision/object_detection_segmentation/yolov4/model/yolov4.onnx -P ./utilities/"
    ]
   },
   {

hip-clang.docker

-FROM ubuntu:18.04
+FROM ubuntu:20.04
 
 ARG PREFIX=/usr/local
 
@@ -6,7 +6,7 @@ ARG PREFIX=/usr/local
 RUN dpkg --add-architecture i386
 
 # Add rocm repository
-RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/4.5/ ubuntu main > /etc/apt/sources.list.d/rocm.list'
+RUN sh -c 'echo deb [arch=amd64 trusted=yes] http://repo.radeon.com/rocm/apt/5.0.2/ ubuntu main > /etc/apt/sources.list.d/rocm.list'
 
 # Install dependencies
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
@@ -20,9 +20,6 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-
     git \
     lcov \
     pkg-config \
-    python \
-    python-dev \
-    python-pip \
     python3 \
     python3-dev \
     python3-pip \

src/CMakeLists.txt

@@ -109,6 +109,7 @@ register_migraphx_ops(
     flatten
     floor
     gather
+    gathernd
     get_tuple_elem
     greater
     gru
@@ -162,10 +163,12 @@ register_migraphx_ops(
     round
     rsqrt
     scalar
-    scatter
-    scatternd_none
+    scatter_add
+    scatter_mul
+    scatter_none
     scatternd_add
     scatternd_mul
+    scatternd_none
     sigmoid
     sign
     sinh

src/adjust_allocation.cpp

@@ -8,9 +8,9 @@
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 
-void adjust_allocation::apply(module& p) const
+void adjust_allocation::apply(module& m) const
 {
-    for(auto ins : iterator_for(p))
+    for(auto ins : iterator_for(m))
     {
         // skip instruction with no input
         if(ins->inputs().empty())
@@ -27,13 +27,13 @@ void adjust_allocation::apply(module& p) const
         // of the instruction, reallocate and replace the previous one
         if(alias_ins->get_shape() == ins->get_shape())
             continue;
-        auto alloc_ins = p.insert_instruction(ins, model.allocate(ins->get_shape()));
-        p.replace_instruction(alias_ins, alloc_ins);
+        auto alloc_ins = m.insert_instruction(ins, model.allocate(ins->get_shape()));
+        m.replace_instruction(alias_ins, alloc_ins);
         // If the memory is an output parameter then copy the memory to the parameter
         if(alias_ins->name() == "@param")
         {
-            auto copy = p.insert_instruction(std::next(ins), make_op(model.copy()), ins, alias_ins);
-            auto tail = range(std::next(copy), p.end());
+            auto copy = m.insert_instruction(std::next(ins), make_op(model.copy()), ins, alias_ins);
+            auto tail = range(std::next(copy), m.end());
             for(auto i : iterator_for(tail))
             {
                 if(contains(i->inputs(), ins))

src/analyze_streams.cpp

@@ -14,31 +14,31 @@ bool happens_before(const std::vector<std::size_t>& e1, const std::vector<std::s
            not std::equal(e1.begin(), e1.end(), e2.begin(), e2.end(), std::greater_equal<>{});
 }
 
-std::vector<stream_race> analyze_streams(const module& p, const stream_model& m)
+std::vector<stream_race> analyze_streams(const module& m, const stream_model& strmm)
 {
     using vector_clock = std::vector<std::size_t>;
     std::vector<stream_race> races;
-    auto nstream = m.get_nstream();
+    auto nstream = strmm.get_nstream();
     std::vector<vector_clock> vclock(nstream, vector_clock(nstream));
     std::unordered_map<instruction_ref, vector_clock> timestamp;
     std::unordered_map<std::size_t, vector_clock> events;
-    for(auto ins : iterator_for(p))
+    for(auto ins : iterator_for(m))
     {
-        if(not m.has_stream(ins))
+        if(not strmm.has_stream(ins))
             continue;
-        std::size_t s = m.get_stream(ins);
+        std::size_t s = strmm.get_stream(ins);
         assert(s < nstream);
         assert(vclock.size() == nstream);
         assert(vclock[s].size() == nstream);
-        if(m.is_record(ins))
+        if(strmm.is_record(ins))
         {
             vclock[s][s]++;
-            auto event    = m.get_event_id(ins);
+            auto event    = strmm.get_event_id(ins);
             events[event] = vclock[s];
         }
-        else if(m.is_wait(ins))
+        else if(strmm.is_wait(ins))
         {
-            auto event = m.get_event_id(ins);
+            auto event = strmm.get_event_id(ins);
             if(not contains(events, event))
                 MIGRAPHX_THROW("Event is waited on before being recorded: " +
                                std::to_string(event));
@@ -57,21 +57,21 @@ std::vector<stream_race> analyze_streams(const module& p, const stream_model& m)
         }
         timestamp[ins] = vclock[s];
     }
-    for(auto ins : iterator_for(p))
+    for(auto ins : iterator_for(m))
     {
-        if(not m.has_stream(ins))
+        if(not strmm.has_stream(ins))
             continue;
         if(ins->inputs().empty())
             continue;
-        std::size_t s = m.get_stream(ins);
+        std::size_t s = strmm.get_stream(ins);
         // Find inputs from different streams
         std::vector<instruction_ref> inputs;
         fix([&](auto self, auto start) {
             for(auto input : start->inputs())
             {
-                if(not m.has_stream(input))
+                if(not strmm.has_stream(input))
                     self(input);
-                else if(m.get_stream(input) != s)
+                else if(strmm.get_stream(input) != s)
                     inputs.push_back(input);
             }
         })(ins);

src/api/api.cpp

@@ -401,7 +401,8 @@ extern "C" struct migraphx_instruction;
 struct migraphx_instruction
 {
     template <class... Ts>
-    migraphx_instruction(Ts&&... xs) : object(std::forward<Ts>(xs)...)
+    migraphx_instruction(Ts&&... xs)
+        : object(std::forward<Ts>(xs)...) // NOLINT(readability-redundant-member-init)
     {
     }
     migraphx::instruction_ref object;
@@ -411,7 +412,8 @@ extern "C" struct migraphx_instructions;
 struct migraphx_instructions
 {
     template <class... Ts>
-    migraphx_instructions(Ts&&... xs) : object(std::forward<Ts>(xs)...)
+    migraphx_instructions(Ts&&... xs)
+        : object(std::forward<Ts>(xs)...) // NOLINT(readability-redundant-member-init)
     {
     }
     std::vector<migraphx::instruction_ref> object;
@@ -421,7 +423,8 @@ extern "C" struct migraphx_modules;
 struct migraphx_modules
 {
     template <class... Ts>
-    migraphx_modules(Ts&&... xs) : object(std::forward<Ts>(xs)...)
+    migraphx_modules(Ts&&... xs)
+        : object(std::forward<Ts>(xs)...) // NOLINT(readability-redundant-member-init)
     {
     }
     std::vector<migraphx::module*> object;
@@ -1069,6 +1072,22 @@ migraphx_module_add_instruction_with_mod_args(migraphx_instruction_t* out,
     return api_error_result;
 }
 
+extern "C" migraphx_status migraphx_module_add_literal(migraphx_instruction_t* out,
+                                                       migraphx_module_t module,
+                                                       const_migraphx_shape_t shape,
+                                                       const char* buffer)
+{
+    auto api_error_result = migraphx::try_([&] {
+        if(module == nullptr)
+            MIGRAPHX_THROW(migraphx_status_bad_param, "Bad parameter module: Null pointer");
+        if(shape == nullptr)
+            MIGRAPHX_THROW(migraphx_status_bad_param, "Bad parameter shape: Null pointer");
+        *out = allocate<migraphx_instruction_t>(
+            (module->object).add_literal((shape->object), (buffer)));
+    });
+    return api_error_result;
+}
+
 extern "C" migraphx_status migraphx_module_add_parameter(migraphx_instruction_t* out,
                                                          migraphx_module_t module,
                                                          const char* name,
@@ -1691,6 +1710,16 @@ extern "C" migraphx_status migraphx_context_finish(const_migraphx_context_t cont
     return api_error_result;
 }
 
+extern "C" migraphx_status migraphx_context_get_queue(void** out, migraphx_context_t context)
+{
+    auto api_error_result = migraphx::try_([&] {
+        if(context == nullptr)
+            MIGRAPHX_THROW(migraphx_status_bad_param, "Bad parameter context: Null pointer");
+        *out = (context->object).get_queue().unsafe_get();
+    });
+    return api_error_result;
+}
+
 extern "C" migraphx_status
 migraphx_experimental_custom_op_destroy(migraphx_experimental_custom_op_t experimental_custom_op)
 {

src/api/include/migraphx/migraphx.h

@@ -258,6 +258,11 @@ migraphx_status migraphx_module_add_instruction_with_mod_args(migraphx_instructi
                                                               migraphx_instructions_t args,
                                                               migraphx_modules_t module_refs);
 
+migraphx_status migraphx_module_add_literal(migraphx_instruction_t* out,
+                                            migraphx_module_t module,
+                                            const_migraphx_shape_t shape,
+                                            const char* buffer);
+
 migraphx_status migraphx_module_add_parameter(migraphx_instruction_t* out,
                                               migraphx_module_t module,
                                               const char* name,
@@ -433,6 +438,8 @@ migraphx_status migraphx_quantize_int8(migraphx_program_t prog,
 
 migraphx_status migraphx_context_finish(const_migraphx_context_t context);
 
+migraphx_status migraphx_context_get_queue(void** out, migraphx_context_t context);
+
 migraphx_status
 migraphx_experimental_custom_op_destroy(migraphx_experimental_custom_op_t experimental_custom_op);