Merge branch 'develop' into enable_navi_32_ci

.github/workflows/ci.yaml

@@ -281,6 +281,7 @@ jobs:
           -DBUILD_DEV=On \
           -DCMAKE_CXX_COMPILER_LAUNCHER=/usr/local/bin/ccache \
           -DCMAKE_C_COMPILER_LAUNCHER=/usr/local/bin/ccache \
+          -DCMAKE_CXX_FLAGS="-Werror" \
           -DGPU_TARGETS=gfx908 \
           ..
         make -j$(nproc) tests driver

CMakeLists.txt

@@ -27,20 +27,18 @@ if("${CMAKE_SOURCE_DIR}" STREQUAL "${CMAKE_BINARY_DIR}")
     message(FATAL_ERROR "The binary and source directroy cannot be the same")
 endif()
 
-get_property(_GENERATOR_IS_MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
+# Setup valid strings for build type
+if (NOT CMAKE_CONFIGURATION_TYPES)
+    set(CMAKE_CONFIGURATION_TYPES "Debug;Release;RelWithDebInfo;MinSizeRel" CACHE STRING "Configs")
+endif()
 
+get_property(MIGRAPHX_GENERATOR_IS_MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
 # This has to be initialized before the project() command appears
 # Set the default of CMAKE_BUILD_TYPE to be release, unless user specifies with -D.  MSVC_IDE does not use CMAKE_BUILD_TYPE
-if(_GENERATOR_IS_MULTI_CONFIG)
-    if (NOT CMAKE_CONFIGURATION_TYPES)
-        set(CMAKE_CONFIGURATION_TYPES "Debug;Release;RelWithDebInfo;MinSizeRel" CACHE STRING
-            "Available build types (configurations) on multi-config generators")
-    endif()
-else()
-    if(NOT CMAKE_BUILD_TYPE)
+if(NOT MIGRAPHX_GENERATOR_IS_MULTI_CONFIG)
     set(CMAKE_BUILD_TYPE Release CACHE STRING
         "Choose the type of build, options are: None Debug Release RelWithDebInfo MinSizeRel.")
-    endif()
+    set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS ${CMAKE_CONFIGURATION_TYPES})
 endif()
 
 set(CMAKE_INSTALL_PREFIX "/opt/rocm" CACHE PATH "")
@@ -53,6 +51,12 @@ include(CTest)
 find_package(ROCM REQUIRED)
 find_package(Threads REQUIRED)
 
+if(WIN32)
+option(MIGRAPHX_ENABLE_PYTHON "Enable python bindings" OFF)
+else()
+option(MIGRAPHX_ENABLE_PYTHON "Enable python bindings" ON)
+endif()
+
 find_path(HALF_INCLUDE_DIR half.hpp PATH_SUFFIXES half)
 if (NOT HALF_INCLUDE_DIR)
     message(FATAL_ERROR "Could not find half.hpp - Please check that the install path of half.hpp has been added to CMAKE_PREFIX_PATH")
@@ -71,8 +75,8 @@ include(ROCMSetupVersion)
 
 option(BUILD_DEV "Build for development purpose only" OFF)
 
-rocm_setup_version(VERSION 2.7.0)
-set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}.${PROJECT_VERSION_PATCH})
+rocm_setup_version(VERSION 2.8.0)
+set(MIGRAPHX_SO_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR})
 
 option( BUILD_SHARED_LIBS "Build as a shared library" ON )
 
@@ -261,8 +265,6 @@ rocm_enable_cppcheck(
         MIGRAPHX_USE_CLANG_TIDY
 )
 
-enable_testing()
-
 include(ROCMCreatePackage)
 include(ROCMTest)
 
@@ -283,7 +285,6 @@ file(MAKE_DIRECTORY ${DEST_DIR}/lib/onnx_migraphx)
 foreach(py_file ${backend_files})
     configure_file(${py_file} ${DEST_DIR}/lib/onnx_migraphx/. COPYONLY)
 endforeach(py_file)
-configure_file(${CMAKE_SOURCE_DIR}/test/py/onnx_backend_test.py ${DEST_DIR}/onnx_backend_test.py COPYONLY)
 
 rocm_create_package(
     NAME MIGraphX

Jenkinsfile

+def getgputargets() {
+    targets="gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102"
+    return targets
+}
 
 // def rocmtestnode(variant, name, body, args, pre) {
 def rocmtestnode(Map conf) {
@@ -105,46 +109,55 @@ def rocmnode(name, body) {
     }
 }
 
-rocmtest clang_debug: rocmnode('cdna') { cmake_build ->
+rocmtest clang_debug: rocmnode('mi100+') { cmake_build ->
     stage('hipRTC Debug') {
+        // Disable MLIR since it doesnt work with all ub sanitizers
+        withEnv(['MIGRAPHX_DISABLE_MLIR=1']) {
             def sanitizers = "undefined"
             def debug_flags = "-g -O2 -fsanitize=${sanitizers} -fno-sanitize-recover=${sanitizers}"
-        cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}' -DMIGRAPHX_USE_HIPRTC=On -DGPU_TARGETS=$(/opt/rocm/bin/rocminfo | grep -o -m1 'gfx.*')", gpu_debug: true)
+            def gpu_targets = getgputargets()
+            cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}' -DMIGRAPHX_USE_HIPRTC=On -DGPU_TARGETS='${gpu_targets}'", gpu_debug: true)
+        }
     }
 }, clang_release: rocmnode('mi100+') { cmake_build ->
     stage('Hip Clang Release') {
-        cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DGPU_TARGETS=$(/opt/rocm/bin/rocminfo | grep -o -m1 'gfx.*')")
+        def gpu_targets = getgputargets()
+        cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DGPU_TARGETS='${gpu_targets}'")
         stash includes: 'build/*.deb', name: 'migraphx-package'
     }
 // }, hidden_symbols: rocmnode('cdna') { cmake_build ->
 //     stage('Hidden symbols') {
 //         cmake_build(flags: "-DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_VISIBILITY_PRESET=hidden -DCMAKE_C_VISIBILITY_PRESET=hidden")
 //     }
-}, all_targets_debug : rocmnode('cdna') { cmake_build ->
+}, all_targets_debug : rocmnode('mi100+') { cmake_build ->
     stage('All targets Release') {
-        cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DMIGRAPHX_ENABLE_FPGA=On -DGPU_TARGETS=$(/opt/rocm/bin/rocminfo | grep -o -m1 'gfx.*')") 
+        def gpu_targets = getgputargets()
+        cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DMIGRAPHX_ENABLE_GPU=On -DMIGRAPHX_ENABLE_CPU=On -DMIGRAPHX_ENABLE_FPGA=On -DGPU_TARGETS='${gpu_targets}'")
     }
-}, mlir_debug: rocmnode('cdna') { cmake_build ->
+}, mlir_debug: rocmnode('mi100+') { cmake_build ->
     stage('MLIR Debug') {
-        withEnv(['MIGRAPHX_ENABLE_MLIR=1']) {
+        withEnv(['MIGRAPHX_ENABLE_EXTRA_MLIR=1']) {
             def sanitizers = "undefined"
             // Note: the -fno-sanitize= is copied from upstream LLVM_UBSAN_FLAGS.
             def debug_flags_cxx = "-g -O2 -fsanitize=${sanitizers} -fno-sanitize=vptr,function -fno-sanitize-recover=${sanitizers}"
             def debug_flags = "-g -O2 -fsanitize=${sanitizers} -fno-sanitize=vptr -fno-sanitize-recover=${sanitizers}"
-            cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_MLIR=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags_cxx}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}' -DGPU_TARGETS=$(/opt/rocm/bin/rocminfo | grep -o -m1 'gfx.*')")
+            def gpu_targets = getgputargets()
+            cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_MLIR=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags_cxx}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}' -DGPU_TARGETS='${gpu_targets}'")
         }
     }
-}, ck_release: rocmnode('mi100+') { cmake_build ->
-    stage('CK Release') {
-        withEnv(['MIGRAPHX_ENABLE_CK=1', 'MIGRAPHX_TUNE_CK=1']) {
-            cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DGPU_TARGETS=$(/opt/rocm/bin/rocminfo | grep -o -m1 'gfx.*')")
+}, ck_hiprtc: rocmnode('mi100+') { cmake_build ->
+    stage('CK hipRTC') {
+        withEnv(['MIGRAPHX_ENABLE_CK=1', 'MIGRAPHX_TUNE_CK=1', 'MIGRAPHX_DISABLE_MLIR=1']) {
+            def gpu_targets = getgputargets()
+            cmake_build(flags: "-DCMAKE_BUILD_TYPE=release -DMIGRAPHX_USE_HIPRTC=On -DGPU_TARGETS='${gpu_targets}'")
         }
     }
 }, clang_asan: rocmnode('nogpu') { cmake_build ->
     stage('Clang ASAN') {
         def sanitizers = "undefined,address"
         def debug_flags = "-g -O2 -fno-omit-frame-pointer -fsanitize=${sanitizers} -fno-sanitize-recover=${sanitizers}"
-        cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=Off -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}'")
+        def gpu_targets = getgputargets()
+        cmake_build(flags: "-DCMAKE_BUILD_TYPE=debug -DMIGRAPHX_ENABLE_PYTHON=Off -DMIGRAPHX_ENABLE_GPU=Off -DMIGRAPHX_ENABLE_CPU=On -DCMAKE_CXX_FLAGS_DEBUG='${debug_flags}' -DCMAKE_C_FLAGS_DEBUG='${debug_flags}' -DGPU_TARGETS='${gpu_targets}'")
     }
 }, clang_release_navi: rocmnode('navi32') { cmake_build ->
     stage('HIP Clang Release Navi32') {

cmake/Embed.cmake

@@ -77,16 +77,17 @@ function(generate_embed_source EMBED_NAME)
         list(GET PARSE_FILES ${idx} FILE)
 
         set(START_SYMBOL "_binary_${SYMBOL}_start")
-        set(END_SYMBOL "_binary_${SYMBOL}_end")
+        set(LENGTH_SYMBOL "_binary_${SYMBOL}_length")
         if(EMBED_USE_LD)
             string(APPEND EXTERNS "
-                extern const char ${START_SYMBOL}[];
-                extern const char ${END_SYMBOL}[];
+extern const char ${START_SYMBOL}[];
+extern const size_t _binary_${SYMBOL}_size;
+const auto ${LENGTH_SYMBOL} = reinterpret_cast<size_t>(&_binary_${SYMBOL}_size);
             ")
         else()
             string(APPEND EXTERNS "
-                extern const char ${START_SYMBOL}[];
-                extern const char* ${END_SYMBOL};
+extern const char ${START_SYMBOL}[];
+extern const size_t ${LENGTH_SYMBOL};
             ")
         endif()
 
@@ -97,23 +98,22 @@ function(generate_embed_source EMBED_NAME)
         endif()
 
         string(APPEND INIT_KERNELS "
-            { \"${BASE_NAME}\", { ${START_SYMBOL}, ${END_SYMBOL}} },
-        ")
+            { \"${BASE_NAME}\", { ${START_SYMBOL}, ${LENGTH_SYMBOL}} },")
     endforeach()
 
     file(WRITE "${PARSE_HEADER}" "
+#include <string_view>
 #include <unordered_map>
-#include <string>
 #include <utility>
-const std::unordered_map<std::string, std::pair<const char*,const char*>>& ${EMBED_NAME}();
+std::unordered_map<std::string_view, std::string_view> ${EMBED_NAME}();
 ")
 
     file(WRITE "${PARSE_SRC}" "
 #include <${EMBED_NAME}.hpp>
 ${EXTERNS}
-const std::unordered_map<std::string, std::pair<const char*,const char*>>& ${EMBED_NAME}()
+std::unordered_map<std::string_view, std::string_view> ${EMBED_NAME}()
 {
-    static const std::unordered_map<std::string, std::pair<const char*,const char*>> result = {${INIT_KERNELS}};
+    static std::unordered_map<std::string_view, std::string_view> result = {${INIT_KERNELS}};
     return result;
 }
 ")
@@ -154,9 +154,10 @@ function(embed_file OUTPUT_FILE OUTPUT_SYMBOL FILE)
             # removes trailing comma
             string(REGEX REPLACE ", $" "" ARRAY_VALUES ${ARRAY_VALUES})
             file(WRITE "${OUT_FILE}" "
-                extern const char _binary_${SYMBOL}_start[] = { ${ARRAY_VALUES} };
-                extern const char* _binary_${SYMBOL}_end = _binary_${SYMBOL}_start + sizeof(_binary_${SYMBOL}_start);
-            \n")
+#include <cstddef>
+extern const char _binary_${SYMBOL}_start[] = { ${ARRAY_VALUES} };
+extern const size_t _binary_${SYMBOL}_length = sizeof(_binary_${SYMBOL}_start);
+")
         endif()
     endforeach()
 endfunction()

docs/.sphinx/requirements.txt

@@ -35,7 +35,7 @@ fastjsonschema==2.16.3
     # via rocm-docs-core
 gitdb==4.0.10
     # via gitpython
-gitpython==3.1.32
+gitpython==3.1.37
     # via rocm-docs-core
 idna==3.4
     # via requests
@@ -87,7 +87,7 @@ requests==2.28.2
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==0.24.2
+rocm-docs-core==0.26.0
     # via -r requirements.in
 smmap==5.0.0
     # via gitdb
@@ -130,7 +130,7 @@ sphinxcontrib-serializinghtml==1.1.5
     # via sphinx
 typing-extensions==4.5.0
     # via pydata-sphinx-theme
-urllib3==1.26.15
+urllib3==1.26.18
     # via requests
 wrapt==1.15.0
     # via deprecated

requirements.txt

@@ -28,5 +28,5 @@ ROCmSoftwarePlatform/half@rocm-5.6.0
 pybind/pybind11@d159a563383d10c821ba7b2a71905d1207db6de4 --build
 msgpack/msgpack-c@cpp-3.3.0 -DMSGPACK_BUILD_TESTS=Off
 sqlite3@3.17 -DCMAKE_POSITION_INDEPENDENT_CODE=On
-ROCmSoftwarePlatform/composable_kernel@5172ec5280f14974beee2acf1af1db3b2670244c -DCK_BUILD_JIT_LIB=On -DCMAKE_POSITION_INDEPENDENT_CODE=On
-ROCmSoftwarePlatform/rocMLIR@a48dfb1f163fb0b38369e73e580968b72e85b594 -DBUILD_FAT_LIBROCKCOMPILER=On
+ROCmSoftwarePlatform/composable_kernel@70eefcf4f263aa5c25f3c9ff0db8f6f199ef0fb9 -DCK_BUILD_JIT_LIB=On -DCMAKE_POSITION_INDEPENDENT_CODE=On
+ROCmSoftwarePlatform/rocMLIR@507bb94ce7873786486d296ec81d2eadaab49003 -DBUILD_FAT_LIBROCKCOMPILER=On
\ No newline at end of file

src/CMakeLists.txt

@@ -282,7 +282,9 @@ add_subdirectory(driver)
 add_subdirectory(onnx)
 add_subdirectory(tf)
 
+if(MIGRAPHX_ENABLE_PYTHON)
 add_subdirectory(py)
+endif()
 add_subdirectory(targets/ref)
 target_link_libraries(migraphx_all_targets INTERFACE migraphx_ref)
 if(MIGRAPHX_ENABLE_CPU)

src/api/CMakeLists.txt

@@ -32,6 +32,10 @@ migraphx_generate_export_header(migraphx_c DIRECTORY migraphx/api)
 # bumped when binary compatibility is broken. 
 rocm_set_soversion(migraphx_c 3.0)   
 
+if(BUILD_TESTING)
+    target_compile_definitions(migraphx_c PRIVATE MIGRAPHX_BUILD_TESTING)
+endif()
+
 rocm_clang_tidy_check(migraphx_c)
 target_link_libraries(migraphx_c PRIVATE migraphx migraphx_tf migraphx_onnx)
 

src/api/api.cpp

@@ -38,26 +38,32 @@
 #include <migraphx/register_op.hpp>
 #include <migraphx/json.hpp>
 #include <migraphx/convert_to_json.hpp>
+#include <array>
 #include <algorithm>
 #include <cstdarg>
+
 namespace migraphx {
 
+#ifdef MIGRAPHX_BUILD_TESTING
 static thread_local bool disable_exception_catch = false; // NOLINT
 
 extern "C" MIGRAPHX_C_EXPORT void migraphx_test_private_disable_exception_catch(bool b)
 {
     disable_exception_catch = b;
 }
+#endif
 
 template <class F>
 migraphx_status try_(F f, bool output = true) // NOLINT
 {
+#ifdef MIGRAPHX_BUILD_TESTING
     if(disable_exception_catch)
     {
         f();
     }
     else
     {
+#endif
         try
         {
             f();
@@ -81,7 +87,9 @@ migraphx_status try_(F f, bool output = true) // NOLINT
         {
             return migraphx_status_unknown_error;
         }
+#ifdef MIGRAPHX_BUILD_TESTING
     }
+#endif
     return migraphx_status_success;
 }
 

src/api/include/migraphx/migraphx.h

@@ -26,6 +26,7 @@
 
 #include <stdlib.h>
 #include <stdbool.h>
+#include <stdint.h>
 
 #include <migraphx/api/export.h>
 

src/api/include/migraphx/migraphx.hpp

@@ -66,7 +66,7 @@ template <class PrivateMigraphTypeNameProbe>
 std::string compute_type_name()
 {
     std::string name;
-#ifdef _MSC_VER
+#if defined(_MSC_VER) && !defined(__clang__)
     name = typeid(PrivateMigraphTypeNameProbe).name();
     name = name.substr(7);
 #else

src/compile_src.cpp

@@ -46,7 +46,7 @@ std::vector<char> src_compiler::compile(const std::vector<src_file>& srcs) const
         fs::path full_path   = td.path / src.path;
         fs::path parent_path = full_path.parent_path();
         fs::create_directories(parent_path);
-        write_buffer(full_path.string(), src.content.first, src.len());
+        write_buffer(full_path.string(), src.content.data(), src.content.size());
         if(src.path.extension().string() == ".cpp")
         {
             params += " " + src.path.filename().string();

src/cpp_generator.cpp

@@ -213,13 +213,13 @@ cpp_generator::function cpp_generator::generate_module(const module& m,
                 ins->get_literal().visit([&](auto v) {
                     assert(v.size() == 1);
                     auto x = v.front();
-                    if(std::isinf(x))
+                    if(std::isinf(static_cast<double>(x)))
                     {
                         string_literal = "__builtin_huge_val()";
                         if(x < 0)
                             string_literal = "-__builtin_huge_val()";
                     }
-                    else if(std::isnan(x))
+                    else if(std::isnan(static_cast<double>(x)))
                         string_literal = "__builtin_nan()";
                     else
                         string_literal = ins->get_literal().to_string();

src/driver/CMakeLists.txt

@@ -48,7 +48,12 @@ rocm_clang_tidy_check(driver)
 file(STRINGS "${CMAKE_SOURCE_DIR}/test/onnx/.onnxrt-commit" String_output)
 target_compile_definitions(driver PUBLIC MIGRAPHX_ORT_SHA1="${String_output}")
 
-target_link_libraries(driver migraphx_all_targets migraphx_onnx migraphx_tf migraphx_py)
+target_link_libraries(driver migraphx_all_targets migraphx_onnx migraphx_tf)
+
+if(MIGRAPHX_ENABLE_PYTHON)
+    target_link_libraries(driver migraphx_py)
+    target_compile_definitions(driver PRIVATE MIGRAPHX_ENABLE_PYTHON)
+endif()
 
 rocm_install_targets(
   TARGETS driver

src/driver/argument_parser.hpp

@@ -187,6 +187,13 @@ struct value_parser
     }
 };
 
+// version for std::optional object
+template <class T>
+struct value_parser<std::optional<T>>
+{
+    static T apply(const std::string& x) { return value_parser<T>::apply(x); }
+};
+
 struct argument_parser
 {
     struct argument

src/driver/main.cpp

@@ -32,7 +32,9 @@
 
 #include <migraphx/tf.hpp>
 #include <migraphx/onnx.hpp>
+#ifdef MIGRAPHX_ENABLE_PYTHON
 #include <migraphx/py.hpp>
+#endif
 #include <migraphx/stringutils.hpp>
 #include <migraphx/convert_to_json.hpp>
 #include <migraphx/load_save.hpp>
@@ -281,10 +283,12 @@ struct loader
                 options.format = "json";
                 p              = migraphx::load(file, options);
             }
+#ifdef MIGRAPHX_ENABLE_PYTHON
             else if(file_type == "py")
             {
                 p = migraphx::load_py(file);
             }
+#endif
             else if(file_type == "migraphx")
             {
                 p = migraphx::load(file);
@@ -536,19 +540,17 @@ struct params : command<params>
 struct verify : command<verify>
 {
     compiler c;
-    migraphx::verify::tolerance tols;
+    std::optional<double> rms_tol;
+    std::optional<double> atol;
+    std::optional<double> rtol;
     bool per_instruction = false;
     bool reduce          = false;
     void parse(argument_parser& ap)
     {
         c.parse(ap);
-        ap(tols.rms_tol, {"--rms-tol"}, ap.help("Tolerance for the RMS error (Default: 0.001)"));
-        ap(tols.atol,
-           {"--atol"},
-           ap.help("Tolerance for the elementwise absolute difference (Default: 0.001)"));
-        ap(tols.rtol,
-           {"--rtol"},
-           ap.help("Tolerance for the elementwise relative difference (Default: 0.001)"));
+        ap(rms_tol, {"--rms-tol"}, ap.help("Tolerance for the RMS error"));
+        ap(atol, {"--atol"}, ap.help("Tolerance for the elementwise absolute difference"));
+        ap(rtol, {"--rtol"}, ap.help("Tolerance for the elementwise relative difference"));
         ap(per_instruction,
            {"-i", "--per-instruction"},
            ap.help("Verify each instruction"),
@@ -567,9 +569,18 @@ struct verify : command<verify>
 
         auto quantize = precision::fp32;
         if(c.to_fp16)
+        {
             quantize = precision::fp16;
+        }
         if(c.to_int8)
+        {
             quantize = precision::int8;
+        }
+
+        auto tols = get_tolerances(p, quantize, rms_tol, atol, rtol);
+        std::cout << "rms_tol: " << tols.rms_tol << std::endl;
+        std::cout << "atol: " << tols.atol << std::endl;
+        std::cout << "rtol: " << tols.rtol << std::endl;
 
         if(per_instruction)
         {

src/driver/verify.cpp

@@ -36,6 +36,42 @@ namespace migraphx {
 namespace driver {
 inline namespace MIGRAPHX_INLINE_NS {
 
+/**
+ * Gives tolerances based on user input (`rms_tol`, `atol`, `rtol` parameters) and defaults.
+ * Sets to fp16 tolerances if `quantize` input is fp16 or any fp16 instruction in found in the
+ * model.
+ */
+verify::tolerance get_tolerances(const program& p,
+                                 precision quantize,
+                                 std::optional<double> rms_tol,
+                                 std::optional<double> atol,
+                                 std::optional<double> rtol)
+{
+    bool has_fp16 = any_of(p.get_modules(), [](auto&& m) {
+        return any_of(*m, [](auto&& ins) { return (ins.get_shape().type() == shape::half_type); });
+    });
+    migraphx::verify::tolerance result{};
+    if(has_fp16 or quantize == precision::fp16)
+    {
+        result.rms_tol = 8e-2;
+        result.atol    = 4e-2;
+        result.rtol    = 4e-2;
+    }
+    if(rms_tol)
+    {
+        result.rms_tol = *rms_tol;
+    }
+    if(atol)
+    {
+        result.atol = *atol;
+    }
+    if(rtol)
+    {
+        result.rtol = *rtol;
+    }
+    return result;
+}
+
 std::vector<argument> run_ref(program p, const parameter_map& inputs)
 {
     p.compile(migraphx::make_target("ref"));

src/driver/verify.hpp

@@ -32,6 +32,12 @@ namespace migraphx {
 namespace driver {
 inline namespace MIGRAPHX_INLINE_NS {
 
+verify::tolerance get_tolerances(const program& p,
+                                 precision quantize,
+                                 std::optional<double> rms_tol,
+                                 std::optional<double> atol,
+                                 std::optional<double> rtol);
+
 void verify_program(const std::string& name,
                     const program& p,
                     const target& t,

src/dynamic_loader.cpp

@@ -27,11 +27,20 @@
 #include <migraphx/file_buffer.hpp>
 #include <migraphx/tmp_dir.hpp>
 #include <utility>
+
+#ifdef _WIN32
+// cppcheck-suppress definePrefix
+#define WIN32_LEAN_AND_MEAN
+#include <Windows.h>
+#else
 #include <dlfcn.h>
+#endif
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 
+#ifndef _WIN32
+
 void check_load_error(bool flush = false)
 {
     char* error_msg = dlerror();
@@ -81,6 +90,48 @@ fs::path dynamic_loader::path(void* address)
     return p;
 }
 
+#else
+
+struct dynamic_loader_impl
+{
+    dynamic_loader_impl() = default;
+    dynamic_loader_impl(const fs::path& p, tmp_dir t = {})
+        : handle{LoadLibrary(p.string().c_str())}, temp{std::move(t)}
+    {
+        if(handle == nullptr)
+        {
+            MIGRAPHX_THROW("Error loading DLL: " + p.string() + " (" +
+                           std::to_string(GetLastError()) + ")");
+        }
+    }
+
+    dynamic_loader_impl(const dynamic_loader_impl&)            = delete;
+    dynamic_loader_impl& operator=(const dynamic_loader_impl&) = delete;
+
+    dynamic_loader_impl(dynamic_loader_impl&&) = default;
+
+    ~dynamic_loader_impl()
+    {
+        if(handle != nullptr)
+        {
+            FreeLibrary(handle);
+        }
+    }
+
+    static std::shared_ptr<dynamic_loader_impl> from_buffer(const char* image, std::size_t size)
+    {
+        auto t = tmp_dir{"migx-dynload"};
+        auto f = t.path / "tmp.dll";
+        write_buffer(f.string(), image, size);
+        return std::make_shared<dynamic_loader_impl>(f, std::move(t));
+    }
+
+    HMODULE handle = nullptr;
+    tmp_dir temp;
+};
+
+#endif
+
 optional<dynamic_loader> dynamic_loader::try_load(const fs::path& p)
 {
     try
@@ -109,12 +160,19 @@ dynamic_loader::dynamic_loader(const std::vector<char>& buffer)
 
 std::shared_ptr<void> dynamic_loader::get_symbol(const std::string& name) const
 {
+#ifndef _WIN32
     // flush any previous error messages
     check_load_error(true);
     void* symbol = dlsym(impl->handle.get(), name.c_str());
     if(symbol == nullptr)
         check_load_error();
     return {impl, symbol};
+#else
+    FARPROC addr = GetProcAddress(impl->handle, name.c_str());
+    if(addr == nullptr)
+        MIGRAPHX_THROW("Symbol not found: " + name + " (" + std::to_string(GetLastError()) + ")");
+    return {impl, reinterpret_cast<void*>(addr)};
+#endif
 }
 
 } // namespace MIGRAPHX_INLINE_NS

src/include/migraphx/argument.hpp

@@ -46,7 +46,7 @@ struct MIGRAPHX_EXPORT argument : raw_data<argument>
 {
     argument() = default;
 
-    argument(const shape& s);
+    explicit argument(const shape& s);
 
     template <class F, MIGRAPHX_REQUIRES(std::is_pointer<decltype(std::declval<F>()())>{})>
     argument(shape s, F d)