Complete makefile changes; Add comments

INSTALL.md

@@ -78,7 +78,7 @@ is available as an alternative to the manual setup.
 4. Set up MKL:
 
     _Note: if you used the general oneAPI setvars script from a Base Toolkit installation, this step will not be necessary as oneMKL will already have been set up._
-    
+
     Download and install [Intel(R) oneMKL](https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl.html).
     Set the environment variables for for Intel(R) oneMKL. For example:
 
@@ -149,7 +149,15 @@ It is possible to build oneDAL libraries with selected set of algorithms and/or
 
             make -f makefile daal PLAT=win32e CORE.ALGORITHMS.CUSTOM=low_order_moments REQCPU=avx2 -j16
 
+On **Linux\*** it is possible to build debug version of oneDAL or the version that allows to do kernel profiling using <ittnotify.h>.
+
+- To build debug version of oneDAL, run:
+
+            make -f makefile daal oneapi_c PLAT=lnx32e REQDBG=yes
+
+- To build oneDAL with kernel profiling information, run:
 
+            make -f makefile daal oneapi_c PLAT=lnx32e REQPROFILE=yes
 
 ---
 **NOTE:** Built libraries are located in the `__release_{os_name}[_{compiler_name}]/daal` directory.

cpp/daal/src/externals/service_profiler.h

@@ -24,9 +24,8 @@
 #ifndef __SERVICE_PROFILER_H__
 #define __SERVICE_PROFILER_H__
 
-// #define ONEDAL_KERNEL_PROFILER
-
 #ifdef ONEDAL_KERNEL_PROFILER
+/* Here if oneDAL kernel profiling is enabled in the build */
 #include <ittnotify.h>
 #endif
 
@@ -36,47 +35,83 @@
 #define DAAL_ITTNOTIFY_UNIQUE_ID __LINE__
 
 #define DAAL_ITTNOTIFY_SCOPED_TASK(name) \
-    daal::internal::ProfilerTask DAAL_ITTNOTIFY_CONCAT(__profiler_taks__, DAAL_ITTNOTIFY_UNIQUE_ID) = daal::internal::Profiler::startTask(#name);
+    daal::internal::ProfilerTask DAAL_ITTNOTIFY_CONCAT(__profiler_task__, DAAL_ITTNOTIFY_UNIQUE_ID) = daal::internal::Profiler::startTask(#name);
 
 namespace daal
 {
 namespace internal
 {
+/**
+ * Defines a logical unit of work to be tracked by performance profilier.
+ */
 class ProfilerTask
 {
 public:
+    /**
+     * Constructs a task with a given name.
+     * \param[in] taskName   Name of the task.
+     */
     ProfilerTask(const char * taskName);
     ~ProfilerTask();
 
 private:
     const char * _taskName;
 #ifdef ONEDAL_KERNEL_PROFILER
-    __itt_string_handle* _handle;
-    __itt_domain* _domain;
+    /* Here if oneDAL kernel profiling is enabled */
+    __itt_string_handle * _handle;  /* The task string handle */
+    __itt_domain * _domain;         /* Pointer to the domain of the task */
 #endif
 };
 
-// This class is a stub in the library. Its redefinition will be in Bechmarks
+/**
+ * Global performance profiler.
+ *
+ * By default this class is a stub in the library and its redefinition will be in C++ Bechmarks.
+ * If oneDAL kernel profiling is enabled, the profiler uses Task API from <ittnotify.h>
+ */
 class Profiler
 {
 public:
+    /**
+     * Start the task to be profiled.
+     * \param[in] taskName   Name of the task.
+     */
     static ProfilerTask startTask(const char * taskName);
+
+    /**
+     * Start the task to profile.
+     * \param[in] taskName   Name of the task.
+     */
     static void endTask(const char * taskName);
+
 #ifdef ONEDAL_KERNEL_PROFILER
+    /* Here if oneDAL kernel profiling is enabled */
+
+    /**
+     * Get pointer to a global profiler state.
+     * \return Pointer to a global profiler state.
+     */
     static Profiler* getInstance() {
         static Profiler  instance;
         return &instance;
     }
 
-    static __itt_domain* getDomain() {
+    /**
+     * Get pointer to the ITT domain associated with the profiler.
+     * \return Pointer to the ITT domain.
+     */
+    static __itt_domain * getDomain() {
         return (getInstance())->_domain;
     }
 private:
+    /**
+     * Construct the profiler.
+     */
     Profiler() {
         _domain = __itt_domain_create("oneDAL");
     }
     ~Profiler() {}
-    __itt_domain* _domain;
+    __itt_domain * _domain;     /* Pointer to the ITT domain */
 #endif
 };
 

makefile

@@ -306,15 +306,12 @@ include dev/make/deps.$(BACKEND_CONFIG).mk
 #=============================== VTune SDK folders ======================================
 
 ifeq ($(REQPROFILE), yes)
+    -DPROFILER := -DONEDAL_KERNEL_PROFILER
     VTUNESDK.include := $(VTUNE_PROFILER_DIR)/sdk/include
-
     VTUNESDK.libia := $(if $(OS_is_lnx), $(VTUNE_PROFILER_DIR)/sdk/lib64,)
-
     VTUNESDK.LIBS_A := $(if $(OS_is_lnx), $(VTUNESDK.libia)/libittnotify.a,)
 endif
 
-$(info VTUNESDK: $(VTUNESDK.include))
-
 #===============================================================================
 # Release library names
 #===============================================================================
@@ -506,7 +503,7 @@ $(WORKDIR.lib)/$(core_y):                   $(daaldep.math_backend.ext) $(VTUNES
                                             $(CORE.tmpdir_y)/$(core_y:%.$y=%_link.txt) ; $(LINK.DYNAMIC) ; $(LINK.DYNAMIC.POST)
 
 $(CORE.objs_a): $(CORE.tmpdir_a)/inc_a_folders.txt
-$(CORE.objs_a): COPT += $(-fPIC) $(-cxx17) $(-Zl) $(-DEBC) $(-DMKL_ILP64)
+$(CORE.objs_a): COPT += $(-fPIC) $(-cxx17) $(-Zl) $(-DEBC) $(-DMKL_ILP64) $(-DPROFILER)
 $(CORE.objs_a): COPT += -D__TBB_NO_IMPLICIT_LINKAGE -DDAAL_NOTHROW_EXCEPTIONS \
                         -DDAAL_HIDE_DEPRECATED -DTBB_USE_ASSERT=0 -D_ENABLE_ATOMIC_ALIGNMENT_FIX \
                         $(if $(CHECK_DLL_SIG),-DDAAL_CHECK_DLL_SIG)
@@ -515,7 +512,7 @@ $(CORE.objs_a): COPT += @$(CORE.tmpdir_a)/inc_a_folders.txt
 $(eval $(call append_uarch_copt,$(CORE.objs_a)))
 
 $(CORE.objs_y): $(CORE.tmpdir_y)/inc_y_folders.txt
-$(CORE.objs_y): COPT += $(-fPIC) $(-cxx17) $(-Zl) $(-DEBC) $(-DMKL_ILP64)
+$(CORE.objs_y): COPT += $(-fPIC) $(-cxx17) $(-Zl) $(-DEBC) $(-DMKL_ILP64) $(-DPROFILER)
 $(CORE.objs_y): COPT += -D__DAAL_IMPLEMENTATION \
                         -D__TBB_NO_IMPLICIT_LINKAGE -DDAAL_NOTHROW_EXCEPTIONS \
                         -DDAAL_HIDE_DEPRECATED -DTBB_USE_ASSERT=0 -D_ENABLE_ATOMIC_ALIGNMENT_FIX \
@@ -1117,7 +1114,8 @@ Flags:
       possible values: $(CORE.ALGORITHMS.CUSTOM.AVAILABLE)
   REQCPU - list of CPU optimizations to be included into library
       possible values: $(CPUs)
-  REQDBG - Flag that enables build in debug mode
+  REQDBG - flag that enables build in debug mode
+  REQPROFILE - flag that enables kernel profiling using <ittnotify.h>
 endef
 
 daal_dbg: