DML release v0.1.5-beta

CMakeLists.txt

@@ -14,7 +14,7 @@
 #
 
 cmake_minimum_required(VERSION 3.12 FATAL_ERROR)
-project(Dml VERSION 0.1.4 LANGUAGES C CXX)
+project(Dml VERSION 0.1.5 LANGUAGES C CXX)
 
 set(PROJECT_SOVERSION 0)
 
@@ -29,35 +29,9 @@ endif()
 
 # TODO: Remove all options below
 option(LIB_ACCEL_3_2 "Use libaccel-3.2" OFF)
-option(SANITIZE_MEMORY "Enables memory sanitizing" OFF)
-option(SANITIZE_THREADS "Enables threads sanitizing" OFF)
 option(LOG_HW_INIT "Enables HW initialization log" OFF)
 option(EFFICIENT_WAIT "Enables usage of umonitor/umwait" OFF)
 
-if (SANITIZE_MEMORY)
-    message(STATUS "Memory sanitizing build is enabled")
-
-    if (WIN32)
-        message(FATAL_ERROR "Memory sanitizing is not supported in Windows")
-    else ()
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=address -fsanitize=leak -g")
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=address -fsanitize=leak -g")
-    endif ()
-endif ()
-
-if (SANITIZE_THREADS)
-    message(STATUS "Threads sanitizing build is enabled")
-
-    if (WIN32)
-        message(FATAL_ERROR "Threads sanitizing is not supported in Windows")
-    else ()
-        set(CMAKE_THREAD_PREFER_PTHREAD TRUE)
-        set(THREADS_PREFER_PTHREAD_FLAG TRUE)
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=thread -g")
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=thread -g")
-    endif ()
-endif ()
-
 include(cmake/CompileOptions.cmake)
 include(cmake/utils.cmake)
 include(CMakePackageConfigHelpers)
@@ -93,6 +67,8 @@ endif()
 get_git_revision()
 
 add_subdirectory(sources)
+
+# Testing
 add_subdirectory(examples)
 
 # Install rules

Doxyfile

@@ -53,7 +53,7 @@ PROJECT_NAME           = "Intel DML Library"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = "v0.1.4-beta"
+PROJECT_NUMBER         = "v0.1.5-beta"
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

doc/DML_REFERENCE_MANUAL.md

@@ -107,8 +107,8 @@ Hardware path is required to set up environment to utilize Intel DSA accelerator
 
 ### System Requirements
 - Software path requirements:
-    - Minimal: x86-64 CPU with Intel® Advanced Vector Extensions 2 support (Intel® microarchitecture code name Broadwell) not optimized code.
-    - Optimal: x86-64 CPU with Intel® Advanced Vector Extensions 512 (Intel® AVX-512) support (Intel® microarchitecture code name Skylake (Server) processor or higher).
+    - Minimal: x86-64 CPU with Intel® Advanced Vector Extensions 2 support (Intel® microarchitecture code name Broadwell) for not optimized code.
+    - Optimal: x86-64 CPU with Intel® Advanced Vector Extensions 512 support (Intel® microarchitecture code name Skylake (Server) processor or higher).
 
 - Hardware path requirements:
     - Minimal: CPU with Intel® DSA (Sapphire Rapids and higher).

doc/LOW_LEVEL_API_GUIDE.md

@@ -148,8 +148,8 @@ You can read more about the flags in our docs
 
 
 A complete description of the job structure fields can be found in Section (N).
-| Field Name | Description                                                                              |
-| ------------------------- | ---------------------- | ------------------------------------------------------------------------------------ |
+| Field Name | Description |
+| ------------------------- |------------------------------------------------------------------------------------ |
 | `source_first_ptr` | The starting point of the input data.|
 | `source_length` | Length of the input data. On completion, the `source_length` field contains the number of bytes remained unprocessed in case of any library/HW issue, or it contains zero if everything is successful.|
 | `source_second_ptr` | The starting point of the input data in case if the second buffer is required (length is described with `source_length`).|

doc/RELEASE_NOTES.md

@@ -1,6 +1,22 @@
 Intel® Data Mover Library (Intel® DML) Release Notes
 ===============================================================================
 
+### Intel® DML v0.1.5-beta
+
+**Date: November 2021**
+
+**Note**: Release introduces unification of underlying implementation for both C and C++ APIs
+
+**Features**:
+* Added internal device selection logic to C API (the same as for C++ API)
+  * Selector considers submitting thread's NUMA node id
+  * Selector switches devices and work queues with each submission
+* Improved range checking for C and C++ APIs
+
+**Bug fix**:
+* Lowered memory size requirements for job structure by ~100x.
+
+
 ### Intel® DML v0.1.4-beta
 
 **Date: October 2021**
@@ -11,7 +27,7 @@ Intel® Data Mover Library (Intel® DML) Release Notes
 * Fixed crashes when C++ API was used without root permissions.
 * Fixed crashes when HW initialization failed.
 * Fixed documentation section about system requirements.
-* Fixed project install-paths. They are now follow GNU guidelines.
+* Fixed project install-paths. They are now following GNU guidelines.
 * Fixed wrong path for loading `libaccel-config.so` to `/usr/lib64/libaccel-config.so`.
 
 **Warnings**:

examples/dml_job_api/job_wrapper_examples.c

@@ -430,7 +430,7 @@ dml_status_t dml_copy_batch_8u(uint8_t *source_first_ptr,
                                dml_job_t *const dml_job_ptr)
 {
     dml_status_t status = DML_STATUS_OK;
-    uint32_t operations_count    = 2u;
+    uint32_t operations_count    = 4u;
     uint32_t batch_buffer_length = 0u;
 
     status = dml_get_batch_size(dml_job_ptr, operations_count, &batch_buffer_length);
@@ -470,6 +470,30 @@ dml_status_t dml_copy_batch_8u(uint8_t *source_first_ptr,
         return status;
     }
 
+    status = dml_batch_set_mem_move_by_index(dml_job_ptr,
+                                             2u,
+                                             source_second_ptr,
+                                             destination_second_ptr,
+                                             second_length,
+                                             0u);
+    if (DML_STATUS_OK != status)
+    {
+        free(batch_buffer_ptr);
+        return status;
+    }
+
+    status = dml_batch_set_mem_move_by_index(dml_job_ptr,
+                                             3u,
+                                             source_second_ptr,
+                                             destination_second_ptr,
+                                             second_length,
+                                             0u);
+    if (DML_STATUS_OK != status)
+    {
+        free(batch_buffer_ptr);
+        return status;
+    }
+
     status = dml_execute_job(dml_job_ptr);
 
     free(batch_buffer_ptr);