Merge pull request #1102 from arrayfire/devel

Release 3.2.0

ArrayFireConfig.cmake.in

@@ -9,12 +9,14 @@
 #
 # ----------------------------------------------------------------------------
 #
-# ArrayFire_CPU_FOUND        - True of the ArrayFire CPU library has been found.
-# ArrayFire_CPU_LIBRARIES    - Location of ArrayFire's CPU library, if found
-# ArrayFire_CUDA_FOUND       - True of the ArrayFire CUDA library has been found.
-# ArrayFire_CUDA_LIBRARIES   - Location of ArrayFire's CUDA library, if found
-# ArrayFire_OpenCL_FOUND     - True of the ArrayFire OpenCL library has been found.
-# ArrayFire_OpenCL_LIBRARIES - Location of ArrayFire's OpenCL library, if found
+# ArrayFire_CPU_FOUND         - True of the ArrayFire CPU library has been found.
+# ArrayFire_CPU_LIBRARIES     - Location of ArrayFire's CPU library, if found
+# ArrayFire_CUDA_FOUND        - True of the ArrayFire CUDA library has been found.
+# ArrayFire_CUDA_LIBRARIES    - Location of ArrayFire's CUDA library, if found
+# ArrayFire_OpenCL_FOUND      - True of the ArrayFire OpenCL library has been found.
+# ArrayFire_OpenCL_LIBRARIES  - Location of ArrayFire's OpenCL library, if found
+# ArrayFire_Unified_FOUND     - True of the ArrayFire Unified library has been found.
+# ArrayFire_Unified_LIBRARIES - Location of ArrayFire's Unified library, if found
 #
 #=============================================================================
 # Copyright (c) 2015, ArrayFire
@@ -48,17 +50,23 @@
 
 get_filename_component(ArrayFire_INCLUDE_DIRS "@INCLUDE_DIR@" ABSOLUTE)
 
-# keep in the backends in the slowest to fastest order
-foreach(backend CPU OpenCL CUDA)
-  string(TOLOWER "${backend}" lowerbackend)
+macro(find_backend backend libname)
   set(targetFile ${CMAKE_CURRENT_LIST_DIR}/@BACKEND_DIR@/ArrayFire${backend}.cmake)
   if(EXISTS ${targetFile})
     include(${targetFile})
     set(ArrayFire_${backend}_FOUND ON)
-    set(ArrayFire_${backend}_LIBRARIES af${lowerbackend})
+    set(ArrayFire_${backend}_LIBRARIES af${libname})
     # set the default backend
-    set(ArrayFire_LIBRARIES af${lowerbackend})
+    set(ArrayFire_LIBRARIES af${libname})
   else()
     set(ArrayFire_${backend}_FOUND OFF)
   endif()
+endmacro()
+
+# keep in the backends in the slowest to fastest order
+foreach(backend CPU OpenCL CUDA)
+  string(TOLOWER "${backend}" lowerbackend)
+  find_backend("${backend}" "${lowerbackend}")
 endforeach()
+
+find_backend("Unified" "")

CMakeLists.txt

@@ -35,6 +35,8 @@ OPTION(BUILD_SIFT "Build ArrayFire nonfree algorithms" OFF)
 
 MARK_AS_ADVANCED(BUILD_SIFT)
 
+OPTION(BUILD_UNIFIED "Build Backend-Independent ArrayFire API" ON)
+
 # Set a default build type if none was specified
 if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
     set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
@@ -163,6 +165,11 @@ IF(${BUILD_OPENCL})
     ADD_SUBDIRECTORY(src/backend/opencl)
 ENDIF()
 
+IF(${BUILD_UNIFIED})
+    ADD_DEFINITIONS(-DAF_UNIFIED)
+    ADD_SUBDIRECTORY(src/api/unified)
+ENDIF()
+
 IF(${BUILD_DOCS})
     ADD_SUBDIRECTORY(docs)
 ENDIF()

CMakeModules/Version.cmake

@@ -2,8 +2,8 @@
 # Make a version file that includes the ArrayFire version and git revision
 #
 SET(AF_VERSION_MAJOR "3")
-SET(AF_VERSION_MINOR "1")
-SET(AF_VERSION_PATCH "3")
+SET(AF_VERSION_MINOR "2")
+SET(AF_VERSION_PATCH "0")
 
 SET(AF_VERSION "${AF_VERSION_MAJOR}.${AF_VERSION_MINOR}.${AF_VERSION_PATCH}")
 SET(AF_API_VERSION_CURRENT ${AF_VERSION_MAJOR}${AF_VERSION_MINOR})

CMakeModules/build_clBLAS.cmake

@@ -14,7 +14,7 @@ ENDIF()
 ExternalProject_Add(
     clBLAS-external
     GIT_REPOSITORY https://github.com/arrayfire/clBLAS.git
-    GIT_TAG 47662a6ac1186c756508109d7fef8827efab4504
+    GIT_TAG 102c832825e8e4d60ad73ca97e95668463294068
     PREFIX "${prefix}"
     INSTALL_DIR "${prefix}"
     UPDATE_COMMAND ""

CMakeModules/build_forge.cmake

@@ -22,7 +22,7 @@ ENDIF()
 ExternalProject_Add(
     forge-ext
     GIT_REPOSITORY https://github.com/arrayfire/forge.git
-    GIT_TAG af3.1.2
+    GIT_TAG af3.2.0
     PREFIX "${prefix}"
     INSTALL_DIR "${prefix}"
     UPDATE_COMMAND ""

CMakeModules/osx_install/OSXInstaller.cmake

@@ -76,7 +76,7 @@ PKG_BUILD(  PKG_NAME        ArrayFireCPU
             SCRIPT_DIR      ${OSX_INSTALL_DIR}/cpu_scripts
             IDENTIFIER      com.arrayfire.pkg.arrayfire.cpu.lib
             PATH_TO_FILES   package/lib
-            FILTERS         opencl cuda)
+            FILTERS         opencl cuda unified)
 
 PKG_BUILD(  PKG_NAME        ArrayFireCUDA
             DEPENDS         afcuda
@@ -85,15 +85,23 @@ PKG_BUILD(  PKG_NAME        ArrayFireCUDA
             SCRIPT_DIR      ${OSX_INSTALL_DIR}/cuda_scripts
             IDENTIFIER      com.arrayfire.pkg.arrayfire.cuda.lib
             PATH_TO_FILES   package/lib
-            FILTERS         cpu opencl)
+            FILTERS         cpu opencl unified)
 
 PKG_BUILD(  PKG_NAME        ArrayFireOPENCL
             DEPENDS         afopencl
             TARGETS         opencl_package
             INSTALL_LOCATION /usr/local/lib
             IDENTIFIER      com.arrayfire.pkg.arrayfire.opencl.lib
             PATH_TO_FILES   package/lib
-            FILTERS         cpu cuda)
+            FILTERS         cpu cuda unified)
+
+PKG_BUILD(  PKG_NAME        ArrayFireUNIFIED
+            DEPENDS         af
+            TARGETS         unified_package
+            INSTALL_LOCATION /usr/local/lib
+            IDENTIFIER      com.arrayfire.pkg.arrayfire.unified.lib
+            PATH_TO_FILES   package/lib
+            FILTERS         cpu cuda opencl)
 
 PKG_BUILD(  PKG_NAME        ArrayFireHeaders
             TARGETS         header_package
@@ -107,5 +115,5 @@ PKG_BUILD(  PKG_NAME        ArrayFireExtra
             IDENTIFIER      com.arrayfire.pkg.arrayfire.extra
             PATH_TO_FILES   package/share)
 
-PRODUCT_BUILD(DEPENDS ${cpu_package} ${cuda_package} ${opencl_package} ${header_package} ${extra_package})
+PRODUCT_BUILD(DEPENDS ${cpu_package} ${cuda_package} ${opencl_package} ${unified_package} ${header_package} ${extra_package})
 

CMakeModules/osx_install/distribution.dist

@@ -4,32 +4,55 @@
     <welcome    file="${WELCOME_FILE_OUT}" />
     <readme     file="${README_FILE_OUT}" mime-type="test/html" />
     <license    file="${CMAKE_MODULE_PATH}/../LICENSE" mime-type="test/plain" />
+    <script>
+      function CheckBackendSelected() {
+      return choices.opencl_lib.selected ||
+            choices.cuda_lib.selected ||
+            choices.cpu_lib.selected;
+      }
+    </script>
 
     <pkg-ref id="com.arrayfire.arrayfire.cpu.lib"       version="${AF_VERSION}" onConclusion="none">ArrayFireCPU.pkg</pkg-ref>
     <pkg-ref id="com.arrayfire.arrayfire.cuda.lib"      version="${AF_VERSION}" onConclusion="none">ArrayFireCUDA.pkg</pkg-ref>
     <pkg-ref id="com.arrayfire.arrayfire.opencl.lib"    version="${AF_VERSION}" onConclusion="none">ArrayFireOPENCL.pkg</pkg-ref>
+    <pkg-ref id="com.arrayfire.arrayfire.unified.lib"   version="${AF_VERSION}" onConclusion="none">ArrayFireUNIFIED.pkg</pkg-ref>
     <pkg-ref id="com.arrayfire.arrayfire.inc"           version="${AF_VERSION}" onConclusion="none">ArrayFireHeaders.pkg</pkg-ref>
     <pkg-ref id="com.arrayfire.arrayfire.extra"         version="${AF_VERSION}" onConclusion="none">ArrayFireExtra.pkg</pkg-ref>
     <options customize="always" require-scripts="false"/>
     <choices-outline>
         <line choice="libs">
-            <line choice="com.arrayfire.arrayfire.cpu.lib"/>
-            <line choice="com.arrayfire.arrayfire.cuda.lib"/>
-            <line choice="com.arrayfire.arrayfire.opencl.lib"/>
+            <line choice="cpu_lib"/>
+            <line choice="cuda_lib"/>
+            <line choice="opencl_lib"/>
+            <line choice="com.arrayfire.arrayfire.unified.lib"/>
         </line>
         <line choice="com.arrayfire.arrayfire.inc"/>
         <line choice="com.arrayfire.arrayfire.extra"/>
     </choices-outline>
     <choice id="libs" title="ArrayFire Libraries" visible="true" />
-    <choice title="CPU Libraries" description="CPU Libraries" id="com.arrayfire.arrayfire.cpu.lib" visible="true" enabled="true">
+    <choice title="CPU Libraries"
+            description="ArrayFire targeting CPUs."
+            id="cpu_lib" visible="true" enabled="true">
         <pkg-ref id="com.arrayfire.arrayfire.cpu.lib"/>
     </choice>
-    <choice title="CUDA Libraries" description="CUDA Libraries" id="com.arrayfire.arrayfire.cuda.lib" visible="true" enabled="true">
+    <choice title="CUDA Libraries"
+            description="ArrayFire which targets the CUDA platform. This platform allows you to to take advantage of the CUDA enabled GPUs to run ArrayFire code."
+            id="cuda_lib" visible="true" enabled="true">
         <pkg-ref id="com.arrayfire.arrayfire.cuda.lib"/>
     </choice>
-    <choice title="OpenCL Libraries" description="OpenCL Libraries" id="com.arrayfire.arrayfire.opencl.lib" visible="true" enabled="true">
+    <choice title="OpenCL Libraries"
+            description="ArrayFire which targets the OpenCL platform. This platform allows you to use the ArrayFire library which targets OpenCL devices. NOTE: Currently ArrayFire does not support OpenCL for the Intel CPU on Apple."
+            id="opencl_lib" visible="true" enabled="true">
         <pkg-ref id="com.arrayfire.arrayfire.opencl.lib"/>
     </choice>
+    <choice title="Unified Library"
+            description="This library will allow you to choose the platform(cpu, cuda, opencl) at runtime. NOTE: This option requires the other platforms to work properly"
+            id="com.arrayfire.arrayfire.unified.lib"
+            selected="CheckBackendSelected()"
+            visible="true"
+            enabled="CheckBackendSelected()">
+        <pkg-ref id="com.arrayfire.arrayfire.unified.lib"/>
+    </choice>
     <choice title="ArrayFire Headers" description="ArrayFire Headers" id="com.arrayfire.arrayfire.inc" visible="true" enabled="true">
         <pkg-ref id="com.arrayfire.arrayfire.inc"/>
     </choice>

README.md

@@ -20,12 +20,10 @@ ArrayFire binary installers can be downloaded at the [ArrayFire Downloads](http:
 * Email: <mailto:technical@arrayfire.com>
 
 ### Build Status
-|                 | Build           | Tests           |
-|-----------------|-----------------|-----------------|
-| Linux x86       | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-linux/devel)](http://ci.arrayfire.org/job/arrayfire-linux/branch/devel/)      | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-linux-test/devel)](http://ci.arrayfire.org/job/arrayfire-linux-test/branch/devel/)              |
-| Linux Tegra     | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegra/devel)](http://ci.arrayfire.org/job/arrayfire-tegra/branch/devel/)      | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegra-test/devel)](http://ci.arrayfire.org/job/arrayfire-tegra-test/branch/devel/)              |
-| Windows         | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-windows/devel)](http://ci.arrayfire.org/job/arrayfire-windows/branch/devel/)  | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-windows-test/devel)](http://ci.arrayfire.org/job/arrayfire-windows-test/branch/devel/)          |
-| OSX             | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-osx/devel)](http://ci.arrayfire.org/job/arrayfire-osx/branch/devel/)          | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-osx-test/devel)](http://ci.arrayfire.org/job/arrayfire-osx-test/branch/devel/)                  |
+|         | Linux x86 | Linux armv7l | Linux aarch64 | Windows | OSX |
+|:-------:|:---------:|:------------:|:-------------:|:-------:|:---:|
+| Build   | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-linux/devel)](http://ci.arrayfire.org/job/arrayfire-linux/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegrak1/devel)](http://ci.arrayfire.org/job/arrayfire-tegrak1/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegrax1/devel)](http://ci.arrayfire.org/job/arrayfire-tegrax1/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-windows/devel)](http://ci.arrayfire.org/job/arrayfire-windows/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-osx/devel)](http://ci.arrayfire.org/job/arrayfire-osx/branch/devel/) |
+| Test    | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-linux-test/devel)](http://ci.arrayfire.org/job/arrayfire-linux-test/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegrak1-test/devel)](http://ci.arrayfire.org/job/arrayfire-tegrak1-test/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-tegrax1-test/devel)](http://ci.arrayfire.org/job/arrayfire-tegrax1-test/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-windows-test/devel)](http://ci.arrayfire.org/job/arrayfire-windows-test/branch/devel/) | [![Build Status](http://ci.arrayfire.org/buildStatus/icon?job=arrayfire-osx-test/devel)](http://ci.arrayfire.org/job/arrayfire-osx-test/branch/devel/) |
 
 Test coverage: [![Coverage Status](https://coveralls.io/repos/arrayfire/arrayfire/badge.svg?branch=HEAD)](https://coveralls.io/r/arrayfire/arrayfire?branch=HEAD)
 

docs/details/algorithm.dox

@@ -15,6 +15,15 @@ This function performs the operation across all batches present in the input sim
 
 Find the sum of values in the input
 
+This table defines the return value types for the corresponding input types
+
+Input Type          | Output Type
+--------------------|---------------------
+f32, f64, c32, c64  | same as input
+s32, u32, s64, u64  | same as input
+s16                 | s32
+u16, u8, b8         | u32
+
 \copydoc batch_detail_algo
 
 
@@ -25,6 +34,15 @@ Find the sum of values in the input
 
 Find the product of values in the input
 
+This table defines the return value types for the corresponding input types
+
+Input Type          | Output Type
+--------------------|---------------------
+f32, f64, c32, c64  | same as input
+s32, u32, s64, u64  | same as input
+s16                 | s32
+u16, u8, b8         | u32
+
 \copydoc batch_detail_algo
 
 
@@ -55,6 +73,8 @@ Find the maximum values and their locations
 
 Find if of all of the values in input are true
 
+Return type is b8 for all input types
+
 \copydoc batch_detail_algo
 
 
@@ -65,6 +85,8 @@ Find if of all of the values in input are true
 
 Find if of any of the values in input are true
 
+Return type is b8 for all input types
+
 \copydoc batch_detail_algo
 
 
@@ -75,6 +97,8 @@ Find if of any of the values in input are true
 
 Count the number of non-zero elements in the input
 
+Return type is u32 for all input types
+
 \copydoc batch_detail_algo
 
 
@@ -85,6 +109,15 @@ Count the number of non-zero elements in the input
 
 Perform exclusive sum along specified dimension
 
+This table defines the return value types for the corresponding input types
+
+Input Type          | Output Type
+--------------------|---------------------
+f32, f64, c32, c64  | same as input
+s32, u32, s64, u64  | same as input
+s16                 | s32
+u16, u8, b8         | u32
+
 \copydoc batch_detail_algo
 
 
@@ -95,6 +128,8 @@ Perform exclusive sum along specified dimension
 
 Locate the indices of non-zero elements
 
+Return type is u32 for all input types
+
 The locations are provided by flattening the input into a linear array.
 
 
@@ -135,7 +170,8 @@ Sort an multi dimensional array
 
 Sort input arrays get the sorted indices
 
-Sort a multi dimensional array and return sorted indices
+Sort a multi dimensional array and return sorted indices. Index array is of
+type u32.
 
 
 

docs/details/backend.dox

@@ -0,0 +1,66 @@
+/**
+\addtogroup arrayfire_func
+@{
+
+\defgroup unified_func_setbackend setBackend
+
+\brief Set the current backend when using Unified backend
+
+This is a noop when using one of CPU, CUDA, or OpenCL backend.
+
+However, when using on of those 3 but trying to set it to a different backend
+will return in an exception.
+
+\ingroup unified_func
+\ingroup arrayfire_func
+
+=======================================================================
+
+\defgroup unified_func_getbackendcount getBackendCount
+
+\brief Get the number of backends whose libraries were successfully loaded.
+
+This will be between 0-3. 0 Being no backends were loaded and 3 being all
+backends loaded successfully.
+
+\ingroup unified_func
+\ingroup arrayfire_func
+
+=======================================================================
+
+\defgroup unified_func_getavailbackends getAvailableBackends
+
+\brief Returns an integer indicating the backends loaded successfully.
+
+The number returned denotes the backends available according to the table:
+
+Return Value | Backends Available
+-------------|-----------------------
+0            | None
+1            | CPU
+2            | CUDA
+3            | CPU and CUDA
+4            | OpenCL
+5            | CPU and OpenCL
+6            | CUDA and OpenCL
+7            | CPU, CUDA and OpenCL
+
+\ingroup unified_func
+\ingroup arrayfire_func
+
+=======================================================================
+
+\defgroup unified_func_getbackendid getBackendId
+
+\brief Get's the backend enum for an array
+
+This will return one of the values from the \ref af_backend enum.
+The return value specifies which backend the array was created on.
+
+\ingroup unified_func
+\ingroup arrayfire_func
+
+=======================================================================
+
+@}
+*/

docs/details/image.dox

@@ -329,6 +329,9 @@ distance as well as the color distance.
 The bilateral filter requires the size of the filter (in pixels) and the upper
 bound on color values, N, where pixel values range from 0–N inclusively.
 
+The return type of the array is f64 for f64 input, f32 for all other input
+types.
+
 =======================================================================
 
 \defgroup image_func_erode erode