[arrayfire] annotated tag upstream/3.3_alpha+dfsg1 created (now dacef43)

Ghislain Vaillant ghisvail-guest at moszumanska.debian.org
Sun Feb 7 19:01:34 UTC 2016


This is an automated email from the git hooks/post-receive script.

ghisvail-guest pushed a change to annotated tag upstream/3.3_alpha+dfsg1
in repository arrayfire.

        at  dacef43   (tag)
   tagging  cb95bdabf7079eebb4133687460aee2b93af3689 (commit)
  replaces  upstream/v3.2.2+dfsg1
 tagged by  Ghislain Antony Vaillant
        on  Fri Feb 5 10:29:27 2016 +0000

- Log -----------------------------------------------------------------
Upstream version 3.3~alpha+dfsg1

Ghislain Antony Vaillant (2):
      Merge tag 'v3.3.alpha' into dfsg-clean
      Clean source tree from non-dfsg components.

Mani Chandra (1):
      Fixes issues when compiling with icc

Pavan Yalamanchili (69):
      Merge pull request #1116 from syurkevi/forge_scatter
      Merge pull request #1133 from shehzan10/devel
      Merge pull request #1140 from 9prady9/ext_ocl_context
      Merge pull request #1163 from shehzan10/devel
      Merge pull request #1178 from shehzan10/readme-badges
      Merge pull request #1190 from pentschev/perspective_transform
      Merge remote-tracking branch 'arrayfire/master' into devel
      Merge pull request #1174 from umar456/async
      Merge pull request #1208 from shehzan10/clang-warnings
      Merge pull request #1207 from shehzan10/memory
      Build fix for CUDA backend when using boost 1.60
      Merge pull request #1216 from 9prady9/threads_uri_fix
      Merge pull request #1215 from shehzan10/devel
      Merge pull request #1211 from pavanky/compile_fixes
      Always use freeimage flags instead of hardcoded offsets
      Add the ability to sort OpenCL devices Using the following criterion
      Using proper offsets for loadImageNative and saveImageNative
      Adding environment variables to choose OpenCL device
      Cleaning up exception handling in src/api/c
      Updating docs for new AF_OPENCL_*_TYPE environment variables
      Freeimage only requires the flags for 24 / 32 bit images
      Merge pull request #1202 from shehzan10/clupdates
      Moving dispatch.hpp / dispatch.cpp to src/backend/
      Renaming a few internal functions
      Adding a unified memory manager for all backends
      Remove unnecessary line from CMakeLists
      Cleaning up error messages in loading and saving files
      Fixing CUDA platform manager to sort devices in a more saner manner.
      Adding lock to memory allocated using af_alloc_device / af::alloc
      Adding documentation for AF_MEM_DEBUG
      Additional sanitizing for mutex locks
      Removing unnecessary returns from void functions
      Fixing issue where garbageCollect was only called on current device
      BUGFIX: Initialize buffer counts to 0
      af_set_device now only warns when device > 0 on CPU
      Merge pull request #1229 from shehzan10/imageio16-test
      Merge pull request #1221 from shehzan10/fallback-opts
      Compile fixes for gcc 5.3
      Fixing compiler warnings
      Functions to get opencl device type and platforms
      Work around for a bug in AMD's clBuildProgram
      Putting transform coordinates within version guards
      Merge branch 'devel' of shehzan10/arrayfire into devel
      Merge branch 'compile_fixes' of pavanky/arrayfire into devel
      Merge pull request #1220 from shehzan10/clBLAS-updates
      Fixes to getMappedPtr in OpenCL backend
      Fixes to build with MKL when INTEL_MKL_ROOT is exported
      Merge branch 'memory' of shehzan10/arrayfire into devel
      Proper exception handling for memory manager
      Removing unneeded cudaDeviceSynchronize()
      Adding compute 37 to list of accepted CUDA computes
      BUGFIX: incorrect index for 3rd dimension in select / replace
      TEST: Adding tests for 3D and 4D select and replace
      Changes required to build tests in a single file
      Merge pull request #1251 from pavanky/bug_fixes
      Merge pull request #1250 from pavanky/misc_fixes
      Merge pull request #1252 from 9prady9/docs_fixes
      Merge branch 'fixes_to_compile_with_icc' of mchandra/arrayfire into devel
      Removing unnecessary option "BUILD_GTEST"
      Updating release notes for 3.3 pre-release
      BUGFIX: max_bytes were being set incorrectly in MemoryManager
      Cleaning up internal API for memory functions
      JIT evaluation can now be tweaked by environment variables
      BUGFIX: Fixing error in where for OpenCL backend
      Changes to internal memory manager
      Fixes to random.hpp to work in multi-threaded environment
      OpenCL JIT now launches more threads per work group for CPU devices
      Merge pull request #1262 from pavanky/memory_manager_fixes
      Merge pull request #1260 from pavanky/where_fix

Peter Andreas Entschev (14):
      Added API support for perspective transform
      Added perspective transform to CPU backend
      Added perspective transform to CUDA backend
      Added perspective transform to OpenCL backend
      Updated transform documentation
      Added perspective transform unit tests
      Updated test data
      Fixed wrong data type in OpenCL transform
      Simplified test for perspective transform in API
      Added tranform coordinates functionality
      Added transform coordinates to unified backend
      Added transform coordinates documentation
      Added transform coordinates unit tests
      Updated test data

Pradeep (1):
      additional style changes

Pradeep Garigipati (1):
      Merge pull request #1152 from syurkevi/gravity_example

Shehzan Mohammed (89):
      Increment version for devel to 3.3
      Fixes for scatter
      Merge remote-tracking branch 'upstream/master' into devel
      DOC resolve markerType enum in graphics
      Remove unused variable warning in homography cuda kernel
      FEAT added allocHost and freeHost functions
      FEAT added infoString function to return info as string
      Merge remote-tracking branch 'upstream/hotfixes-3.2.2' into devel
      Using af_alloc_host when allocating user-return string
      toString now uses af_alloc_host to allocate memory
      Add overload of toString that returns a string
      FEAT add af_get_revision to get commit instead of AF_REVISION
      Using c/version.cpp in unified
      Update README.md with updated status badges
      Cleanup util.cpp
      Using getEnvVar instead of getenv
      Cleanup/improve backend test
      Not building info for unified. backend does the same as info
      FEAT Added isImageIOAvailable function to check support
      FEAT Added isLAPACKAvailable function to check support
      Use isImageIOAvailable in testHelper
      Use isLAPACKAvailable in testHelper
      Add missing af_err to string
      af_get_last_error supports NULL as valid argument for len
      Improvements to af::exception messages
      Add version guards around allocHost and freeHost
      Deprecate af_(lock/unlock)_device_ptr. Use af_(lock/unlock)_array
      Add memFreeUnlinked to free locked device ptrs
      FEAT Add printMemInfo to print memory information
      Renamed is_free -> mngr_lock and is_unlinked -> user_lock in cpu memory mngr
      Renamed is_free -> mngr_lock and is_unlinked -> user_lock in cuda memory mngr
      Renamed is_free -> mngr_lock and is_unlinked -> user_lock in opencl memory mngr
      Reverse conditions for freeing in memory managers
      Renamed internal memFree functions
      Fix AF_DISABLE_GRAPHICS condition (Fixes e19a6be)
      Fix clang warnings (std::abs, pragma ignores)
      Add missing isLAPACKAvailable implementation in CPU backend
      Set revision to "default" when git is not available
      Documentation fixes
      Fix NONFREE Build CMake Options
      Handle compute_53 (tegra x1) for cuda lapack
      Add definition for each compute type
      Added helper functions for device type and unified mem in OpenCL
      Added getMappedPtr and unmapPtr functions in opencl memory
      Added matmul offloading to CPU
      Fix blas header types in cpu
      Fix bug in OpenCL JIT when calling functions that return same value
      Fix bug in CUDA JIT when calling functions that return same value
      dot in CUDA/OpenCL now uses mul followed by reduction
      Remove cpu dot fallback. Slower than opencl dot always
      Add OpenCL-CPU fallback for LU
      Merge pull request #1218 from pavanky/freeimage_fix
      Add OpenCL-CPU fallback for Cholesky
      Add OpenCL-CPU fallback for QR
      Add OpenCL-CPU fallback for SVD
      Add OpenCL-CPU fallback for solve
      Add OpenCL-CPU fallback for inverse
      Clean up header files in opencl/cpu/
      Update environment variables doc
      Update boost compute release tag
      Update clFFT release tag
      Move MappedPtr into the Array class
      FEAT Add getActiveBackend function
      Merge pull request #1222 from pavanky/opencl_device_sort
      Merge pull request #1225 from pavanky/memory_manager
      Re-enable disabled sort tests from issue #995
      BUGFIX Handle 16-bit data in saveImage
      Fix saveImageNative for 1-channel images
      Update test data submodule commit
      Add tests for 16-bit images for ImageIO+Native
      Handle CUDA devices locked in exclusive mode
      Add try/catch around cuda::setDevice in Pinned Memory Manager
      Merge remote-tracking branch 'upstream/devel' into fallback-opts
      Using device independent vector for cuda Pinned Memory Manager
      Cleanup for opencl-cpu offload
      Clean up cusolver finding in cmake
      Move asserts inside try/catch in indexer functions in util
      Fix OpenCL-CPU offload when OpenCL is built without lapack
      Update clBLAS release tag
      Fixes in magma potrf (opencl cholesky)
      BUGFIX Fix how streams are created in setActiveDevice (CUDA)
      Add getHostMemorySize and getDeviceMemorySize functions
      Updates to Memory Manager and Garbage Collection Heuristics
      Move ArrayFireConfig, CPack (as CPackConfig) into CMakeModules
      Move /bigobj definitions into main CMakeList (windows)
      BUGFIX Add/remove entries for platform when adding external device/context
      BUGFIX Fix CUDA device management and free at destructor
      Fix Tests: ORB, Meanshift, basic_c, solve
      Add documentation for deviceInfo

Umar Arshad (20):
      Setup async project; Setup basic queues;
      Convert CPU blas to use async queues
      Async CPU approx1 and approx2
      Async CPU Assign
      Async CPU Bilateral
      Async CPU Convolve
      Async CPU diff1 and diff2
      Avoid sending references to queued lambdas
      Async CPU Copy, Assign, and Index
      Async CPU diagonal
      Async FFT for the CPU backend
      Add eval to copyData
      Async random on CPU backend
      Async where on the CPU backe
      Async CPU reduce and ireduce
      Async CPU Transpose. Fix bug in eval
      Merge additional changes for async CPU.
      Synchronize when AF_SYNCHRONOUS_CALLS is set to 1
      Revert "Added ENQUEUE macro in cpu backend"
      Fix resize unit test.

pradeep (70):
      Merge branch 'devel' into async
      async cpu::index function
      Fixed asynchronous indexing & assignment in cpu backend
      converted cpu tile to asychronous call
      converted sort_index cpu function to asynchronous call
      converted triangle fn in cpu backend to async call
      converted lu & cholesky decomposition functions to async calls
      svd cpu backend is async now
      converted qr & solve cpu functions to async calls
      Converted wrap & unwrap cpu fns to async calls
      converted transform to async call
      converted susan fn in cpu backend to asynchronous call
      convert sort & sort_by_key cpu fns to async calls
      sobel cpu fn is async fn after this change
      Converted resize & shift cpu fns to async calls
      convert select & rotate cpu fns to async calls
      Convert morph & range cpu fns to async calls
      Converted cpu scan function to async call
      fix for async sift cpu function
      converted matchTemplate, meanShift & medfilt to async calls
      nearest neighbour cpu func is asyn call now
      conversion of listed functions to async calls
      converted join cpu func to async call
      converted cpu regions function to asynchronous call
      Removed dead code from opencl::DeviceManager class
      Enables use of ArrayFire with external context & command queue
      cpp wrappers for opencl external context related fns
      unit tests for afcl::{pushDevice, setDevice, popDevice} fns
      Style changes in opencl header
      api name change for afcl external context functionality
      Added OpenCL include dir for unit tests
      Merge branch 'devel' into async
      Converted iir, fir, fftconvolve to async calls
      Change to gfx to handle Arrays created by async calls
      Fixed harris & homography cpu fns to work with async fns
      Fix in Array::device method
      Fix for getDeviceMemInfo function in cpu
      Fixes for asynchronous cpu copy && set functions
      Merge branch 'devel' into async
      Added missing eval for input Array's in cpu backend fns
      threads library is now a submodule in cpu backend
      Moved fns enqueued onto async queue to separate folder
      Removed obselete fn of evalArray from all backends
      cpu::Array::eval queue work moved to kerenel namespace
      moved assign cpu async fn to kernel space
      moved kernel namespace in cpu backend inside cpu namespace
      moved bilateral, convolve, fftconvolve to cpu kernel namespace
      moved copy queue fns from cpu backend to kernel namespace
      Moved diagonal cpu implementation to kernel namespace
      moved diff, fast, gradient, harris, histogram to kernel namespace
      moved rgb_hsv & identity fns to kernel namespace
      moved indexing & assignment fns to kernel namespace
      template parameters style fixes in cpu kernel namespace fns
      Moved more cpu fns implementations to kernel namespace
      moved the left over fns to cpu kernel namespace
      Fixed the bug in cpu ireduce kernel function
      Fixed orb async cpu fn
      cmake fix to check for threads submodule
      moved fft cpu fns implementations to kernel namespace
      moved dot cpu implementation to kernel namespace
      moved fftconvolve reorder helper fn to kernel namespace
      fixed cpu::kernel::dot fn signature
      Added ENQUEUE macro in cpu backend
      Removed obsolete queue sync in cpu::padArray fn
      Fixed cmake condition for threads submodule check
      Merge branch 'devel' into async
      Replaced ssh based url with http url for threads submodule
      Documentation fix in matchTemplate function
      syntax+typo fix in opencl backend
      Updated test data for meanshift, bilateral & morph

syurkevi (8):
      adds scatter function
      update plot2d example to include scatter plot
      cleanup and scatter example update
      adds scatter3
      initial gravity example
      add collisions, split vectors into components for performance
      remove state globals
      remove windows pause ifdef

-----------------------------------------------------------------------

No new revisions were added by this update.

-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/arrayfire.git



More information about the debian-science-commits mailing list