Sorry, you need to enable JavaScript to visit this website.


Beignet is an open source implementation of the OpenCL specification, supporting the Intel OpenCL runtime library and compiler. OpenCL defines an implementation of parallel computation, one of the most dramatic shifts in the industry providing new performance opportunity through software engineering. With OpenCL, many programs (2D rendering, image/video processing, etc.) can be dramatically accelerated by the GPU.


Release Date: 
Oct 26, 2017
source package
Beignet 1.3.2 has been released. It's a bug fix release including the
following highlighted improvements:
1. Add some Gemini Lake devices support.
2. Add llvm4.0 and llvm5.0 support.
3. Add cl_intel_required_subgroup_size extension support.
4. Fix some other bugs.
Git tag: Release_v1.3.2
md5sum: a577ab18d67a891c8767b8ea62253543  beignet-1.3.2-source.tar.gz
sha1sum: ac7225429f4adab3bbac6d6c4a39be7629717867  beignet-1.3.2-source.tar.gz
sha256sum: c555edba149de23243965a703f39f79fa06f15bcff5a3762b5544b0925010b43  beignet-1.3.2-source.tar.gz
Guo Yejun (2):
      refresh DAG when an arg has both direct and indirect read
      keep GEN IR as SSA style
Ismo Puustinen (1):
      build: fix cmake code generation dependencies.
Jan Beich (2):
      Properly check return value from __cxa_demangle
      Limit get_program_global_data() calls to OpenCL 2.0
Jan Vesely (2):
      libocl: Add shuffle and shuffle2 builtins for half type
      libocl: Consider only bottom ilogb(2m-1)+1 bits
Luo Xionghu (2):
      fix build error log not output issue.
      llvm3.9 will assert if ouput is empty string.
Pan Xiuli (16):
      Backend: Fix flag and subflag seting for src 3 instruction
      Backend: Remove old llvm support code.
      Backend: Fix an include file error problem
      Backend: Refine GEP lowering code
      Backend: Refine LLVM version check macro
      Backend: Refine FCmp one and une
      Backend: Add LLVM40 support
      Backend: Fix llvm40 assert about literal structs
      Backend: Fix performance regression with sampler refine fro LLVM40
      Backend: Add sel ir output for MATH function
      Runtime: Fix a mssing llvm version marco for LLVM40+
      Backend: Add intel_reqd_sub_group_size support
      Runtime: Add new API enums for cl_intel_required_subgroup_size extension
      Utset: Add test case for cl_intel_required_subgroup_size extension
      Runtime: Add missing SKL deivce ID
      Fix GCC6 build bug
Patrick Beaulieu (1):
      Fix context leak with internal kernels
Rebecca N. Palmer (2):
      Docs: Fix spelling and grammar
      Docs: Fix broken link
Ruiling Song (1):
      utest: modify compiler_if_else to not rely on compiler behaviour.
Song, Ruiling (1):
      backend: use simd-1 for scalar dst in indirectMov.
Yang Rong (11):
      intel: Check that we can reserve the zero-offset
      utest: fix image qualifier of compiler_fill_gl_image test.
      GLK: add Geminilake pciids.
      GLK: add geminilake backend support.
      GLK: add geminilake runtime support.
      Runtime: implement clEnqueueAcquireGLObjects and clEnqueueReleaseGLObjects.
      libocl: enable llvm5.0 support.
      GBE: enable llvm5.0 support.
      GBE: remove static context to fix Segmentation fault.
      GBE: Fix a TBAA issue against llvm5.0.
      Bump version to 1.3.2.
Yang, Rong R (8):
      GBE: set memcpy and memset functions's linkage to LinkOnceAnyLinkage at last call.
      Runtime: fix a recurrent release context error.
      Utest: fix a build-in program leak.
      Runtime: remove ctx's useless fileds.
      GBE: clean llvm module's clone and release.
      Runtime: fix the context ref is not 0 assert when delete.
      GBE: fix a errMsg uninitialized build warning.
      Runtime: fix a build warning.
rander (4):
      backend: add sqrt-div pattern to instruction select
      utests: add utest for sqrt-div optimization
      Backend: Add optimization for negative modifier
      utests: added for optimization negativeAdd (1):
      backend: refine the local copy propagation.