- Update to version 21.1.0.
* Move (add|sub)_sat to CLC; optimize
* Move several integer functions to CLC library
* Remove use of symlinks
* Move mad_sat to CLC; optimize for vector types
* Move rotate to CLC library; optimize
* Allow default path when looking for llvm-spirv
* Have all targets build all CLC functions
* Move sign to the CLC builtins library
* Move conversion builtins to the CLC library
* [cmake] Fix per-target *_convert.cl dependencies
* Fix dependencies on generated convert builtins
* Disable external-calls testing for clspv targets
* Suppress data-layout warnings during linking
* Use CLC conversion builtins in CLC functions
* Replace gfx940 and gfx941 with gfx942
* Improve nextafter behaviour around zero
* Clean up directory search procedure
* Explicitly qualify private address spaces
* Move modf to the CLC library
* Move frexp to CLC library; optimize half vecs
* Define macros for users of gentype.inc
* Fix int<->float conversion builtins
* Move fma to the CLC library
* Remove clspv-specific clc conversions
* Move nan to the CLC library
* Stop using asm declarations for r600 on amdgcn for
get_global_size
* Move log/log2/log10 to CLC library
* Move round to CLC library
* Move __clc_ldexp to CLC library
* Make CLC library warning-free
* Move sqrt to CLC library
* Move rsqrt to the CLC library
* Move hypot to CLC library; optimize
* Reduce bithacking in CLC frexp
* Stop installing CLC headers
* Add missing gfx950 target
* Update license headers
* add --only-needed to llvm-link when INTERNALIZE flag is set
* link_bc target should depends on target
builtins.link.clc-arch_suffix
* Add license headers to files missing them
* Relicense gen_convert.py
* erfc: fix fp32 implementation in FTZ mode
* Move fp32 sincos helpers to CLC library
* Add missing license headers to source IR files
* Move asin/acos/atan to the CLC library
* Move asinpi/acospi/atanpi to the CLC library
* Move log1p/asinh/acosh/atanh to the CLC library
* Move atan2/atan2pi to the CLC library
* Move fmod, remainder & remquo to the CLC library
* Pass -fapprox-func when compiling 'native' builtins
* Move pow, powr & pown to the CLC library
* Implement native_exp2 via AMD builtin
* Fix unresolved reference to missing table
* Move rootn to the CLC library; optimize
* Move several 'native' builtins to CLC library
* Move sinpi/cospi/tanpi to the CLC library
* Move exp10 to the CLC library
* Move exp, exp2 and expm1 to the CLC library
* Move cbrt to the CLC library; vectorize
* Move lgamma, lgamma_r & tgamma to CLC library
* Move sinh, cosh & tanh to the CLC library
* Move native_(exp10|powr|tan) to CLC library
* clspv: add a dummy implememtation for mul_hi
* frexp: fix implementation regarding denormals
* Move shuffle/shuffle2 to the CLC library
* clspv: fma: remove fp16 implementation
* Fix commands in compile_to_bc are executed sequentially
* Add ctz built-in implementation to clc and generic
* Build for OpenCL 3.0
* Fix unguarded use of image types
* Re-enable compiler warning
* only check filename part of the source for avoiding duplication
* Remove unnecessary clcmacros.h
* Remove (vload|vstore)_half helpers
* Move fmin & fmax to CLC library
* Move fract to the CLC library
* Move fdim to CLC library; simplify
* Avoid casting NANs & literals to 'gentype'
* Add v3 variants of async_work_group_copy/
async_work_group_strided_copy/prefetch
* Clean up unnecessary #undef __CLC_BODYs
* Move minmag & maxmag to the CLC library
* Move log2/log10 tables to CLC tables impl
* Move sin, cos & sincos to CLC library
* Move tan to the CLC library
* Move 'half' builtins to CLC library
* Move (fast) length & distance to CLC library
* Move logb/ilogb to CLC library; optimize
* Move cross to CLC library; add missing half overloads
* Improving vector code generated from scalar code
* Append file_specific_compile_options after ARG_COMPILE_FLAGS
* Mov erf & erfc to CLC library
* Reorganize OpenCL builtins
* Fix header inclusion issues
* Re-use shuffle_decl.inc in OpenCL shuffle2 declaration
* Move all remquo address spaces to CLC library
* Support the generic address space
* Move step to the CLC library; add missing half variants
* Move vload & vstore to CLC library
* Move prefetch to clc library
* clspv: do not set generic_addrspace_val
* Add (fast) normalize to CLC; add half overloads
* Support LLVM_ENABLE_RUNTIMES when building
* Avoid out-of-range float-to-int.
* Add __clc_nan implementation with signed nancode argument
* Add missing clc_lgamma_r with generic address space pointer arg
* Reduce include usage in OpenCL builtins
* Fix typo in OpenCL header math/sincos.h
* Tighten OpenCL builtin include strategy
* Remove catch-all opencl/clc.h
* Declare workitem built-ins in clc, move ptx-nvidiacl workitem
built-ins into clc
- Headers are no longer included.
OBS-URL: https://build.opensuse.org/request/show/1304686
OBS-URL: https://build.opensuse.org/package/show/X11:XOrg/libclc?expand=0&rev=82
- Update to version 0.2.0+llvm11.0.0.
The repository that we extracted the tarball from isn't updated
any longer. So we take the tarballs from LLVM releases instead.
- The build now uses CMake instead of a custom Python script.
- Remove dependencies on gcc, libstdc++-devel, ncurses and zlib.
- The provided package consists of LLVM bitcode files, which are
not necessarily backwards-compatible across major versions.
(https://llvm.org/docs/DeveloperPolicy.html#ir-backwards-compatibility)
So we let the package provide a symbol libclc(llvmXX).
- The library files have moved from %{_libdir} to %{_datadir}.
- Make noarch: the bitcode files don't depend on the host CPU.
OBS-URL: https://build.opensuse.org/request/show/845078
OBS-URL: https://build.opensuse.org/package/show/X11:XOrg/libclc?expand=0&rev=32
- Update to version 0.2.0+git.20190313:
* sincos: Simplify declaration headers.
* fdim: Use binary_decl_tt.inc instead of custom inc file.
* nextafter: Use binary_decl_tt.inc instead of custom inc file.
* copysign: Use binary_decl_tt.inc instead of custom inc file.
* atan2pi: Use binary_decl_tt.inc instead of custom inc file.
* atan2: Use binary_decl_tt.inc instead of custom inc file.
* hypot: Use binary_decl_tt.inc instead of custom inc file
* Move unary_instrinsic.inc to private headers.
* Move binary_intrinsic.h to private headers.
* Move ternary_intrinsic.h to private headers.
* trunc: Remove llvm intrinsic from the header.
* round: Remove llvm intrinsic from the header
* rint: Remove llvm intrinsic from the header.
* floor: Remove llvm isntrinsic from the header.
* fabs: Remove llvm intrinsic from the header.
* ceil: Remove llvm intrinsic from the header.
* sqrt: Split function generation to a shared inc file.
* mad: Convert to standard ternary header
* travis: Deduplicate LLVM_CONFIG variable
* travis: Use gcc-6 for llvm-7 build
* Adjust documentation for git migration.
* Fix typos throughout the license files that somehow I and my reviewers all missed!
* Update the file headers across all of the LLVM projects in the monorepo to reflect the new license.
* Install new LLVM license structure and new developer policy.
* Update year in license files
* cmake: Install libraries to DATADIR from GNUInstallDirs
OBS-URL: https://build.opensuse.org/request/show/684790
OBS-URL: https://build.opensuse.org/package/show/X11:XOrg/libclc?expand=0&rev=27
- Update to version 0.2.0+git.20180915:
* configure: Rework support for gfx9+ devices that were added post LLVM 3.9
* .travis: Add llvm-7 build
* .travis: Use source whitelist alias for llvm-6 repository
* amdgcn: Use __constant AS for amdgcn builtins.
* atom: Use volatile pointers for cl_khr_{global,local}_int32_{base,extended}_atomics
* atom: Consolidate cl_khr_{local,global}_int32_{base,extended}_atomics implementation
* atomic: Provide function implementation of atomic_{dec,inc}
* atom: Consolidate cl_khr_int64_{base,extended}_atomics declarations
* atom: Consolidate cl_khr_{local,global}_int32_{base,extended}_atomics declarations
* atomic: Cleanup atomic_cmpxchg header
* atomic: Move define cleanup to shared include
* Update copyright year to 2018.
* r600/fmin: Flush denormals before calling builtin.
* r600/fmax: Flush denormals before calling builtin.
* math/fma: Add fp32 software implementation
* Add initial support for half precision builtins
* rootn: Use denormal path only
* remquo: Flush denormals if not supported
* remquo: Port from amd builtins
* math: Add helper function to flush denormals if not supported.
* clc_sqrt: Reuse unary_decl.inc
* relational/select: Condition types for half are short/ushort, not char/uchar
* log10: Use sw implementation from amd builtins
* powr: Use denormal path only
* pown: Use denormal path only
* pow: Use denormal path only
* amdgcn/fmin: Fix typos that reduced precision
* exp10: Port from amd builtins
* hypot: Port from amd builtins
OBS-URL: https://build.opensuse.org/request/show/641350
OBS-URL: https://build.opensuse.org/package/show/X11:XOrg/libclc?expand=0&rev=23
Will fail to build in devel, but needed for accept llvm4 to Factory.
- Update to version 0.2.0+git.20170225:
* Fix build since llvm r286566 and require at least llvm 4.0
* Fix build since r286752.
* math: Add expm1 builtin function
* math: Add logb builtin
* math: Add native_rsqrt builtin function
* Add the correct prefixes to the cl_khr_fp64 pragma
* Move BufferPtr into the block where it it being used
* math: Add native_tan as wrapper to tan
* .gitignore: Ignore amdgcn-mesa object directory
* math: Implement sinh function
OBS-URL: https://build.opensuse.org/request/show/487675
OBS-URL: https://build.opensuse.org/package/show/X11:XOrg/libclc?expand=0&rev=14
