Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package libclc for openSUSE:Factory checked in at 2026-04-23 17:05:33 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/libclc (Old) and /work/SRC/openSUSE:Factory/.libclc.new.11940 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "libclc" Thu Apr 23 17:05:33 2026 rev:37 rq:1348453 version:0.2.0+llvm22.1.0 Changes: -------- --- /work/SRC/openSUSE:Factory/libclc/libclc.changes 2026-01-26 10:44:49.113990259 +0100 +++ /work/SRC/openSUSE:Factory/.libclc.new.11940/libclc.changes 2026-04-23 17:09:35.717052440 +0200 @@ -1,0 +2,87 @@ +Tue Apr 21 09:03:23 UTC 2026 - Stefan Dirsch <[email protected]> + +- use 0%{?suse_version} >= 1699 for TW (> 1600 now includes Leap 16.1) + +------------------------------------------------------------------- +Mon Apr 20 17:32:45 UTC 2026 - Aaron Puchert <[email protected]> + +- Update to version 22.1.0. For now we build the tarball ourselves, + because upstream no longer provides it. (With LLVM 23, it seem + that a standalone build will be unsupported anyway.) + * Remove llvm-link --override flag and make implementation + self-contained. + * Initial support for cross-compiling OpenCL libraries. + * Add support for `nvptx64-nvidia-cuda` triple. + * Remove bitcode prepare utility from OpenCL library build. + * Update default triple support for AMD and deprecate 32-bit PTX. + * Compile clc_convert_float2int.cl and clc_convert_int2float.cl + separately. + * Fix incorrect argument format. + * Suppress AMDGCN code object version for library code. + * Improve __clc_min/max/clamp implementation. + * Use clc functions in clspv/shared/vstore_half.cl. + * Fix clspv/shared/vstore_half.cl. + * Add OpenCL atomic_*_explicit builtins. + * Fix memory fence scope mapping for OpenCL. + * Fix build in atomic_def.inc. + * Fix bitfield_insert implementation. + * Use __scoped_atomic_udec/uinc_wrap to implement + _clc_atomic_dec/inc. + * Add atomic_init, atomic_flag_clear and + atomic_flag_test_and_set. + * Use CLC atomic functions for legacy OpenCL atom/atomic + builtins. + * Fix link to source in index.html. + * Fix floating-point __clc_atomic_store/exchange cast mismatch. + * Implement integer __clc_abs using __builtin_elementwise_abs. + * Fix double NAN_MASK in __clc_nan. + * Move functions definition from header clc_sincos_piby4.inc into + clc_sincos_helpers.cl. + * Change libclc install dir to ${clang_resource_dir}/lib/libclc + in in-tree build. + * Remove -fno-builtin from compile options. + * Add gfx1250 and gfx1251 to amdgpu target list. + * Remove __attribute__((always_inline)). + * Create LIBCLC_OUTPUT_LIBRARY_DIR directory before build. + * Remove HAVE_LLVM version macros. + * Implement erf/erfc vector function with loop since scalar + function is large. + * Override generic symbol using llvm-link --override flag instead + of using weak linkage. + * Update __clc_mem_fence: add MemorySemantic arg and use + __builtin_amdgcn_fence for AMDGPU. + * Only create a target per each compile command for cmake MSVC + generator. + * Use elementwise ctlz/cttz builtins for CLC clz/ctz. + * Implement __clc_get_local_size/__clc_get_max_sub_group_size for + amdgcn. + * Fix out-of-bound value for workitem functions according to + OpenCL spec. + * Add __attribute__((const)) to functions that don't access + memory. + * Fix libclc install on Windows when MSVC generator is used. + * Implement clc_log/sinpi/sqrt with __nv_* functions. + * Move mem_fence and barrier to clc library. + * Refine id in async_work_group_copy STRIDED_COPY. + * Add an option to build SPIR-V targets with the LLVM backend. + * Optimize generic CLC fmin/fmax. + * Fix building top-level 'libclc' target. + * Add generic native half implementation of __clc_normalize. + * Simplify unary_def_scalarize.inc's use in + __clc_erf/erfc/tgamma. + * Fix installed symlinks to be relative again. + * Add generic implementation of some atomic functions in OpenCL + spec section 6.15.12.7. + * Add generic implementation of + bitfield_insert/extract,bit_reverse. + * Enable `clang fp reciprocal` in + clc_native_divide/recip/rsqrt/tan. + * Enable -fdiscard-value-names build flag to reduce bitcode size. +- Add fix-subnormal-build.patch to fix build of subnormal runtimes. + +------------------------------------------------------------------- +Mon Apr 20 16:50:53 UTC 2026 - Aaron Puchert <[email protected]> + +- Use clang/llvm-devel in Factory. + +------------------------------------------------------------------- Old: ---- libclc-21.1.0.src.tar.xz libclc-21.1.0.src.tar.xz.sig New: ---- fix-subnormal-build.patch libclc-22.1.0.src.tar.xz ----------(New B)---------- New: * Enable -fdiscard-value-names build flag to reduce bitcode size. - Add fix-subnormal-build.patch to fix build of subnormal runtimes. ----------(New E)---------- ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ libclc.spec ++++++ --- /var/tmp/diff_new_pack.SFaNDH/_old 2026-04-23 17:09:36.337077950 +0200 +++ /var/tmp/diff_new_pack.SFaNDH/_new 2026-04-23 17:09:36.341078114 +0200 @@ -16,7 +16,7 @@ # -%define _libclc_llvm_ver 21.1.0 +%define _libclc_llvm_ver 22.1.0 %define _version %_libclc_llvm_ver%{?_rc:rc%_rc} %define _tagver %_libclc_llvm_ver%{?_rc:-rc%_rc} @@ -27,13 +27,20 @@ License: Apache-2.0 WITH LLVM-exception AND (BSD-3-Clause OR MIT) Group: Development/Libraries/C and C++ URL: https://libclc.llvm.org/ -Source0: https://github.com/llvm/llvm-project/releases/download/llvmorg-%{_tagver}/%{name}-%{_version}.src.tar.xz -Source1: https://github.com/llvm/llvm-project/releases/download/llvmorg-%{_tagver}/%{name}-%{_version}.src.tar.xz.sig +# Built manually via "git archive --prefix=%{name}-%{_version}.src/ llvmorg-%{_tagver} . +# | xz -T0 >../%{name}-%{_version}.src.tar.xz" until upstream provides tarballs again. +Source0: %{name}-%{_version}.src.tar.xz Source100: %{name}-rpmlintrc Source101: https://releases.llvm.org/release-keys.asc#/%{name}.keyring -BuildRequires: clang21-devel +Patch1: fix-subnormal-build.patch BuildRequires: cmake +%if 0%{?suse_version} >= 1699 +BuildRequires: clang-devel +BuildRequires: llvm-devel +%else +BuildRequires: clang21-devel BuildRequires: llvm21-devel +%endif BuildRequires: python3-base BuildRequires: pkgconfig(LLVMSPIRVLib) Provides: libclc(llvm%{_llvm_sonum}) ++++++ fix-subnormal-build.patch ++++++ Fixes build error: [100%] Generating /subnormal_use_default.bc /usr/bin/llvm-link -o /subnormal_use_default.bc [...]/opencl/lib/generic/subnormal_use_default.ll error: Permission denied diff --git a/CMakeLists.txt b/CMakeLists.txt index d060b71..c17b350 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -239,6 +239,7 @@ configure_file( libclc.pc.in libclc.pc @ONLY ) install( FILES ${CMAKE_CURRENT_BINARY_DIR}/libclc.pc DESTINATION "${CMAKE_INSTALL_DATADIR}/pkgconfig" ) if( ENABLE_RUNTIME_SUBNORMAL ) + set( LIBCLC_ARCH_OBJFILE_DIR "${LIBCLC_OBJFILE_DIR}" ) foreach( file IN ITEMS subnormal_use_default subnormal_disable ) link_bc( TARGET ${file} ++++++ libclc-21.1.0.src.tar.xz -> libclc-22.1.0.src.tar.xz ++++++ ++++ 25850 lines of diff (skipped)
