Use virtual/opencl-icd instead of ocl-icd explicitly.
Signed-off-by: Zoltán Böszörményi <zboszor@gmail.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Ignore the buildpath being embedded in bitcode by clang until we've
found a way to fix.
| QA Issue: File /usr/lib/libigc.so.1.0.1 in package intel-graphics-compiler contains reference to TMPDIR [buildpaths]
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Clone SPIRV-Tools and Headers in S instead of WORKDIR so they don't
escape file-prefix-map substitutions.
Fixes buildpaths warnings:
| WARNING: intel-graphics-compiler-1.0.12812.9-r0 do_package_qa: QA Issue: File /usr/lib/.debug/libigc.so.1.0.1 in package intel-graphics-compiler-dbg contains reference to TMPDIR [buildpaths]
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Fixes:
/git/IGC/VectorCompiler/lib/GenXCodeGen/GenXSimdCFRegion.cpp:412:31: error: ‘experimental_vector_reduce_or’ is not a member of ‘llvm::Intrinsic’; did you mean ‘experimental_vector_insert’?
[2022-12-06T21:01:52.813Z] | 412 | Intrinsic::experimental_vector_reduce_or;
[2022-12-06T21:01:52.813Z] | | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[2022-12-06T21:01:52.813Z] | | experimental_vector_insert
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Upgrade to the latest tag and backport 3 patches to fix build issues
with LLVM 15 [1].
Remove IGC_OPTION__USE_KHRONOS_SPIRV_TRANSLATOR_IN_VC option as it has
been removed upstream [2].
And, make sure CLANG_EXE points to clang in native sysroot.
[1] https://github.com/intel/intel-graphics-compiler/issues/263
[2] 3fde0acae8
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Build opencl-clang with llvm-15 and point to latest from ocl-open-150
branch.
Signed-off-by: Naveen Saini <naveen.kumar.saini@intel.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
This is due to recent packaging chagnes in meta-clang
3d56a85afe
Error log:
do_package_qa: QA Issue: /usr/bin/ispc contained in package ispc
requires libclang-cpp.so.15()(64bit), but no providers found
in RDEPENDS:ispc? [file-rdeps]
Signed-off-by: Naveen Saini <naveen.kumar.saini@intel.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Upstream has force pushed master and the commit we were fetching is
no longer present on that branch. Remove the branch parameter and switch
to using nobranch to continue using that commit.
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
* Don't build with clang specifically and let user decide.
* Drop already merged patch: 0001-Enable-LLVM-15.0-support.patch
* ispc generated headers include a comment will full path to the header. Patch the code to drop this comment.
* Tweak bison and flex invocation to make sure they don't include #line directives.
* Add flex-native to DEPENDS.
* Release notes:
https://github.com/ispc/ispc/releases/tag/v1.18.0
Signed-off-by: Yogesh Tyagi <yogesh.tyagi@intel.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
* Remove upstreamed patches and refresh others.
* Include the LICENSE.md file and remove the cpp file which just points to
MIT license.
* Update the license to MIT and Apache-2.0
* Install vcb tool from -native recipe.
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
We no longer support building with older branches of OE-Core/meta-clang
so remove LLVM 10 specific configurations and patches.
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Backport patches to fix compile problems like:
| /build/build/tmp/work/corei7-64-poky-linux/intel-graphics-compiler/1.0.10395-r0/git/visa/LocalScheduler/SWSB_G4IR.cpp:1459:49: error: expected primary-expression before 'int'
| 1459 | int maxTokenDelay = std::numeric_limits<int>::min(); //The delay may cause if reuse
| | ^~~
| /build/build/tmp/work/corei7-64-poky-linux/intel-graphics-compiler/1.0.10395-r0/git/visa/LocalScheduler/SWSB_G4IR.cpp:1460:37: error: 'numeric_limits' is not a member of 'std'
| 1460 | int minTokenDistance = std::numeric_limits<int>::max(); //The distance from the reused node
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
*Fixed performance regression in binary primitive with broadcast (b972174, ff75122).
*Fixed issue with SYCL device properties initialization (cabc5ca, 095f13e).
*Fixed issue in matmul primitive with zero points (3157354).
*Fixed segmentation fault in depthwise convolution primitive for shapes with huge spatial size for processors with Intel AVX-512 support (6834764, 1d2addc).
*Fixed issue in forward convolution primitive for processors with Intel AVX2 support (d691137).
*Fixed performance regression on GPUs with SYCL runtime (d8364e5).
Signed-off-by: ezhilarasan s <ezhilarasanx.s@intel.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>