Andreas Beckmann

Results 32 issues of Andreas Beckmann

quoting from https://bugs.debian.org/946477 > povray fails to cross build from source, because it uses AC_RUN_IFELSE > in a few places. Getting rid of these isn't obvious. It's used for >...

compatibility
OS: Unix

Some platforms (e.g. mipsel) need to link with -latomic for 64-bit atomics. Switch to the modern __atomic_*() builtins (available from gcc 4.7 onwards) that are backed by -latomic (from gcc)...

Hi, I've been debugging the printf issues a bit further, looking at `ulong16` failing where `ulong2` succeeds. The issue I'm describing here happens on `x86_64` and is not a bug...

In https://bugs.debian.org/975931 a segmentation fault in LLVM 10/11 (but not 9) on armhf was reported where pocl was used as the backend for some libgpuarray/libclblas/pyopencl based test. I've reduced that...

ARM

Subdevice creation obviously requires multiple CUs, but if only a single CU is available the test should not fail but rather be skipped. Test in CMake can have the `SKIP_RETURN_CODE`...

I noticed this bug in 1.2 and could reproduce it in current git master. This happens on 32-bit x86, i.e. Debian i386 and with both llvm 6.0 and llvm 7....

i686

While looking further into https://github.com/pocl/pocl/issues/801 on Debian i386 I found some inconsistencies in the alignment of `cl_*` types resulting in a `struct {cl_int; cl_long;};` to have size `12` on i386...

`examples/demo_cdpSimplePrint.py` comes with a problematic license: ``` * Adapted from NVIDIA's "cdpSimplePrint - Simple Print (CUDA Dynamic Parallelism)" sample * http://docs.nvidia.com/cuda/cuda-samples/index.html#simple-print--cuda-dynamic-parallelism- * http://ecee.colorado.edu/~siewerts/extra/code/example_code_archive/a490dmis_code/CUDA/cuda_work/samples/0_Simple/cdpSimplePrint/cdpSimplePrint.cu * * From cdpSimplePrint.cu (not sure if...

bug

In an attempt to work around #1176 I tried building with `-DEXTRA_HOST_CLANG_FLAGS=-frounding-math` (or `-ffp-model=strict`, both fix the plain C reproducer). That causes 137 tests to fail, mostly with `undefined symbol:...

Right now we have BUILD_EXCLUSIVE_KERNEL as a regex that whitelists kernel versions. With the recently added BUILD_EXCLUSIVE_KERNEL_MIN/MAX that whitelist regex probably gets less used. But sometimes I'd rather like to...