[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [UNIKRAFT/EIGEN PATCH] Prevent kludgeoning by barrier macro.
Hi,
Please ignore this patch, the error was a result of my local set up.
Thanks,
-- Felipe
On Fri, Apr 17, 2020 at 10:23 PM Felipe Huici <felipe.huici@xxxxxxxxx> wrote:
>
> Add patch to prevent the definition of the macro "barrier" in lcpu.h
> from kludgeoning the use of the word barrier in eigen code.
>
> Signed-off-by: Felipe Huici <felipe.huici@xxxxxxxxx
> ---
> patches/0002-prevent-barrier-macro-code-kludge.patch | 10 ++++++++++
> 1 file changed, 10 insertions(+)
> create mode 100644 patches/0002-prevent-barrier-macro-code-kludge.patch
>
> diff --git a/patches/0002-prevent-barrier-macro-code-kludge.patch b/patches/0002-prevent-barrier-macro-code-kludge.patch
> new file mode 100644
> index 0000000..0b3fe78
> --- /dev/null
> +++ b/patches/0002-prevent-barrier-macro-code-kludge.patch
> @@ -0,0 +1,10 @@
> +--- a/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h 2020-04-17 17:13:19.072796790 +0200
> ++++ b/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h 2021-04-17 17:13:41.800543414 +0200
> +@@ -98,6 +98,7 @@
> + const char* src_ptr = static_cast<const char*>(src);
> + char* dst_ptr = static_cast<char*>(dst);
> + const size_t blocksize = (n + (num_threads - 1)) / num_threads;
> ++#undef barrier
> + Barrier barrier(static_cast<int>(num_threads - 1));
> + // Launch the last 3 blocks on worker threads.
> + for (size_t i = 1; i < num_threads; ++i) {
> --
> 2.20.1
>
>
|