New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Tensorflow 2.16.1 fails to compile with -march=native on Xeon 4410Y CPU #64221
Comments
I just tested building without cuda. I ran configure again, provided "-march=native" as the "opt" compiler flag, and then built with --config=opt, without modifying anything other than setting TF_PYTHON_VERSION=3.10. That failed with the same static_cast error in MathFunctions.h. I changed the issue subject to remove the mention of CUDA since it happens without it too. |
Can you try disabling avx512_fp16? For |
@QuesarVII, |
Yes, avx512_fp16 triggers the issue. Using -mno-avx512fp16 results in a successful build. Thanks. |
@QuesarVII |
Shouldn't the code base be fixed to resolve this issue before closing it? Disabling compiler flags is more of a workaround than a real fix. Thanks! |
@QuesarVII, |
This issue is stale because it has been open for 7 days with no activity. It will be closed if no further activity occurs. Thank you. |
This issue was closed because it has been inactive for 7 days since being marked as stale. Please reopen if you'd like to work on this further. |
Issue type
Build/Install
Have you reproduced the bug with TensorFlow Nightly?
Yes
Source
source
TensorFlow version
2.16.1
Custom code
No
OS platform and distribution
x86_64 Ubuntu 22.04
Mobile device
No response
Python version
3.10
Bazel version
6.5.0
GCC/compiler version
clang 17.0.6 (ubuntu package)
CUDA/cuDNN version
12.4 / 8.9.7
GPU model and memory
(2) RTX A4500 20GB
Current behavior?
Tensorflow 2.16.1 fails to compile with -march=native when building with CUDA support on Xeon 4410Y CPUs with clang 17.0.6. I run configure saying yes to cuda support, then edit .tf_configure.bazelrc and add these lines:
build:cuda --copt=-march=native
build:cuda --host_copt=-march=native
build:cuda --copt=-Wno-error=unused-command-line-argument (see issue 62459 regarding "unused"
and set TF_PYTHON_VERSION=3.10, the bazel build errors out on tensorflow/core/kernels/matmul_op_real.cc in external/eigen_archive/Eigen/src/Core/MathFunctions.h. The error output is in the log field.
This appears to be the same problem as issue 62047.
Standalone code to reproduce the issue
Relevant log output
The text was updated successfully, but these errors were encountered: