| commit | 41e8c71efb7efdb55fe83eab1bfde55b3a46bad4 | [log] [tgz] |
|---|---|---|
| author | Deven Desai <deven.desai.amd@gmail.com> | Sun Jan 31 01:52:09 2021 +0000 |
| committer | Deven Desai <deven.desai.amd@gmail.com> | Wed Feb 03 14:14:25 2021 +0000 |
| tree | e20307f8df175b013b52c05dbf3e7fc16069a4e5 | |
| parent | bff7bf1e48d40a5cc95c291fea042c93e5c20af1 [diff] |
[ROCm] Workaround for a LLVM crash when doing codegen for MLIR generated Cast kernel
When you enable (uncomment) the `gen_kernel_library` rule for `cast` in `tensorflow/core/kernels/mlir_generated/BUILD`, the build would fail on the ROCm platform with the following error
```
LLVM ERROR: Cannot select: 0x56134e3c5b10: i1 = fp_to_sint 0x56134d1f53d8
0x56134d1f53d8: f16 = bitcast 0x56134e3c58a0
0x56134e3c58a0: i16,ch = load<(load 2 from %ir.lsr.iv)> 0x56134eaa3788, 0x56134e3c5698, undef:i64
0x56134e3c5698: i64,ch = CopyFromReg 0x56134eaa3788, Register:i64 %16
0x56134d41d620: i64 = Register %16
0x56134d1f5850: i64 = undef
In function: Cast_f16_i1_kernel
TensorFlow crashed, please file a bug on https://github.com/tensorflow/tensorflow/issues with the trace below.
Stack dump:
0. Program arguments: bazel-out/host/bin/tensorflow/compiler/mlir/tools/kernel_gen/tf_to_kernel --unroll_factors=4 --tile_sizes=256 --arch=gfx803,gfx900,gfx906,gfx908 --input=bazel-out/k8-opt/bin/tensorflow/core/kernels/mlir_generated/cast_f16_i1.mlir --output=bazel-out/k8-opt/bin/tensorflow/core/kernels/mlir_generated/cast_f16_i1_kernel_generator_kernel.o --enable_ftz=False
1. 2. Running pass 'CallGraph Pass Manager' on module 'acme'.
3. Running pass 'AMDGPU DAG->DAG Pattern Instruction Selection' on function '@Cast_f16_i1_kernel'
...
...
```
Jack (@whchung) identified the root cause of this crash as lack of support for FPTOSI (fp16 to i1) instruction in the AMDGPU LLVM backend. The correct fix for this bug, is to add support for the same in the AMDGPU LLVM backend. Jack is in the process of upstreaming that fix to the LLVM repo.
In the meantime (i.e. until the TF LLVM commit pointer is updated to point to a commit that includes Jack's fix), we need to workaround this on the TF side, by adding a pass that converts the `fptosi f16 to i1` op to `fptosi f16 to i16` + `trunci i16 to i1`, which is what this commit does.
Documentation |
|---|
 |
TensorFlow is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art in ML and developers easily build and deploy ML-powered applications.
TensorFlow was originally developed by researchers and engineers working on the Google Brain team within Google's Machine Intelligence Research organization to conduct machine learning and deep neural networks research. The system is general enough to be applicable in a wide variety of other domains, as well.
TensorFlow provides stable Python and C++ APIs, as well as non-guaranteed backward compatible API for other languages.
Keep up-to-date with release announcements and security updates by subscribing to announce@tensorflow.org. See all the mailing lists.
See the TensorFlow install guide for the pip package, to enable GPU support, use a Docker container, and build from source.
To install the current release, which includes support for CUDA-enabled GPU cards (Ubuntu and Windows):
$ pip install tensorflow
A smaller CPU-only package is also available:
$ pip install tensorflow-cpu
To update TensorFlow to the latest version, add --upgrade flag to the above commands.
Nightly binaries are available for testing using the tf-nightly and tf-nightly-cpu packages on PyPi.
$ python
>>> import tensorflow as tf >>> tf.add(1, 2).numpy() 3 >>> hello = tf.constant('Hello, TensorFlow!') >>> hello.numpy() b'Hello, TensorFlow!'
For more examples, see the TensorFlow tutorials.
If you want to contribute to TensorFlow, be sure to review the contribution guidelines. This project adheres to TensorFlow's code of conduct. By participating, you are expected to uphold this code.
We use GitHub issues for tracking requests and bugs, please see TensorFlow Discuss for general questions and discussion, and please direct specific questions to Stack Overflow.
The TensorFlow project strives to abide by generally accepted best practices in open-source software development:
| Build Type | Status | Artifacts |
|---|---|---|
| Linux CPU |  | PyPI |
| Linux GPU |  | PyPI |
| Linux XLA |  | TBA |
| macOS |  | PyPI |
| Windows CPU |  | PyPI |
| Windows GPU |  | PyPI |
| Android |  |  |
| Raspberry Pi 0 and 1 |  | Py3 |
| Raspberry Pi 2 and 3 |  | Py3 |
| Libtensorflow MacOS CPU |  | Nightly GCS Official GCS |
| Libtensorflow Linux CPU |  | Nightly GCS Official GCS |
| Libtensorflow Linux GPU |  | Nightly GCS Official GCS |
| Libtensorflow Windows CPU |  | Nightly GCS Official GCS |
| Libtensorflow Windows GPU |  | Nightly GCS Official GCS |
| Build Type | Status | Artifacts |
|---|---|---|
| Linux AMD ROCm GPU Nightly | Nightly | |
| Linux AMD ROCm GPU Stable Release | Release 1.15 / 2.x | |
| Linux s390x Nightly | Nightly | |
| Linux s390x CPU Stable Release | Release | |
| Linux ppc64le CPU Nightly | Nightly | |
| Linux ppc64le CPU Stable Release | Release 1.15 / 2.x | |
| Linux ppc64le GPU Nightly | Nightly | |
| Linux ppc64le GPU Stable Release | Release 1.15 / 2.x | |
| Linux aarch64 CPU Nightly (Linaro) | Nightly | |
| Linux aarch64 CPU Stable Release (Linaro) | Release 1.x & 2.x | |
| Linux aarch64 CPU Nightly (OpenLab) Python 3.6 | Nightly | |
| Linux aarch64 CPU Stable Release (OpenLab) | Release 1.15 / 2.x | |
| Linux CPU with Intel oneAPI Deep Neural Network Library (oneDNN) Nightly | Nightly | |
| Linux CPU with Intel oneAPI Deep Neural Network Library (oneDNN) Stable Release | Release 1.15 / 2.x | |
| Red Hat® Enterprise Linux® 7.6 CPU & GPU Python 2.7, 3.6 | 1.13.1 PyPI |
| Container Type | Status | Artifacts |
|---|---|---|
| TensorFlow aarch64 Neoverse-N1 CPU Stable (Linaro) Debian | Static | Release 2.3 |
Learn more about the TensorFlow community and how to contribute.