iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CPU] Enable tcmalloc for s390x (#50841)

Browse Source 
Signed-off-by: Rehan Khan <Rehan.Khan7@ibm.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
This commit is contained in:
Rehan Khan
2026-08-05 10:55:39 +08:00
committed by GitHub
co-authored by Li, Jiang <jiang1.li@intel.com>
parent 2cb3ff88cc
commit 0187f4c88e
4 changed files with 39 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/getting_started/installation/cpu.s390x.inc.md
+21
View File
@@ -117,6 +117,27 @@ VLLM_TARGET_DEVICE=cpu VLLM_CPU_MOE_PREPACK=0 python setup.py bdist_wheel && \
pip install dist/*.whl
```
!!! warning "set `LD_PRELOAD` for TCMalloc"
For best memory allocation performance, build and install
[gperftools](https://github.com/gperftools/gperftools) (TCMalloc) from
source and add it to `LD_PRELOAD`:
```bash
# Build and install TCMalloc from source
curl -LO https://github.com/gperftools/gperftools/releases/download/gperftools-2.16/gperftools-2.16.tar.gz
tar -xzf gperftools-2.16.tar.gz
cd gperftools-2.16
./configure --enable-minimal && make -j$(nproc) && sudo make install
sudo ldconfig
cd ..
# Add to LD_PRELOAD
export LD_PRELOAD="/usr/local/lib/libtcmalloc_minimal.so.4:$LD_PRELOAD"
```
The Docker image (`Dockerfile.s390x`) already includes TCMalloc and sets
`LD_PRELOAD` automatically.
!!! warning "Protobuf workaround for s390x"
The C++ protobuf extension crashes on s390x. After installation, set the
following environment variable and remove the C++ extensions: