Skip to content

Commit

Permalink
Bump default HPC-X version to 2.16 (#464)
Browse files Browse the repository at this point in the history
And adjust for another package naming change.
  • Loading branch information
samcmill authored Aug 31, 2023
1 parent 60fd2a5 commit 8c75c5d
Show file tree
Hide file tree
Showing 3 changed files with 90 additions and 54 deletions.
25 changes: 17 additions & 8 deletions docs/building_blocks.md
Original file line number Diff line number Diff line change
Expand Up @@ -1468,15 +1468,19 @@ Stage1 += h.runtime()
hpcx(self, **kwargs)
```
The `hpcx` building block downloads and installs the [Mellanox
HPC-X](https://www.mellanox.com/page/products_dyn?product_family=189&mtag=hpc-x)
HPC-X](https://developer.nvidia.com/networking/hpc-x)
component.

__Parameters__


- __buildlabel__: The build label assigned by Mellanox to the tarball.
This value is ignored for HPC-X version 2.10 and earlier. The
default value is `cuda12-gdrcopy2-nccl2.17`.
For version 2.16 the default value is `cuda12-gdrcopy2-nccl2.18`.
For version 2.15 the default value is `cuda12-gdrcopy2-nccl2.17`.
For version 2.14 the default value is `cuda11-gdrcopy2-nccl2.16`.
For versions 2.12 and 2.13 the default value is `cuda11-gdrcopy2-nccl2.12`.
For versions 2.10 and 2.11 the default value is `cuda11-gdrcopy2-nccl2.11`.
This value is ignored for HPC-X version 2.9 and earlier.

- __environment__: Boolean flag to specify whether the environment
should be modified to include HPC-X. This option is only
Expand Down Expand Up @@ -1508,19 +1512,24 @@ library directories. This value is ignored if `hpcxinit` is

- __mlnx_ofed__: The version of Mellanox OFED that should be matched.
This value is ignored if Inbox OFED is selected. The default
value is `5` for HPC-X version 2.11 and later, and `5.2-2.2.0.0`
value is `5` for HPC-X version 2.10 and later, and `5.2-2.2.0.0`
for earlier HPC-X versions.

- __multi_thread__: Boolean flag to specify whether the multi-threaded
version of Mellanox HPC-X should be used. The default is `False`.

- __ofedlabel__: The Mellanox OFED label assigned by Mellanox to the
tarball. For version 2.16 and later, the default value is
`gcc-mlnx_ofed`. For earlier versions, the default value is
`gcc-MLNX_OFED_LINUX-5`. This value is ignored if `inbox` is `True`.

- __oslabel__: The Linux distribution label assigned by Mellanox to the
tarball. For Ubuntu, the default value is `ubuntu16.04` for
Ubuntu 16.04, `ubuntu18.04` for Ubuntu 18.04, `ubuntu20.04` for
Ubuntu 20.04, and `ubuntu22.04` for Ubuntu 22.04. For HPC-X
version 2.11 and later and RHEL-based Linux distributions, the
version 2.10 and later and RHEL-based Linux distributions, the
default value is `redhat7` for version 7 and `redhat8` for version
8. For HPC-X version 2.10 and earlier and RHEL-based Linux
8. For HPC-X version 2.9 and earlier and RHEL-based Linux
distributions, the default value is `redhat7.6` for version 7 and
`redhat8.0` for version 8.

Expand All @@ -1534,13 +1543,13 @@ distributions the default values are `bzip2`, `numactl-libs`,
`/usr/local/hpcx`.

- __version__: The version of Mellanox HPC-X to install. The default
value is `2.15`.
value is `2.16`.

__Examples__


```python
hpcx(prefix='/usr/local/hpcx', version='2.15')
hpcx(prefix='/usr/local/hpcx', version='2.16')
```


Expand Down
69 changes: 48 additions & 21 deletions hpccm/building_blocks/hpcx.py
Original file line number Diff line number Diff line change
Expand Up @@ -43,14 +43,18 @@
class hpcx(bb_base, hpccm.templates.envvars, hpccm.templates.ldconfig,
hpccm.templates.rm, hpccm.templates.tar, hpccm.templates.wget):
"""The `hpcx` building block downloads and installs the [Mellanox
HPC-X](https://www.mellanox.com/page/products_dyn?product_family=189&mtag=hpc-x)
HPC-X](https://developer.nvidia.com/networking/hpc-x)
component.
# Parameters
buildlabel: The build label assigned by Mellanox to the tarball.
This value is ignored for HPC-X version 2.10 and earlier. The
default value is `cuda12-gdrcopy2-nccl2.17`.
For version 2.16 the default value is `cuda12-gdrcopy2-nccl2.18`.
For version 2.15 the default value is `cuda12-gdrcopy2-nccl2.17`.
For version 2.14 the default value is `cuda11-gdrcopy2-nccl2.16`.
For versions 2.12 and 2.13 the default value is `cuda11-gdrcopy2-nccl2.12`.
For versions 2.10 and 2.11 the default value is `cuda11-gdrcopy2-nccl2.11`.
This value is ignored for HPC-X version 2.9 and earlier.
environment: Boolean flag to specify whether the environment
should be modified to include HPC-X. This option is only
Expand Down Expand Up @@ -82,19 +86,24 @@ class hpcx(bb_base, hpccm.templates.envvars, hpccm.templates.ldconfig,
mlnx_ofed: The version of Mellanox OFED that should be matched.
This value is ignored if Inbox OFED is selected. The default
value is `5` for HPC-X version 2.11 and later, and `5.2-2.2.0.0`
value is `5` for HPC-X version 2.10 and later, and `5.2-2.2.0.0`
for earlier HPC-X versions.
multi_thread: Boolean flag to specify whether the multi-threaded
version of Mellanox HPC-X should be used. The default is `False`.
ofedlabel: The Mellanox OFED label assigned by Mellanox to the
tarball. For version 2.16 and later, the default value is
`gcc-mlnx_ofed`. For earlier versions, the default value is
`gcc-MLNX_OFED_LINUX-5`. This value is ignored if `inbox` is `True`.
oslabel: The Linux distribution label assigned by Mellanox to the
tarball. For Ubuntu, the default value is `ubuntu16.04` for
Ubuntu 16.04, `ubuntu18.04` for Ubuntu 18.04, `ubuntu20.04` for
Ubuntu 20.04, and `ubuntu22.04` for Ubuntu 22.04. For HPC-X
version 2.11 and later and RHEL-based Linux distributions, the
version 2.10 and later and RHEL-based Linux distributions, the
default value is `redhat7` for version 7 and `redhat8` for version
8. For HPC-X version 2.10 and earlier and RHEL-based Linux
8. For HPC-X version 2.9 and earlier and RHEL-based Linux
distributions, the default value is `redhat7.6` for version 7 and
`redhat8.0` for version 8.
Expand All @@ -108,12 +117,12 @@ class hpcx(bb_base, hpccm.templates.envvars, hpccm.templates.ldconfig,
`/usr/local/hpcx`.
version: The version of Mellanox HPC-X to install. The default
value is `2.15`.
value is `2.16`.
# Examples
```python
hpcx(prefix='/usr/local/hpcx', version='2.15')
hpcx(prefix='/usr/local/hpcx', version='2.16')
```
"""
Expand All @@ -127,27 +136,45 @@ def __init__(self, **kwargs):
self.__baseurl = kwargs.get('baseurl',
'https://content.mellanox.com/hpc/hpc-x')
self.__bashrc = '' # Filled in by __distro()
self.__buildlabel = kwargs.get('buildlabel',
'cuda12-gdrcopy2-nccl2.17')
self.__buildlabel = kwargs.get('buildlabel', None)
self.__hpcxinit = kwargs.get('hpcxinit', True)
self.__inbox = kwargs.get('inbox', False)
self.__mlnx_ofed = kwargs.get('mlnx_ofed', None)
self.__multi_thread = kwargs.get('multi_thread', False)
self.__ofedlabel = kwargs.get('ofedlabel', None)
self.__oslabel = kwargs.get('oslabel', '') # Filled in by __distro()
self.__ospackages = kwargs.get('ospackages', []) # Filled in by _distro()
self.__packages = kwargs.get('packages', [])
self.__prefix = kwargs.get('prefix', '/usr/local/hpcx')
self.__version = kwargs.get('version', '2.15')
self.__version = kwargs.get('version', '2.16')

self.__commands = [] # Filled in by __setup()
self.__wd = kwargs.get('wd', hpccm.config.g_wd) # working directory

if not self.__buildlabel:
if StrictVersion(self.__version) >= StrictVersion('2.16'):
self.__buildlabel = 'cuda12-gdrcopy2-nccl2.18'
elif StrictVersion(self.__version) >= StrictVersion('2.15'):
self.__buildlabel = 'cuda12-gdrcopy2-nccl2.17'
elif StrictVersion(self.__version) >= StrictVersion('2.14'):
self.__buildlabel = 'cuda11-gdrcopy2-nccl2.16'
elif StrictVersion(self.__version) >= StrictVersion('2.12'):
self.__buildlabel = 'cuda11-gdrcopy2-nccl2.12'
elif StrictVersion(self.__version) >= StrictVersion('2.10'):
self.__buildlabel = 'cuda11-gdrcopy2-nccl2.11'

if not self.__mlnx_ofed:
if StrictVersion(self.__version) >= StrictVersion('2.11'):
if StrictVersion(self.__version) >= StrictVersion('2.10'):
self.__mlnx_ofed = '5'
else:
self.__mlnx_ofed = '5.2-2.2.0.0'

if not self.__ofedlabel:
if StrictVersion(self.__version) >= StrictVersion('2.16'):
self.__ofedlabel = 'gcc-mlnx_ofed'
else:
self.__ofedlabel = 'gcc-MLNX_OFED_LINUX-{}'.format(self.__mlnx_ofed)

# Output toolchain
self.toolchain = toolchain(CC='mpicc', CXX='mpicxx', F77='mpif77',
F90='mpif90', FC='mpifort')
Expand Down Expand Up @@ -192,12 +219,12 @@ def __distro(self):
elif hpccm.config.g_linux_distro == linux_distro.CENTOS:
if not self.__oslabel:
if hpccm.config.g_linux_version >= StrictVersion('8.0'):
if StrictVersion(self.__version) >= StrictVersion('2.11'):
if StrictVersion(self.__version) >= StrictVersion('2.10'):
self.__oslabel = 'redhat8'
else:
self.__oslabel = 'redhat8.0'
else:
if StrictVersion(self.__version) >= StrictVersion('2.11'):
if StrictVersion(self.__version) >= StrictVersion('2.10'):
self.__oslabel = 'redhat7'
else:
self.__oslabel = 'redhat7.6'
Expand Down Expand Up @@ -227,7 +254,7 @@ def __setup(self):

if self.__inbox:
# Use inbox OFED
if StrictVersion(self.__version) >= StrictVersion('2.11'):
if StrictVersion(self.__version) >= StrictVersion('2.10'):
# Version 2.11 and later include an extra label
self.__label = 'hpcx-v{0}-gcc-inbox-{1}-{2}-{3}'.format(
self.__version, self.__oslabel, self.__buildlabel,
Expand All @@ -237,13 +264,13 @@ def __setup(self):
self.__version, self.__oslabel, self.__arch)
else:
# Use MLNX OFED
if StrictVersion(self.__version) >= StrictVersion('2.11'):
# Version 2.11 and later include an extra label
self.__label = 'hpcx-v{0}-gcc-MLNX_OFED_LINUX-{1}-{2}-{3}-{4}'.format(
self.__version, self.__mlnx_ofed, self.__oslabel, self.__buildlabel, self.__arch)
if StrictVersion(self.__version) >= StrictVersion('2.10'):
# Version 2.10 and later include an extra label
self.__label = 'hpcx-v{0}-{1}-{2}-{3}-{4}'.format(
self.__version, self.__ofedlabel, self.__oslabel, self.__buildlabel, self.__arch)
else:
self.__label = 'hpcx-v{0}-gcc-MLNX_OFED_LINUX-{1}-{2}-{3}'.format(
self.__version, self.__mlnx_ofed, self.__oslabel, self.__arch)
self.__label = 'hpcx-v{0}-{1}-{2}-{3}'.format(
self.__version, self.__ofedlabel, self.__oslabel, self.__arch)

tarball = self.__label + '.tbz'
url = '{0}/v{1}/{2}'.format(self.__baseurl, version_string, tarball)
Expand Down
50 changes: 25 additions & 25 deletions test/test_hpcx.py
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ def test_defaults_ubuntu18(self):
"""Default hpcx building block"""
h = hpcx()
self.assertEqual(str(h),
r'''# Mellanox HPC-X version 2.15
r'''# Mellanox HPC-X version 2.16
RUN apt-get update -y && \
DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
bzip2 \
Expand All @@ -47,12 +47,12 @@ def test_defaults_ubuntu18(self):
tar \
wget && \
rm -rf /var/lib/apt/lists/*
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.15/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu18.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu18.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu18.04-cuda12-gdrcopy2-nccl2.17-x86_64 /usr/local/hpcx && \
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.16/hpcx-v2.16-gcc-mlnx_ofed-ubuntu18.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu18.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu18.04-cuda12-gdrcopy2-nccl2.18-x86_64 /usr/local/hpcx && \
echo "source /usr/local/hpcx/hpcx-init-ompi.sh" >> /etc/bash.bashrc && \
echo "hpcx_load" >> /etc/bash.bashrc && \
rm -rf /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu18.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu18.04-cuda12-gdrcopy2-nccl2.17-x86_64''')
rm -rf /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu18.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu18.04-cuda12-gdrcopy2-nccl2.18-x86_64''')

@x86_64
@ubuntu20
Expand All @@ -61,7 +61,7 @@ def test_defaults_ubuntu20(self):
"""Default hpcx building block"""
h = hpcx()
self.assertEqual(str(h),
r'''# Mellanox HPC-X version 2.15
r'''# Mellanox HPC-X version 2.16
RUN apt-get update -y && \
DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
bzip2 \
Expand All @@ -70,12 +70,12 @@ def test_defaults_ubuntu20(self):
tar \
wget && \
rm -rf /var/lib/apt/lists/*
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.15/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu20.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu20.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu20.04-cuda12-gdrcopy2-nccl2.17-x86_64 /usr/local/hpcx && \
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.16/hpcx-v2.16-gcc-mlnx_ofed-ubuntu20.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu20.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu20.04-cuda12-gdrcopy2-nccl2.18-x86_64 /usr/local/hpcx && \
echo "source /usr/local/hpcx/hpcx-init-ompi.sh" >> /etc/bash.bashrc && \
echo "hpcx_load" >> /etc/bash.bashrc && \
rm -rf /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu20.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu20.04-cuda12-gdrcopy2-nccl2.17-x86_64''')
rm -rf /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu20.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu20.04-cuda12-gdrcopy2-nccl2.18-x86_64''')

@x86_64
@centos
Expand All @@ -84,20 +84,20 @@ def test_defaults_centos7(self):
"""Default mlnx_ofed building block"""
h = hpcx()
self.assertEqual(str(h),
r'''# Mellanox HPC-X version 2.15
r'''# Mellanox HPC-X version 2.16
RUN yum install -y \
bzip2 \
numactl-libs \
openssh-clients \
tar \
wget && \
rm -rf /var/cache/yum/*
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.15/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat7-cuda12-gdrcopy2-nccl2.17-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat7-cuda12-gdrcopy2-nccl2.17-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat7-cuda12-gdrcopy2-nccl2.17-x86_64 /usr/local/hpcx && \
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.16/hpcx-v2.16-gcc-mlnx_ofed-redhat7-cuda12-gdrcopy2-nccl2.18-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat7-cuda12-gdrcopy2-nccl2.18-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat7-cuda12-gdrcopy2-nccl2.18-x86_64 /usr/local/hpcx && \
echo "source /usr/local/hpcx/hpcx-init-ompi.sh" >> /etc/bashrc && \
echo "hpcx_load" >> /etc/bashrc && \
rm -rf /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat7-cuda12-gdrcopy2-nccl2.17-x86_64.tbz /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat7-cuda12-gdrcopy2-nccl2.17-x86_64''')
rm -rf /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat7-cuda12-gdrcopy2-nccl2.18-x86_64.tbz /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat7-cuda12-gdrcopy2-nccl2.18-x86_64''')

@x86_64
@centos8
Expand All @@ -106,20 +106,20 @@ def test_defaults_centos8(self):
"""Default mlnx_ofed building block"""
h = hpcx()
self.assertEqual(str(h),
r'''# Mellanox HPC-X version 2.15
r'''# Mellanox HPC-X version 2.16
RUN yum install -y \
bzip2 \
numactl-libs \
openssh-clients \
tar \
wget && \
rm -rf /var/cache/yum/*
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.15/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat8-cuda12-gdrcopy2-nccl2.17-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat8-cuda12-gdrcopy2-nccl2.17-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat8-cuda12-gdrcopy2-nccl2.17-x86_64 /usr/local/hpcx && \
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.16/hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64 /usr/local/hpcx && \
echo "source /usr/local/hpcx/hpcx-init-ompi.sh" >> /etc/bashrc && \
echo "hpcx_load" >> /etc/bashrc && \
rm -rf /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat8-cuda12-gdrcopy2-nccl2.17-x86_64.tbz /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-redhat8-cuda12-gdrcopy2-nccl2.17-x86_64''')
rm -rf /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64.tbz /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64''')

@x86_64
@ubuntu
Expand Down Expand Up @@ -291,7 +291,7 @@ def test_runtime(self):
h = hpcx()
r = h.runtime()
self.assertEqual(r,
r'''# Mellanox HPC-X version 2.15
r'''# Mellanox HPC-X version 2.16
RUN apt-get update -y && \
DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
bzip2 \
Expand All @@ -300,9 +300,9 @@ def test_runtime(self):
tar \
wget && \
rm -rf /var/lib/apt/lists/*
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.15/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu22.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu22.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu22.04-cuda12-gdrcopy2-nccl2.17-x86_64 /usr/local/hpcx && \
RUN mkdir -p /var/tmp && wget -q -nc --no-check-certificate -P /var/tmp https://content.mellanox.com/hpc/hpc-x/v2.16/hpcx-v2.16-gcc-mlnx_ofed-ubuntu22.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz && \
mkdir -p /var/tmp && tar -x -f /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu22.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz -C /var/tmp -j && \
cp -a /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu22.04-cuda12-gdrcopy2-nccl2.18-x86_64 /usr/local/hpcx && \
echo "source /usr/local/hpcx/hpcx-init-ompi.sh" >> /etc/bash.bashrc && \
echo "hpcx_load" >> /etc/bash.bashrc && \
rm -rf /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu22.04-cuda12-gdrcopy2-nccl2.17-x86_64.tbz /var/tmp/hpcx-v2.15-gcc-MLNX_OFED_LINUX-5-ubuntu22.04-cuda12-gdrcopy2-nccl2.17-x86_64''')
rm -rf /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu22.04-cuda12-gdrcopy2-nccl2.18-x86_64.tbz /var/tmp/hpcx-v2.16-gcc-mlnx_ofed-ubuntu22.04-cuda12-gdrcopy2-nccl2.18-x86_64''')

0 comments on commit 8c75c5d

Please sign in to comment.