Skip to content

perf: use jemalloc for make_examples in the runtime image #1087

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.

Sign up for GitHub

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

Open
tfenne wants to merge 2 commits into
base: r1.10
Choose a base branch
from
+34 −10
Open

perf: use jemalloc for make_examples in the runtime image #1087

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
da4f616
perf: use jemalloc for make_examples in the runtime image
tfenne Jun 22, 2026
a632502
perf: use jemalloc for make_examples in the DeepTrio, DeepSomatic and…
tfenne Jun 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 10 additions & 4 deletions Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,7 +102,7 @@ ENV DV_BIN_PATH=/opt/deepvariant/bin

# Install libraries
RUN apt-get -y update && \
apt-get install -y parallel python3-pip unzip && \
apt-get install -y parallel python3-pip unzip libjemalloc2 && \
PATH="${HOME}/.local/bin:$PATH" python3 -m pip install absl-py==0.13.0 && \
apt-get clean autoclean && \
apt-get autoremove -y --purge && \
Expand Down Expand Up @@ -140,11 +140,17 @@ COPY --from=builder \
/opt/deepvariant/bin/

# Create shell wrappers for python zip files for easier use.
#
# The make_examples family is wrapped with LD_PRELOAD=libjemalloc.so.2: jemalloc
# meaningfully reduces make_examples wall-clock (its pileup/realignment work is
# allocation-heavy) while having no measurable effect on call_variants (TF
# inference), so the preload is scoped to just those wrappers. The bare soname
# keeps it architecture-portable (resolved from the default linker search path).
RUN \
BASH_HEADER='#!/bin/bash' && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'/usr/bin/python3 /opt/deepvariant/bin/make_examples.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 /usr/bin/python3 /opt/deepvariant/bin/make_examples.zip "$@"' > \
/opt/deepvariant/bin/make_examples && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand All @@ -168,7 +174,7 @@ RUN \
/opt/deepvariant/bin/runtime_by_region_vis && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'/usr/bin/python3 /opt/deepvariant/bin/multisample_make_examples.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 /usr/bin/python3 /opt/deepvariant/bin/multisample_make_examples.zip "$@"' > \
/opt/deepvariant/bin/multisample_make_examples && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand All @@ -180,7 +186,7 @@ RUN \
/opt/deepvariant/bin/convert_to_saved_model && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'/usr/bin/python3 -u /opt/deepvariant/bin/make_examples_somatic.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 /usr/bin/python3 -u /opt/deepvariant/bin/make_examples_somatic.zip "$@"' > \
/opt/deepvariant/bin/make_examples_somatic && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand Down
10 changes: 8 additions & 2 deletions Dockerfile.deepsomatic
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,7 +130,7 @@ ENV DV_BIN_PATH=/opt/deepvariant/bin

# Install libraries
RUN apt-get -y update && \
apt-get install -y parallel python3-pip unzip && \
apt-get install -y parallel python3-pip unzip libjemalloc2 && \
PATH="${HOME}/.local/bin:$PATH" python3 -m pip install absl-py==0.13.0 && \
apt-get clean autoclean && \
apt-get autoremove -y --purge && \
Expand Down Expand Up @@ -164,11 +164,17 @@ COPY --from=builder \
/opt/deepvariant/bin/

# Create shell wrappers for python zip files for easier use.
#
# The make_examples family is wrapped with LD_PRELOAD=libjemalloc.so.2: jemalloc
# meaningfully reduces make_examples wall-clock (its pileup/realignment work is
# allocation-heavy) while having no measurable effect on call_variants (TF
# inference), so the preload is scoped to just those wrappers. The bare soname
# keeps it architecture-portable (resolved from the default linker search path).
RUN \
BASH_HEADER='#!/bin/bash' && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'python3 -u /opt/deepvariant/bin/make_examples_somatic.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 python3 -u /opt/deepvariant/bin/make_examples_somatic.zip "$@"' > \
/opt/deepvariant/bin/make_examples_somatic && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand Down
10 changes: 8 additions & 2 deletions Dockerfile.deeptrio
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,11 +72,17 @@ RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTH
update-alternatives --install /usr/bin/python python /usr/bin/python${PYTHON_VERSION} 0

# Create shell wrappers for python zip files for easier use.
#
# The make_examples family is wrapped with LD_PRELOAD=libjemalloc.so.2: jemalloc
# meaningfully reduces make_examples wall-clock (its pileup/realignment work is
# allocation-heavy) while having no measurable effect on call_variants (TF
# inference), so the preload is scoped to just those wrappers. The bare soname
# keeps it architecture-portable (resolved from the default linker search path).
RUN \
BASH_HEADER='#!/bin/bash' && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'python3 /opt/deepvariant/bin/deeptrio/make_examples.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 python3 /opt/deepvariant/bin/deeptrio/make_examples.zip "$@"' > \
/opt/deepvariant/bin/deeptrio/make_examples && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand Down Expand Up @@ -215,7 +221,7 @@ ENV PATH="${PATH}":/opt/conda/bin:/opt/conda/envs/bio/bin:/opt/deepvariant/bin/d
ENV TF_USE_LEGACY_KERAS=1

RUN apt-get -y update && \
apt-get install -y parallel python3-pip && \
apt-get install -y parallel python3-pip libjemalloc2 && \
PATH="${HOME}/.local/bin:$PATH" python3 -m pip install absl-py==0.13.0 && \
apt-get clean autoclean && \
apt-get autoremove -y --purge && \
Expand Down
10 changes: 8 additions & 2 deletions Dockerfile.pangenome_aware_deepvariant
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,11 +73,17 @@ RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTH
update-alternatives --install /usr/bin/python python /usr/bin/python${PYTHON_VERSION} 0

# Create shell wrappers for python zip files for easier use.
#
# The make_examples family is wrapped with LD_PRELOAD=libjemalloc.so.2: jemalloc
# meaningfully reduces make_examples wall-clock (its pileup/realignment work is
# allocation-heavy) while having no measurable effect on call_variants (TF
# inference), so the preload is scoped to just those wrappers. The bare soname
# keeps it architecture-portable (resolved from the default linker search path).
RUN \
BASH_HEADER='#!/bin/bash' && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
'python3 -u /opt/deepvariant/bin/make_examples_pangenome_aware_dv.zip "$@"' > \
'LD_PRELOAD=libjemalloc.so.2 python3 -u /opt/deepvariant/bin/make_examples_pangenome_aware_dv.zip "$@"' > \
/opt/deepvariant/bin/make_examples_pangenome_aware_dv && \
printf "%s\n%s\n" \
"${BASH_HEADER}" \
Expand Down Expand Up @@ -140,7 +146,7 @@ ENV PATH="${PATH}":/opt/conda/bin:/opt/conda/envs/bio/bin:/opt/deepvariant/bin/p
ENV TF_USE_LEGACY_KERAS=1

RUN apt-get -y update && \
apt-get install -y parallel python3-pip && \
apt-get install -y parallel python3-pip libjemalloc2 && \
PATH="${HOME}/.local/bin:$PATH" python3 -m pip install absl-py==0.13.0 && \
apt-get clean autoclean && \
apt-get autoremove -y --purge && \
Expand Down