Remove unused oldisim workloads for FeedSim

[lukalt]

This PR depends on #322 which needs to be merged first.

The feedsim code base has three different workloads (simple, search, and ranking) but is only using the ranking workload.
The ranking workload depends on icachebuster and pointerchase components of the search workloads.
This PR removes the simple and search workloads and moves the icachebuster and pointerchase components into ranking to improve code readability and build time.

[meta-codesync]

@charles-typ has imported this pull request. If you are a Meta employee, you can view this in D88255996.

[excelle08]

Hi @lukalt, the CI for feedsim installation on CentOS 9 aarch64 (the job) is failing due to linking error:

/usr/bin/ld: libfolly.a(Futex.cpp.o): undefined reference to symbol '_ZNSt18condition_variable10notify_oneEv@@GLIBCXX_3.4.11'
/usr/bin/ld: /github/home/miniconda/lib/libstdc++.so.6: error adding symbols: DSO missing from command line

Could you look into and fix it?

[lukalt]

Hi @lukalt, the CI for feedsim installation on CentOS 9 aarch64 (the job) is failing due to linking error:

/usr/bin/ld: libfolly.a(Futex.cpp.o): undefined reference to symbol '_ZNSt18condition_variable10notify_oneEv@@GLIBCXX_3.4.11'
/usr/bin/ld: /github/home/miniconda/lib/libstdc++.so.6: error adding symbols: DSO missing from command line

Could you look into and fix it?

Yes, I will have a look at it. I've only tested it on Ubuntu 22.04 so far where the issues does not occur

[lukalt]

Hi @lukalt, the CI for feedsim installation on CentOS 9 aarch64 (the job) is failing due to linking error:

/usr/bin/ld: libfolly.a(Futex.cpp.o): undefined reference to symbol '_ZNSt18condition_variable10notify_oneEv@@GLIBCXX_3.4.11'
/usr/bin/ld: /github/home/miniconda/lib/libstdc++.so.6: error adding symbols: DSO missing from command line

Could you look into and fix it?

Yes, I will have a look at it. I've only tested it on Ubuntu 22.04 so far where the issues does not occur

Hi @excelle08

I tried to reproduce this on a fresh Ubuntu 24.04 container with the following steps.

apt update
apt install -y python3 python3-pip python3-venv git dmidecode wget lshw bc vim sudo numactl gcc-13 g++-13
git clone -b lalt-feedsim-cleanup https://github.com/lukalt/DCPerf.git
cd DCPerf/
./packages/feedsim/install_feedsim.sh

This builds without any issues. I think the issue is related to how FeedSim is built by the GitHub pipeline

[excelle08]

Hi @excelle08

I tried to reproduce this on a fresh Ubuntu 24.04 container with the following steps.

apt update
apt install -y python3 python3-pip python3-venv git dmidecode wget lshw bc vim sudo numactl gcc-13 g++-13
git clone -b lalt-feedsim-cleanup https://github.com/facebookresearch/DCPerf.git
cd DCPerf/
./packages/feedsim/install_feedsim.sh

This builds without any issues. I think the issue is related to how FeedSim is built by the GitHub pipeline

Hi @lukalt, the failed CI job actually runs CentOS 9 in the container, have you tried running the installation on a CentOS server?

[lukalt]

./packages/feedsim/install_feedsim.sh

Thanks for the pointer. I think the issue now occurs on both operating system. However, I was not able to reproduce it on the same centos9 base container and following these steps. It installs successfully.

    1  export TZ=Etc/UTC
    2  dnf --enablerepo=* clean all && dnf config-manager --set-enabled crb --save &&         dnf install -y dnf-plugins-core &&         dnf install -y epel-release
    3  yum install -y python3 python3-pip git dmidecode wget lshw bc vim make cmake gcc g++ xz procps bzip2 numactl sudo sysstat
    4  git clone -b lalt-feedsim-cleanup https://github.com/lukalt/DCPerf.git
    5  cd DCPerf/
    6  ./packages/feedsim/install_feedsim.sh

I don't understand why the changes introduces in the PR cause this issue. The error occurs during the folly installation which has not been touched. Maybe the issue is caused by the parallel build in combination with the conda environment used by the pipeline