From 6973f48638c437b40c250286d2c8448d1c308719 Mon Sep 17 00:00:00 2001 From: Rodrigo Arias Mallo Date: Wed, 3 Mar 2021 12:33:59 +0100 Subject: [PATCH] osu: add an experiment for Intel MPI tunning --- garlic/exp/index.nix | 1 + garlic/exp/osu/impi.nix | 68 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 69 insertions(+) create mode 100644 garlic/exp/osu/impi.nix diff --git a/garlic/exp/index.nix b/garlic/exp/index.nix index 6d4c2a5..955f735 100644 --- a/garlic/exp/index.nix +++ b/garlic/exp/index.nix @@ -101,6 +101,7 @@ latencyMt = latency.override { enableMultithread = true; }; latencyMtShm = latency.override { enableMultithread = true; interNode = true; }; bw = callPackage ./osu/bw.nix { }; + impi = callPackage ./osu/impi.nix { }; bwShm = bw.override { interNode = false; }; }; } diff --git a/garlic/exp/osu/impi.nix b/garlic/exp/osu/impi.nix new file mode 100644 index 0000000..7123560 --- /dev/null +++ b/garlic/exp/osu/impi.nix @@ -0,0 +1,68 @@ +{ + stdenv +, stdexp +, bsc +, targetMachine +, stages + +# Should we test the network (true) or the shared memory (false)? +, interNode ? true +}: + +with builtins; +with stdenv.lib; + +let + + machineConfig = targetMachine.config; + + # Initial variable configuration + varConf = with bsc; { + threshold = [ 8000 16000 32000 64000 ]; + #threshold = [ 4096 8192 10240 ]; + }; + + # Generate the complete configuration for each unit + genConf = with bsc; c: targetMachine.config // rec { + inherit (machineConfig) hw; + nodes = if interNode then 2 else 1; + ntasksPerNode = if interNode then 1 else 2; + mpi = impi; + cpusPerTask = 1; + time = "00:10:00"; + qos = "debug"; + loops = 10; + expName = "osu-impi-rndv"; + unitName = expName + "-${toString threshold}"; + jobName = expName; + inherit (c) threshold; + }; + + # Compute the array of configurations + configs = stdexp.buildConfigs { + inherit varConf genConf; + }; + + exec = {nextStage, conf, ...}: stages.exec { + inherit nextStage; + env = '' + export PSM2_MQ_RNDV_SHM_THRESH=${toString conf.threshold} + export PSM2_MQ_RNDV_HFI_THRESH=${toString conf.threshold} + export PSM2_MQ_EAGER_SDMA_SZ=${toString conf.threshold} + #export PSM2_MTU=${toString conf.threshold} + export PSM2_TRACEMASK=0x101 + ''; + + program = "${nextStage}/bin/osu_bw"; + }; + + program = {nextStage, conf, ...}: bsc.osumb.override { + # Use the specified MPI implementation + inherit (conf) mpi; + }; + + pipeline = stdexp.stdPipeline ++ [ exec program ]; + +in + + stdexp.genExperiment { inherit configs pipeline; }