FIMSMemoryBenchmarks

A repository to benchmark and compare the memory footprint of NOAA-FIMS/FIMS builds.

Repository Layout

• /R: Helper R scripts for setup and benchmark stage execution.
• /scripts: Shell runners for memory profiling tools.
• /outputs: Benchmark and profiler outputs (.gitkeep included).
• /.devcontainer: Codespaces setup for debug-safe R compilation flags.

Debug Build Configuration in Codespaces

This repository includes .devcontainer/postCreate.sh, which configures ~/.R/Makevars with:

• PKG_CXXFLAGS += -g -O0 -fno-omit-frame-pointer -fvisibility=default
• PKG_STRIP = true

These settings preserve symbols and frame pointers for profiler-friendly builds.

Install FIMS in Debug Mode

R/setup_FIMS provides:

install_fims_debug(ref = "main")

It installs NOAA-FIMS/FIMS from GitHub for a chosen branch/tag/commit using source compilation.

Benchmark Stages

R/run_benchmark.R is structured into five stages:

1. Static model construction through MakeADFun()
2. Single evaluation calls (obj$fn() and obj$gr())
3. Full optimization run with nlminb (without sdreport)
4. Full optimization run with nlminb and sdreport
5. Cleanup and retention check via TMB::FreeADFun(obj) and gc()

Before sourcing R/run_benchmark.R, define fims_stage1_builder() so it returns the stage-1 MakeADFun() object.

Run Massif Benchmarks

bash scripts/run_massif.sh

This generates:

• outputs/massif_main.out
• outputs/massif_xptr.out

Use ms_print to inspect each output file.