Bootstrap FIMS memory profiling benchmark scaffold (R stages, Massif runner, Codespaces debug build config)

[Copilot]

This PR initializes the repository as a repeatable benchmark harness for comparing memory behavior across NOAA-FIMS/FIMS refs. It adds the baseline structure, staged R benchmark entrypoints, and profiling-oriented debug build setup for Codespaces.

• Repository scaffold

  • Added:
    • R/ for benchmark/setup helpers
    • scripts/ for profiler runners
    • outputs/.gitkeep as committed output placeholder
    • .devcontainer/ for Codespaces initialization

• FIMS debug install entrypoint

  • Added R/setup_FIMS with install_fims_debug(ref) to install NOAA-FIMS/FIMS from a branch/tag/SHA via source build (remotes::install_github).

• Staged benchmark flow (R)

  • Added R/run_benchmark.R with explicit stages:
    1. Static model construction through MakeADFun()
    2. Single eval calls (obj$fn(), obj$gr())
    3. nlminb run without sdreport
    4. nlminb run with sdreport
    5. Cleanup/retention check (TMB::FreeADFun(obj), gc())

• Branch comparison driver

  • Added R/main.R to run the same benchmark flow for:
    • main
    • xptr-refactor

• Massif execution script

  • Added scripts/run_massif.sh to run Valgrind Massif per ref and emit:
    • outputs/massif_main.out
    • outputs/massif_xptr.out

• Codespaces debug symbol preservation

  • Added .devcontainer/postCreate.sh + .devcontainer/devcontainer.json to ensure ~/.R/Makevars contains:
    • PKG_CXXFLAGS += -g -O0 -fno-omit-frame-pointer -fvisibility=default
    • PKG_STRIP = true

• Documentation

  • Rewrote README.md with setup and execution guidance for debug installs and Massif benchmark runs.

# Run memory profiling for both refs
bash scripts/run_massif.sh
# -> outputs/massif_main.out
# -> outputs/massif_xptr.out

Original prompt

Initialize a repository for memory profiling and benchmarking the NOAA-FIMS/FIMS package

Please set up the following directory structure and starter files:

1. Directory Structure:

   • R/: Helper scripts for running benchmark stages and parsing profiling output.
   • scripts/: Shell scripts to run benchmarks under Valgrind Massif or gproftools.
   • outputs/: Placeholder directory for storing .out files and summary reports (include a .gitkeep).
   • .devcontainer/ (or setup script): Pre-configure user level ~/.R/Makevars inside Codespaces with debug flags (PKG_CXXFLAGS += -g -O0 -fno-omit-frame-pointer -fvisibility=default and PKG_STRIP = true) to prevent stripping debug symbols during builds.

2. Starter Files:

   • R/run_benchmark.R: An R script structured into 3 distinct stages:
     • Stage 1: Static model construction up through MakeADFun().
     • Stage 2: Single evaluation calls (obj$fn() and obj$gr()).
     • Stage 3: a full model run with nlminb but without sdreport
     • Stage 4: a full model run with nlminb and sdreport
     • Stage 5: Memory cleanup/retention check (TMB::FreeADFun(obj) & gc()).
       -R/setup_FIMS: an R script that installs FIMS from the NOAA-FIMS/FIMS repo:

#' Install a specific FIMS branch compiled in Debug Mode
#'
#' @param ref Branch name, tag, or commit hash (e.g. "main", "xptr-refactor")
install_fims_debug <- function(ref = "main") {
message(sprintf("Installing NOAA-FIMS/FIMS@%s in debug mode...", ref))

Ensure remotes is available

if (!requireNamespace("remotes", quietly = TRUE)) {
install.packages("remotes")
}

Force compilation from source with user's ~/.R/Makevars applied

remotes::install_github(
repo = "NOAA-FIMS/FIMS",
ref = ref,
force = TRUE,
build_vignettes = FALSE,
INSTALL_opts = c("--no-multiarch")
)
}

  -`R/main.R`:

1. Benchmark Main

install_fims_debug("main")
source("R/run_benchmark.R") # Outputs to outputs/massif_main.out

2. Benchmark Feature Branch

install_fims_debug("xptr-refactor")
source("R/run_benchmark.R") # Outputs to outputs/massif_xptr.out

   - `scripts/run_massif.sh`: A shell script that executes the benchmark script using `valgrind --tool=massif` and outputs results to the `outputs/` folder.
   - `README.md`: Documenting instructions on how to install FIMS from GitHub/branches in debug mode and run the benchmark suite.

[Copilot]

Done in f8452dd. The script is now named R/install_FIMS_debug and all references were updated.

[Andrea-Havron-NOAA]

Can you squash before pushing commits?