Parallel Compiling

FoBiS.py can compile independent source files concurrently, giving a significant speedup on multi-core machines.

The -j flag

fobis build -j 4

The -j N (or --jobs N) flag enables a pool of N concurrent compile jobs. Files that have no dependency on each other are compiled simultaneously; files that depend on others are compiled after their dependencies are ready.

The pool size should roughly match the number of physical CPU cores available:

# 8-core machine
fobis build -j 8

# Use all available cores (Linux/macOS)
fobis build -j $(nproc)

fobos option

[release]
compiler = gnu
cflags   = -c -O3
jobs     = 8

How it works

FoBiS.py resolves the full dependency graph before starting compilation. Only sources that are ready (all their USEd modules already compiled) are submitted to the pool. This ensures correct ordering while maximising concurrency for independent files.

Note

The pool is not dynamically balanced based on file count — it simply submits all ready files at once. For very uneven dependency trees the speedup is proportional to the width of the graph.

Example

# Debug build, single job (default)
fobis build -compiler gnu -cflags "-c -O0 -g"

# Release build, parallel
fobis build -compiler gnu -cflags "-c -O3" -j 8

With a fobos file:

[modes]
modes = debug release

[debug]
compiler = gnu
cflags   = -c -O0 -g
jobs     = 1

[release]
compiler = gnu
cflags   = -c -O3
jobs     = 8

fobis build -mode debug
fobis build -mode release