Skip to main content

FORTRAN programs

Course and self-directed FORTRAN work for numerical methods, linear algebra, and small scientific simulations of the kind you see in engineering and physics curricula.

Problem​

A lot of legacy and HPC work still depends on numerically careful, compiler-friendly Fortran for tight loops. The point of these pieces was to understand memory layout, array semantics, and floating-point behavior without hiding behind a higher-level framework.

Solution​

Classic algorithms (numerical integration, eigenvalue steps, or PDE discretizations) with clear module boundaries, explicit interfaces where it helped, and build steps you can repeat.

Impact​

  • Builds intuition for the performance-sensitive paths used in research and industry HPC.
  • Numeric cores stay portable enough to wrap later from Python, C, or modern Fortran.

Lessons learned​

  • Compiler flags and optimization levels can nudge answers. Document the test matrix.
  • Prefer modern Fortran (modules, intent, allocatables) over fixed-form legacy patterns when you can.
  • Pair the numerics with small golden-file checks or known reference solutions.

Tech stack​

  • FORTRAN (modern where allowed)
  • Compiler toolchain and simple Makefiles / build scripts
  • Optional: Python for plotting or wrapping drivers