Programming Approaches
“Automatic Parallelization” of Fortran
- Preserves the investment in the “legacy codes”
- Replaces programmer effort with compiler effort
- Should be as successful as vectorization
- Has been demonstrated to achieve 0.9-10 fold speedup, with 2.5 being typical -- Amdahl’s Law
Alternative Languages
- Functional -- SISAL, Id Nouveau, etc.
- Logic -- Prolog
These approaches failed ... the only successful programmers have programmed close to their machine; vectorized Cray Fortran continued to reign