My nerd snipe for the morning: "A High-Level Language Benchmark", p. 180, comparison of the results on p.194 with a few random compilers on the system I'm using

My nerd snipe for the morning: "A High-Level Language Benchmark", p. 180, comparison of the results on p.194 with a few random compilers on the system I'm using today. All results from my 2013 Mac Pro (12 core), compiled with -O3, "total size" binaries stripped, execution times averaged over 100 runs. Changes to BYTE C source: add return type / value to main. Changes to BYTE Fortran source: fix lower bounds on DO loops (FLAGS(0) is invalid), change LOGICAL comparison from .EQ. to .EQV. to silence compiler warning, change WRITE channel to stdout. clang (Xcode 14.2) Compiled Bytes: 1,096 Total Bytes: 49,496 Compile and Load: 0.392 s Execute: 0.210 s Ratio to PL/I-80: 0.015 1/(Ratio to PL/I-80): 66.67 gcc 12.2 (MacPorts) Compiled Bytes: 1,224 Total Bytes: 49,496 Compile and Load: 0.247 s Execute: 0.203 s Ratio to PL/I-80: 0.0145 1/(Ratio to PL/I-80): 68.97 icc 2021.9.0 20230302 Compiled Bytes: 1,582 Total Bytes: 99,664 Compile and Load: 0.675 s Execute: 0.334 s Ratio to PL/I-80: 0.0239 1/(Ratio to PL/I-80): 41.92 gfortran 12.2 (MacPorts) Compiled Bytes: 1,968 Total Bytes: 49,832 Compile and Load: 0.239 s Execute: 0.00832 s Ratio to PL/I-80: 0.000594 1/(Ratio to PL/I-80): 1,682.69 ifort 2021.9.0 20230302 Compiled Bytes: 2,467 Total Bytes: 801,192 Compile and Load: 0.617 s Execute: 0.00667 s Ratio to PL/I-80: 0.000476 1/(Ratio to PL/I-80): 2,098.95 Tentative conclusion: At least on x64 macOS, we should all be writing Fortran 77, compiled with GNU Fortran where code (esp. runtime) size is critical, Intel Fortran otherwise. YMMV.