A core design goal of the SML design effort from its inception was supporting efficient code generation by maintaining a clear formal phase distinction between compile-time and run-time semantics.
Turning that intention into actual high-performance compilers required solving a number of new problems, which ultimately took a decade and a half, but today state of the art SML compilers such as MLton produce code fully competitive with C++ and Java.
Mythryl inherits from SML/NJ the fruits of this research. Depending upon the specific synthetic benchmark, it generates code that will clock in at anywhere from twice as fast to half as fast as a C compiler — and roughly one hundred times faster than scripting languages like Perl, Python or Ruby.
Despite this, due to being an incremental compiler that generates optimized code directly in memory (as opposed to a batch-mode disk-to-disk compiler like gcc), Mythryl offers much of the convenience of scripting languages: Short Mythryl programs may be run simply by putting a #!/usr/bin/mythryl “shebang line” at the top, and small fragments of code may be compiled and executed directly in-process.