I'm developing a new programming language (rylang.net) and I am looking for some suggestions for problems to benchmark in the new programming language. This language has 3 main goals, in increasing order of importance:
1. Develop faster than C++, by being context free and simplify the syntax where it can be done without causing problems with the other 2 goals,
2. Drastically improve compile and link times v.s. C++, and
3. Improve execution speed and efficiency v.s. C++,
With that in mind, there are a few different cases I am interested in:
1. Standard library benchmark suggestions. e.g. "Insert X random entries into a map"
2. Benchmarks for problems that C++ has issues with, where you might turn to assembly instead. For example, if you find that Clang++ and GCC generate sub-optimal assembly when multiplying two matrices even with -O2. Super helpful if you point out what the compiler is doing wrong.
3. Benchmarks for problems you think that might show the language is fit as a high performance language. Example: "Serve X concurrent https requests".
4. Areas where C++ has performance portability issues. E.g. "I can't write a fast matrix multiply that is fast on both ARM and x64 without doing #ifdef __aarch64__ ... etc..
5. Any ideas for performance improvement that requires language frontend support (example, allowing the programmer to assign pointers and references to user-defined alias groups, as opposed to only using alias groups that result from compiler analysis)
1. Develop faster than C++, by being context free and simplify the syntax where it can be done without causing problems with the other 2 goals,
2. Drastically improve compile and link times v.s. C++, and
3. Improve execution speed and efficiency v.s. C++,
With that in mind, there are a few different cases I am interested in:
1. Standard library benchmark suggestions. e.g. "Insert X random entries into a map"
2. Benchmarks for problems that C++ has issues with, where you might turn to assembly instead. For example, if you find that Clang++ and GCC generate sub-optimal assembly when multiplying two matrices even with -O2. Super helpful if you point out what the compiler is doing wrong.
3. Benchmarks for problems you think that might show the language is fit as a high performance language. Example: "Serve X concurrent https requests".
4. Areas where C++ has performance portability issues. E.g. "I can't write a fast matrix multiply that is fast on both ARM and x64 without doing #ifdef __aarch64__ ... etc..
5. Any ideas for performance improvement that requires language frontend support (example, allowing the programmer to assign pointers and references to user-defined alias groups, as opposed to only using alias groups that result from compiler analysis)