6 points | by vishvananda 6 hours ago ago
2 comments
Whereas this is very impressive, I feel “faster than” should be called out as “faster to compile than”. It’s generated code seems “on a par with”.
Yes, agreed. "Faster to compile" can also be interpreted as the time to compile the compiler itself, so rather than try to disambiguate in the title, I just went with the short version.
The breakdown is roughly:
* Compiling a TU (Unoptimized): ~2.5x
* Compiling a TU (Optimized): ~4x
* Generated Code Speed (Unoptimized): ~1.01x
* Generated Code Speed (Optimized): 0.99x
Whereas this is very impressive, I feel “faster than” should be called out as “faster to compile than”. It’s generated code seems “on a par with”.
Yes, agreed. "Faster to compile" can also be interpreted as the time to compile the compiler itself, so rather than try to disambiguate in the title, I just went with the short version.
The breakdown is roughly:
* Compiling a TU (Unoptimized): ~2.5x
* Compiling a TU (Optimized): ~4x
* Generated Code Speed (Unoptimized): ~1.01x
* Generated Code Speed (Optimized): 0.99x