Read the article, 1.2 was 40% as fast on one specific benchmark, and its lack of speed largely came down to having a much worse CRC implementation, not general shortcomings.
Now 1.3 is 63% the speed of gcc O2 on this largely hostile benchmark.
Whoops somehow I missed the 63%. Either way, it's still too slow to be considered production ready. Hare is the only real example of a QBE based language but why put up with something that's C-like with manual memory management but slower than a bunch of managed languages? If I'm going to have to write my own allocations and deal with a static language, it needs to be within spitting distance of C. Not slower than Java, Go or Common Lisp.
It's a cool hobby project though, can't take that away from them. On a technical basis, 63% of -O2 with the amount of LOC written is impressive.
More like no inlining, so-so register allocation, and a simple code generator. I believe also no loop unrolling. Other compilers could handle the table-less CRC implementation.
Now 1.3 is 63% the speed of gcc O2 on this largely hostile benchmark.