#tinycc based #RStats #C #Transpiler, only 4x slower than #{quickr} for the convolution benchmark ! 2-3x faster than naive #C + marshaling. #tinycc transpilation will have an obvious advantage since it can always fallback when transpiling and use `Rf_lang3` since it has access to R's C runtime !

RE: https://bsky.app/profile/did:plc:hmk4f5yf3ohtsvw32sbtwvk3/post/3memwdbpbqs22

#RStats the comparison results with #{quickr} is expected since tinycc is not an optimizing compiler like gfortran/flang, but still since you don t need an additional compiler at runtime and you can iterate (more segfault :D) very fast ! (no windows support for now/never)

RE: https://bsky.app/profile/did:plc:hmk4f5yf3ohtsvw32sbtwvk3/post/3mec4fdma6k2s