undefined | Better HN

0 pointssizzla15y ago0 comments

Advancing compilers is hard. When people argue efficiency as a compiler implementation detail that is going to get worked out, they forget about many who have fallen before them.

You can argue some older languages were written in a way in which it was (reasonably) easy to write a compiler that generates code with little performance overhead when compared to assembly (at worst a factor of 2 to 4, back in the 80s). Some popular languages keep adding abstractions and constructs that the compiler can actually deal with without some new compiler research breakthrough.

0 comments

2 comments · 1 top-level

neilc15y ago· 1 in thread

When people argue efficiency as a compiler implementation detail that is going to get worked out, they forget about many who have fallen before them.

Yes; see also http://prog21.dadgum.com/40.html

sizzlaOP15y ago

Thanks. Beatifully written and I love the print-shop analogy.

It's nice he is into the C&C Portland Wiki, there's golden nuggets of wisdom everywhere. I am reposting the link from his blog. http://c2.com/cgi/wiki?SufficientlySmartCompiler

Personally of the higher-level languages, I found SBCL to be crazy good at optimizing, of course given a few nudges with a compiler directive or two. By inspecting the dissasembly, I could validate it was doing the "right-thing" (TM), but then again, I do the same to check the C/C++ compiled code.

j / k navigate · click thread line to collapse