Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Writing a compiler in Python using Lex, Yacc and LLVM

wiki.alcidesfonseca.com

Post reply on HN
Print view View original

31–40 of 44 posts

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#31

Post by pc »

ilkhd2 wrote: post #29
mahmud wrote: post #24

Earlier quoted context omitted.

Compiler written in python, ruby and other "slow" .. languages going to take eternity to compile linux kernel. What makes you think everyone is out to write a C compiler?

What make you think that I think this way? Linux kernel is just an example of relatively big program, where performance of the compiler is an important issue. Another example of performance-sensitive application of compilers is background Java compiler in Eclipse - even for rather small programs it has to be fast. Ruby is 10-100 (actually more than that) times slower than C/Java, So if you optimizer (wriiten in C/Jav…

Have you actually used Stalin? Compiling anything with it is shockingly slow. It does, of course, produce very fast programs—but it can take minutes to generate them, even in simple cases.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#32

Post by ilkhd2 »

pc wrote: post #31
ilkhd2 wrote: post #29

Earlier quoted context omitted.

What make you think that I think this way? Linux kernel is just an example of relatively big program, where performance of the compiler is an important issue. Another example of performance-sensitive application of compilers is background Java compiler in Eclipse - even for rather small programs it has to be fast. Ruby is 10-100 (actually more than that) times slower than C/Java, So if you optimizer (wriiten in C/Jav…

Have you actually used Stalin? Compiling anything with it is shockingly slow. It does, of course, produce very fast programs—but it can take minutes to generate them, even in simple cases.

No I did not, butI read papers, written by the author. Indeed it is slow, but still be good for final compilation. Alas it is r4rs, and also I heard that it sometimes generate incorrect code. Still Scheme is better choice in my opinion for performance sensitive code, than ruby/python.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#33

Post by ilkhd2 »

pc wrote: post #31
ilkhd2 wrote: post #29

Earlier quoted context omitted.

What make you think that I think this way? Linux kernel is just an example of relatively big program, where performance of the compiler is an important issue. Another example of performance-sensitive application of compilers is background Java compiler in Eclipse - even for rather small programs it has to be fast. Ruby is 10-100 (actually more than that) times slower than C/Java, So if you optimizer (wriiten in C/Jav…

Have you actually used Stalin? Compiling anything with it is shockingly slow. It does, of course, produce very fast programs—but it can take minutes to generate them, even in simple cases.

You probably missed the point - my concern is not the speed of compilation _of_ _the_ _compiler_ _itself_ - it is totally irrelevant, what is important is the performance of _created_ _compiler_, which is going to be ok, taking into account the quality of Stalins optimizer.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#34

Post by judofyr »

daeken wrote: post #5
ilkhd2 wrote: post #2

Impractical: 1) The compiler is gonna be very slow; 2) If you make compiler for low-level language, such as C you need precise correspondence between data types used in cpu, datatypes used in compiler's implementation language and datatypes in the target language.

On the contrary. Compilers written in a high-level language are at a distinct advantage: they're easier to optimize (significantly so), easier to debug, and support forms that are difficult to deal with in a low(er)-level language like C or C++. I do compiler development every day, and I no longer touch low-level compilers. Everything I write is Boo (for my OS), Ruby (for my startup -- by far my favorite language for…

May I ask you what tools you use in Ruby?

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#35

Post by mahmud »

ilkhd2 wrote: post #33
pc wrote: post #31

Earlier quoted context omitted.

Have you actually used Stalin? Compiling anything with it is shockingly slow. It does, of course, produce very fast programs—but it can take minutes to generate them, even in simple cases.

You probably missed the point - my concern is not the speed of compilation _of_ _the_ _compiler_ _itself_ - it is totally irrelevant, what is important is the performance of _created_ _compiler_, which is going to be ok, taking into account the quality of Stalins optimizer.

We can debate the performance of a hypothetical compiler for eternity, or we can cite examples; slow compilers written in a non-C language along with publicly available input data.

i.e. benchmarks or it didn't happen.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#36

Post by mahmud »

ilkhd2 wrote: post #32
pc wrote: post #31

Earlier quoted context omitted.

Have you actually used Stalin? Compiling anything with it is shockingly slow. It does, of course, produce very fast programs—but it can take minutes to generate them, even in simple cases.

No I did not, butI read papers, written by the author. Indeed it is slow, but still be good for final compilation. Alas it is r4rs, and also I heard that it sometimes generate incorrect code. Still Scheme is better choice in my opinion for performance sensitive code, than ruby/python.

Stalin and Mlton are whole-program optimizing compilers. They take a bit longer than other compilers, but the results are impressive.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#37

Post by dkersten »

ilkhd2 wrote: post #29
mahmud wrote: post #24

Earlier quoted context omitted.

Compiler written in python, ruby and other "slow" .. languages going to take eternity to compile linux kernel. What makes you think everyone is out to write a C compiler?

What make you think that I think this way? Linux kernel is just an example of relatively big program, where performance of the compiler is an important issue. Another example of performance-sensitive application of compilers is background Java compiler in Eclipse - even for rather small programs it has to be fast. Ruby is 10-100 (actually more than that) times slower than C/Java, So if you optimizer (wriiten in C/Jav…

Background compilers do not need to perform heavy optimiztion, so can skip a lot of time consuming steps.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#39

Post by ilkhd2 »

mahmud wrote: post #35
ilkhd2 wrote: post #33

Earlier quoted context omitted.

You probably missed the point - my concern is not the speed of compilation _of_ _the_ _compiler_ _itself_ - it is totally irrelevant, what is important is the performance of _created_ _compiler_, which is going to be ok, taking into account the quality of Stalins optimizer.

We can debate the performance of a hypothetical compiler for eternity, or we can cite examples; slow compilers written in a non-C language along with publicly available input data. i.e. benchmarks or it didn't happen.

Man, you are being difficult. Ruby is well known to be a very slow language (if you want numbers check http://shootout.alioth.debian.org/u32q/benchmark.php?test=al...). Optimizing compilers are computationally demanding tasks, and the proof (albeit extreme one) is Stalin Scheme compiler, that can take minutes to compile code, but the result is sometimes faster than C/C++ code, or at least on par. Now, this compiler is written either in Scheme itself or in C i do not remember, but definetelt not in ruby. If it take to compile your code a minute for a compiler written in [C/Stalin Scheme] it will take at least ten minutes for one written in ruby.

Re: Writing a compiler in Python using Lex, Yacc and LLVM

#40

Post by ilkhd2 »

daeken wrote: post #5
ilkhd2 wrote: post #2

Impractical: 1) The compiler is gonna be very slow; 2) If you make compiler for low-level language, such as C you need precise correspondence between data types used in cpu, datatypes used in compiler's implementation language and datatypes in the target language.

On the contrary. Compilers written in a high-level language are at a distinct advantage: they're easier to optimize (significantly so), easier to debug, and support forms that are difficult to deal with in a low(er)-level language like C or C++. I do compiler development every day, and I no longer touch low-level compilers. Everything I write is Boo (for my OS), Ruby (for my startup -- by far my favorite language for…

Ok, the discussion is becoming pointless: you like ruby for compiler construction. Well I'd rather use a language with proper tail recursion, algebraic data types, with pattern matching as a part of it, fast, strongly typed [I beleive algebraic types with pat. matching gives roughly same possibilities as does dyn. typing, for this task], with standard or at least big industrial backing behind - such as F# or Haskell. And at the end of the day get warm feeling that my code is "green" - does not waste joules, by working fast.

I appreciate your choice, but I'd myself do differently.

Post reply on HN