Version française
Home     About     Download     Resources     Contact us    
Browse thread
compiler bug?
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: John Carr <jfc@M...>
Subject: Re: [Caml-list] compiler bug?

On SPARC the presence of the function call in the outer loop causes
the code generated for the inner loop to change so the dummy1 variable
is stored on the stack instead of in a register.  Each loop iteration
loads dummy1, modifies it, and stores it back onto the stack.
The store-load hazard, loading a value that is in the store buffer,
adds a large delay.  The loop runs in half the time if I comment out
either the store or the load in the assembly.

If the inner loop did more computation the effect would be much less.

This is surprising but not strictly a bug.  Xavier Leroy has posted
about similar minor changes causing the compiler to box or unbox a
floating point value with major changes in performance.

>   I would like to report what I think might be a bug in the Ocaml compiler.  But first I wanted to run this by this group in case there's something I'm missing.  I have some very simple code that consists of 2 nested loops.  Inside the inner loop, is a simple statement.  Furthermore, the inner loop is not "tight".  Ie. the number of iterations within the inner loop is very large and the number of iterations of the outer loop is very small.  I then manually time this.  I then change the code by inserting a simple function call between the inner and outer loops.  This should have virtually no effect whatsoever.  However, when I time this, I get exactly twice the time.  This is somewhat inexplicable.  I tried tinkering with the "-inline" option for ocamlopt but this had no effect.  Below is the actual code (main.ml):
>    
>   let main () =
>     let dummy1 = ref 0 in
>   let dummy2 = ref 0.0 in
>     for i = 1 to 4 do
>     for j = 1 to 1000000000 do
>       dummy1 := !dummy1 + 1;
>       dummy1 := !dummy1 - 1
>     done;
>     dummy2 := Unix.gettimeofday ()
>   done
>    
>   let _ = main ()
>    
>   I compile as follows: ocamlopt unix.cmxa main.ml
> and run: ./a.out
>    
>   Is this in fact a bug of the ocamlopt compiler?  Or is there some way currently to make this effect disappear?