[Ping] gcc+gcj build-times? QUESTION.3

[Norman Hendrich <hendrich at informatik dot uni-hamburg dot de>]

Hello all,

I am just too impatient to wait a whole week before pinging, sorry ;-)


So far, I got 10+ answers to my original questions 1) and 2), and I now
understand the reasons for the much longer build-time of gcc-4.2.0 vs.
gcc-4.0.x. No problem, I can always use a machine with lots of memory
for buiding gcc.


However, I got no answer to my question 3) and the related "toy benchmark"
post (http://gcc.gnu.org/ml/java/2006-02/msg00194.html).

Could anyone with a known working installation of gcj please verify
that about "three frames per second" really is the best I should expect 
from the animation, even when compiling with -O3?  After all, I only 
built gcj because jamvm was running the program so slowly.

Note that the speed of the matrix calculations seems pretty competitive,
but the animation runs _slower_ under gcj than jamvm+classpath. Also
note that the animation just calls Graphics.fillRect().

libjavaclasspath/native/jni/gtk-peer/gnu_java_awt_peer_gtk_GdkGraphics.c:

JNIEXPORT void JNICALL
Java_gnu_java_awt_peer_gtk_GdkGraphics_fillRect
  (JNIEnv *env, jobject obj, jint x, jint y, jint width, jint height)
{
  struct graphics *g = NULL;

  gdk_threads_enter ();

  g = (struct graphics *) NSA_GET_G_PTR (env, obj);

  gdk_draw_rectangle (g->drawable, g->gc, TRUE,
                      x + g->x_offset, y + g->y_offset, width, height);
  gdk_flush ();

  gdk_threads_leave ();
}




Thanks in advance,
  Norman
  
  

PS:  Alexandre Olvia writes: 
> I remember running full bootstraps of
> GCC was the reason why I got my Athlon64 notebook upgraded from 512MB
> to 2G of RAM (well over the alleged maximum of 1280MB specified in the
> manual :-) about half a year ago, as soon as the 1-year warranty
> expired. 

Well, I fear most people will rather not build libjava at all, instead
of attempting "brain surgery" on their machines :-)