🚀 go-pugleaf

jacobsd@prism.CS.ORST.EDU (Dana Jacobsen) writes:
> In <MIB.92Feb24165350@geech.gnu.ai.mit.edu> mib@geech.gnu.ai.mit.edu (Michael I Bushnell) writes:

> >From the release announcement by rms:
> > Early testing of GCC Version 2.0 indicates that it produces faster
> > code for SPARC computers than Sun's latest released compilers (both
> > bundled and unbundled).  It is also the fastest known compiler for
> > the Motorola 88k.

> >I don't know in what the "early testing" consists.

>   I will get the new version soon, but I ran some initial tests on the Jan
> 7 version, and it is faster than Sun's bundled compiler, but not the 1.0
> unbundled one.

The released version of gcc2 doesn't seem to do a very good job of scheduling
compared to the sun compiler that comes with sunos (4.1.1).

E.g., 

float dotprod(vec1,vec2,length)
float *vec1, *vec2;
int length;
{
    int i;
    float sum=0;

    for(i=0; i<length; i++)
	sum+=vec1[i]*vec2[i];

    return sum;
}

Here's what the unrolled part of the inner loops look like:

"gcc2 -O2 -funroll-loops"	"cc -O4 -fsingle"                
				                                 
L5:				L77003:                          
	ld [%o0],%f2		        ld      [%o0],%f1        
	ld [%o1],%f1		        ld      [%o5],%f2        
	fmuls %f2,%f1,%f1	        fmuls   %f1,%f2,%f1      
	fadds %f0,%f1,%f0	        ld      [%o0+4],%f4      
	ld [%o0+4],%f2		        fadds   %f30,%f1,%f30    
	ld [%o1+4],%f1		        ld      [%o5+4],%f5      
	fmuls %f2,%f1,%f1	        fmuls   %f4,%f5,%f4      
	fadds %f0,%f1,%f0	        fadds   %f30,%f4,%f30    
	ld [%o0+8],%f2		        ld      [%o0+8],%f7      
	ld [%o1+8],%f1		        ld      [%o5+8],%f8      
	fmuls %f2,%f1,%f1	        fmuls   %f7,%f8,%f7      
	fadds %f0,%f1,%f0	        ld      [%o0+12],%f10    
	ld [%o0+12],%f2		        fadds   %f30,%f7,%f30    
	ld [%o1+12],%f1		        ld      [%o5+12],%f11    
	fmuls %f2,%f1,%f1	        fmuls   %f10,%f11,%f10   
	add %g2,16,%g2		        inc     16,%o5           
	cmp %g2,%o3		        add     %o5,12,%g2       
	add %o0,16,%o0		        fadds   %f30,%f10,%f30   
	add %o1,16,%o1		        cmp     %g2,%o1          
	bl L5			        blu     L77003           
	fadds %f0,%f1,%f0	        inc     16,%o0           

The same instructions, but sun's compiler spreads them more advantageously;
anyway, the resulting program (that just repeatedly calls dotprod) runs
almost twice as fast when compiled with the sun compiler...

[Where should messages like this be posted?  Should this have gone to gnu.gcc.bugs?]

-Miles
--
--
Miles Bader  --  HCRC, University of Edinburgh  --  miles@cogsci.ed.ac.uk
Ich bin ein Virus. Mach' mit und kopiere mich in Deine .signature.
97% of everything is grunge