Re: FSF-GCC-2.7.2 vs EGCS-GCC-971008 on Ultra (more info)

[Oleg Krivosheev <kriol at fnal dot gov>]

On Tue, 14 Oct 1997, Fred Richardson wrote:

> Hi-
> 
> Here's a brief update on my previous email.
> 
> I've built and installed EGCS-971008 with the following flags:
>     --enable-haifa=yes --with-gnu-as --with-gnu-ld --with-stabs
> on a Model 170 UltraSPARC running Solaris 2.5
> 
> I compiled a program with the following flags:
> 
>         -mcpu=ultrasparc -funsigned-char -fno-exceptions -O6
>             -finline-functions -ffast-math -fomit-frame-pointer
>                 -funroll-loops  
> 
> The speed of the program (which does a lot of floating point) was
> about the same as when I compiling it with FSF-GCC-2.7.2 and these
> flags: 
> 
>         -O3 -finline-functions -ffast-math -fomit-frame-pointer
>             -funroll-loops 
> 
> I haven't re-run this test with the SUNW cc yet.  I don't know if this
k
just for you information, i've tried MDBENCH - molecular
dynamic FP intensive Fortran code:

On 200mhz Ultra-II box timings were:

- 16.5sec native sun f77 v4.2
- 17.6sec egcs g77 with haifa and max.optimization (f2c+egcs-gcc was the
same )
- 19.8sec f2c+gcc-2.7.2

regards

OK