Re: Fortran array indexing on 64-bit targets & snapshot 971023

[Toon Moene <toon@moene.indiv.nluug.nl>]

  In message < 9710241223.AA02380@moene.indiv.nluug.nl > I wrote:

  > It seems that Richard Henderson's `expr.c' (get_inner_reference)  
  > patch didn't make into snapshot 971023.
  >
  > Was there a compelling reason to exclude it ?

Jeff:

> Jim had some questions about it, which nobody has answered yet:

[ Complicated discussion about the special handling of zero lower  
bound array references elided ]

Yes, I've looked into this and didn't understand it either.

Nevertheless, to inject some numbers into this discussion (using  
Patick Queutey's CFD benchmark example on a 500 Mhz Alpha):

egcs-971023, bare (3 runs):

7.60user 0.02system 0:20.87elapsed 36%CPU (0avgtext+0avgdata  
0maxresident)k
7.25user 0.01system 0:11.27elapsed 64%CPU (0avgtext+0avgdata  
0maxresident)k
7.78user 0.02system 0:11.18elapsed 69%CPU (0avgtext+0avgdata  
0maxresident)k

After applying Richard Henderson's `exp.c' patch (4 runs):

4.51user 0.14system 0:09.17elapsed 50%CPU (0avgtext+0avgdata  
0maxresident)k
4.53user 0.01system 0:08.20elapsed 55%CPU (0avgtext+0avgdata  
0maxresident)k
4.15user 0.02system 0:07.72elapsed 54%CPU (0avgtext+0avgdata  
0maxresident)k
4.37user 0.01system 0:07.29elapsed 60%CPU (0avgtext+0avgdata  
0maxresident)k

(with your `use' patch I get between 3.6 and 3.9 seconds).

So I hope a decision will be reached (to say the least :-)

Cheers,
Toon.