Re: Poor performance of implied-DO loop with Intel Fortran

From: glen herrmannsfeldt (gah_at_ugcs.caltech.edu)
Date: 05/05/04 

Date: Wed, 05 May 2004 21:21:40 GMT

Aidan wrote:
> glen herrmannsfeldt <gah@ugcs.caltech.edu> wrote:

(snip)

>>Could you compile this with the compiler option that shows
>>the generated assembly code, and post that? It would help
>>to see what the compiler is actually doing.

> You asked for it .. though I am not saying I need this solved -- I
> was just interested to see if this was a widespread problem that I
> should look out for. I have decided it isn't.

It is sometimes useful to know what compilers will generate for
different input.

Does anyone know what f90_allocate3 and f90_allocate1 leave on
the stack? The code isn't too hard to follow, except when
called routines change the stack pointer so that all the stack
offsets are different.

In following it so far, I haven't found that it actually allocates
1000000 elements, so I presume that I followed it wrong.
Knowing the arguments to f90_allocate3 and f90_allocate1,
and for that matter, f90_reallocate1, f90_deallocate1, and
f90_deallocate13, would be nice, too.

-- glen

Relevant Pages

  • Re: Question about function call in a microprocessor?
    ... > or subroutine call in a assembly code. ... > about the stack here. ... > foo uses lets says 5 local varaibles. ... compiler and look at the output listing. ...
    (sci.electronics.design)
  • Re: Ancient history [was Re: Public disclosure ...]
    ... > assembler that withdrawing the assembler wouldn't stop me from ... DIMENSION statements in subroutines, assuming the compiler would ... of inline assembly code when rules didn't allow assembly ...
    (sci.crypt)
  • Re: Just a bit of silliness
    ... snip ... ... recognised as such by the compiler. ... printf for example initially ... Combined with a mechanism to access the stack ...
    (comp.lang.c)
  • Re: what about the option -time for g95?
    ... execution of each subprocess". ... f951 which is the compiler, which generates assembly code, and ...
    (comp.lang.fortran)
  • Re: made it to page 4 of gforth tutorial
    ... the stack space anyway. ... And there are clearly going to be some platform ... came with an embedded C compiler once. ... like I've written before, if one *really* needs recursion (as I have, ...
    (comp.lang.forth)