Re: inlining inefficiencies

[Daniel Berlin <dberlin at dberlin dot org>]

On Thu, 30 May 2002, Dan Nicolaescu wrote:

> Daniel Berlin <dberlin@dberlin.org> writes:
> 
>   > On Thu, 30 May 2002, Kurt Garloff wrote:
>   > 
>   > > Hi Dan,
>   > > 
>   > > On Wed, May 29, 2002 at 10:31:46PM -0700, Dan Nicolaescu wrote:
>   > > > Andreas Jaeger <aj@suse.de> writes:
>   > > > 
>   > > >   > Can you have a look at Kurt Garloff's patches which have been
>   > > >   > discussed here some weeks agao?  A current version is at
>   > > >   > http://www.garloff.de/kurt/freesoft/gcc/
>   > > > 
>   > > > Those patches deal with deciding what functions to inline. 
>   > > > What I've shown is that the code generated after inlining all the
>   > > > functions (desired in that case) is inefficient and the optimizers
>   > > > can't deal with it.
>   > > 
>   > > The problem that I spotted in gcc is that it's unable to remove 
>   > > temporaries too often. I have some Matrix class whose [] operator
>   > > return a temporary object, basically having a pointer to the 
>   > > Matrix and the information which line was selected. This can
>   > > be used to create a Vector or an element with a second [] and
>   > > the temporary should be removed by the compiler.
>   > 
>   > Just out of curiosity, does the patch I posted yesterday to make the 
>   > inliner 
>   > generate proper statement expressions help at all?
> 
> It doesn't help the testcase I posted. The generated code is identical
> before and after your patch.

Yeah, I already tested it and found that out right before i got this.


ssa-ccp seems to help the bar function a lot, by noting that the values 
it's storing into memory are constant.

The end result is that the stores get eliminated later:

[dberlin@dberlin dberlin]$ diff -c3pwBb before.s inlineeff.s
*** before.s    2002-05-30 11:43:47.000000000 -0400
--- inlineeff.s 2002-05-30 12:26:45.000000000 -0400
*************** _Z3barv:
*** 82,99 ****
  .LFB3:
        subl    $20, %esp
  .LCFI2:
-       movl    $123, 8(%esp)
-       movl    8(%esp), %eax
-       movl    $428, 12(%esp)
-       movl    12(%esp), %edx
-       leal    (%eax,%eax), %ecx
-       movl    %ecx, (%esp)
-       leal    (%edx,%edx), %eax
-       movl    %eax, 4(%esp)
-       movl    4(%esp), %ecx
-       movl    (%esp), %eax
-       movl    %ecx, Zi+4
        movl    %eax, Zi
        addl    $20, %esp
        ret
  .LFE3:
--- 82,89 ----
  .LFB3:
        subl    $20, %esp
  .LCFI2:
        movl    %eax, Zi
+       movl    %edx, Zi+4
        addl    $20, %esp
        ret
  .LFE3:
>