dtrace discuss - Sep 2007 - nop instruction generated by sun cc

Dear all:

there''s no sun studio mailling list in opensolaris.org, i raise my
problem
here :)

a simple piece of code, 2 nested while loops, and some if statements inside
each while loop.

i compiled this program in Solaris with sun cc (sun studio 12)  and gcc.
and i compiled it in Debian with gcc as well.

Then I dumped (disassembled) these 3 programs.
I found that there were alway serveral nop instructions followed with the
jmp instruction in the program generated by suncc.
and this problem didn''t exist in the other 2 program generated by gcc.

err... Why does sun cc generate code using this pattern (a jmp followed by
serveral nops)? For what?

TIA


Regards,
TJ
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20070911/9ba0ad5b/attachment.html>

???? TaoJie stated:
<    Dear all:
<     
<    there''s no sun studio mailling list in opensolaris.org, i raise
my problem
<    here :)

 Theres the tools-discuss mailing list for this.

<     
<    a simple piece of code, 2 nested while loops, and some if statements
<    inside each while loop.
<     
<    i compiled this program in Solaris with sun cc (sun studio 12)  and gcc.
<    and i compiled it in Debian with gcc as well.
<     
<    Then I dumped (disassembled) these 3 programs.
<    I found that there were alway serveral nop instructions followed with
the
<    jmp instruction in the program generated by suncc.
<    and this problem didn''t exist in the other 2 program generated
by gcc.
<     
<    err... Why does sun cc generate code using this pattern (a jmp followed
by
<    serveral nops)? For what?

  Ask on the tools-discuss and maybe attach the (small?) c-source too.

Regards,
<     
<    TIA
<     
<     
<    Regards,
<    TJ

< _______________________________________________
< dtrace-discuss mailing list
< dtrace-discuss at opensolaris.org


-- 
Sean.
.

"???? TaoJie" <eulertao at gmail.com> wrote:
> there''s no sun studio mailling list in opensolaris.org, i raise my
problem
> here :)
>
> a simple piece of code, 2 nested while loops, and some if statements inside
> each while loop.
>
> i compiled this program in Solaris with sun cc (sun studio 12)  and gcc.
> and i compiled it in Debian with gcc as well.
>
> Then I dumped (disassembled) these 3 programs.
> I found that there were alway serveral nop instructions followed with the
> jmp instruction in the program generated by suncc.
> and this problem didn''t exist in the other 2 program generated by
gcc.
This is how unoptimized sparc code looks since at least 21 years.

J?rg

-- 
 EMail:joerg at schily.isdn.cs.tu-berlin.de (home) J?rg Schilling D-13353 Berlin
       js at cs.tu-berlin.de                (uni)  
       schilling at fokus.fraunhofer.de     (work) Blog:
http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily

oh! yes! I forgot this...
no optimization option was set when I called cc, it''s my fault :)


2007/9/11, Joerg Schilling <Joerg.Schilling at
fokus.fraunhofer.de>:
>
> "???? TaoJie" <eulertao at gmail.com> wrote:
>
> > there''s no sun studio mailling list in opensolaris.org, i
raise my
> problem
> > here :)
> >
> > a simple piece of code, 2 nested while loops, and some if statements
> inside
> > each while loop.
> >
> > i compiled this program in Solaris with sun cc (sun studio 12)  and
gcc.
> > and i compiled it in Debian with gcc as well.
> >
> > Then I dumped (disassembled) these 3 programs.
> > I found that there were alway serveral nop instructions followed with
> the
> > jmp instruction in the program generated by suncc.
> > and this problem didn''t exist in the other 2 program
generated by gcc.
>
> This is how unoptimized sparc code looks since at least 21 years.
>
> J?rg
>
> --
> EMail:joerg at schily.isdn.cs.tu-berlin.de (home) J?rg Schilling D-13353
> Berlin
>       js at cs.tu-berlin.de                (uni)
>       schilling at fokus.fraunhofer.de     (work) Blog:
> http://schily.blogspot.com/
> URL:  http://cdrecord.berlios.de/old/private/
> ftp://ftp.berlios.de/pub/schily
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20070912/8bcd4319/attachment.html>

On 9/11/07, ?? TaoJie <eulertao at gmail.com>
wrote:
> I found that there were alway serveral nop instructions followed with the
> jmp instruction in the program generated by suncc.
>
> err... Why does sun cc generate code using this pattern (a jmp followed by
> serveral nops)? For what?
In some cases, the compiler *needs* to put a nop instruction right
after the branch because SPARC implements delayed branches.

If the compiler could not find a useful instruction to fill the delay
slot, then it puts a nop there. With no optimizations turned on, the
compiler will not spend time to do any of that, so that it can trade
code quality for faster compile time.

Rayson


>
> TIA
>
>
> Regards,
> TJ
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>