Re: Simple Question on CPU specific Optimization Options

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



"Hofman, Gertjan (BI35)" <Gertjan.Hofman@xxxxxxxxxxxxx> writes:

> I have noticed that any optimization flag (-0 and up) removes
> consequtive read/write lines i.e. 
> 
>   *((unsigned char *) virtaddr) =  0xF0;
>   *((unsigned char *) virtaddr) =  0xFE;  
> 
> Even though they are writing different data to memory. Kind of scary.
> With -O0, the code runs as anticipated.

This is normal.  Think about use with a normal pointer.  The first
assignment is overwritten by the second before anything could possibly
read it.

You need to use a volatile qualified pointer here.  This is standard
C; it has nothing to do with gcc as such.

> In order to find out what is causing the removal I replaced -O with all
> the -fxxxxx options listed in man gcc. This does not replicate the
> problem ! In otherwords, -O does more that the sum of all the -fxxx
> options.

Correct.  Some code transformations are controlled only be the
presence of the -O option.

> 1. Is there a way to have the compiler tell me which options are being
> turned on using -O ?

There will be in gcc 4.3, but it won't help answer your actual
question.

> 2. Are there architecture specific optimizations which are not described
> in the man page ?

Yes.  Many.

Ian

[Index of Archives]     [Linux C Programming]     [Linux Kernel]     [eCos]     [Fedora Development]     [Fedora Announce]     [Autoconf]     [The DWARVES Debugging Tools]     [Yosemite Campsites]     [Yosemite News]     [Linux GCC]

  Powered by Linux