Re: Issue with DI and SI mode

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

 



Hi!

On Tue, Jan 21, 2020 at 04:09:44PM +0100, Henri Cloetens wrote:
> long long int mem_read( long long int *a)
> {
> return(a[1000]) ;
> }
> 
> and the compiler (gcc 9.2.0) implements that as 2 32-bit reads.
> 
> The "movdi" is defined, the optimizers are turned on, and in the
> rtl dump, I see the "split2" step is splitting the "DI" in 2x "SI".
> 
> Any idea why this happens ?. Where should I look for this ?.

The split2 dump file will tell you what splitter was used (where it
was defined, what md file and line no, and what function in insn-emit.c
it ended up in).

A splitter is used if its pattern matches and its split condition is
true.  It isn't checked if the result is cheaper some way; it isn't
even checked if the resulting insns are valid at all.


Segher



[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