[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [avr-gcc-list] Can someone benchmark this option please
From: |
Weddington, Eric |
Subject: |
RE: [avr-gcc-list] Can someone benchmark this option please |
Date: |
Wed, 23 Jul 2008 12:12:18 -0600 |
> -----Original Message-----
> From:
> address@hidden
> [mailto:address@hidden
> org] On Behalf Of Weddington, Eric
> Sent: Wednesday, July 23, 2008 11:18 AM
> To: Andy H; AVR-GCC
> Subject: RE: [avr-gcc-list] Can someone benchmark this option please
>
>
>
> > -----Original Message-----
> > From:
> > address@hidden
> > [mailto:address@hidden
> > org] On Behalf Of Andy H
> > Sent: Tuesday, July 22, 2008 6:40 PM
> > To: AVR-GCC
> > Subject: [avr-gcc-list] Can someone benchmark this option please
> >
> > Hi,
> >
> > I have noted problem where gcc is optimizing if-then-else
> > constructs -
> > with disastrous results. For example:
> >
> > if (a >= 0)
> > return 8;
> > else
> > return 0;
> >
> > (This is now reported as bug.)
> >
> > Can folks try the following gcc option
> >
> > -fno-if-conversion
> >
> > on their own code at -Os optimization, and see if this produce
> > better/worse or same code? This will help figure if the whole pass
> > should be skipped - or whether it still contains useful
> optimizations.
> >
>
> Hi Andy,
>
> I tried it on the BC100 Kit code for gcc, which uses these options:
>
> -mmcu=attiny861
> -gdwarf-2
> -Os
> -funsigned-char
> -funsigned-bitfields
> -fpack-struct
> -fshort-enums
> -Wall
> -Wstrict-prototypes
> -std=gnu99
> -ffunction-sections
> -fno-inline-small-functions
> -fno-split-wide-types
>
> When I added -fno-if-conversion, there was no change in code
> size: 5588 bytes.
I also tried it on the Butterfly Kit code for gcc. It saved a whopping 2 bytes:
Baseline: 13876 bytes
With -fno-if-conversion flag: 13874 bytes.
Eric