On Wed, 13 Oct 2004 at 15:24 +0200, Daniel Schürmann <schuerm...@svea.de> wrote:
> From: Daniel Schürmann <schuerm...@svea.de>
>
> i might have found a problem with code generation and optimization.
>
> I use the command:
> msp430-gcc -mmcu=msp430x1132 -g -O2 -Wall -c -o Merten_Argus.o
> Merten_Argus.c
>
> This shows insight im mixed View:
> <<MA_get_brigtness.mix>>
> Here are the *.i and *.s files:
> <<Merten_Argus.i>> <<Merten_Argus.s>>
> In Funktion MA_get_brightness, the statement:
>
> buff *= im_SegA.im.MA_gain_multipier;
>
> is not compiled correctly. But when I delete the irrelevant line
>
> buff = (U32)(MA_brightness & 0x07ff );
>
> or switch off the optimization, everything works as desired.
Are you asking why mspgcc seems to be assuming that
im_SegA.im.MA_gain_multipier is a constant 32?
(The compiler sees that buff was initialized to 500,
and 16000/500 is 32)
The assignment of 500 to buff is after the line you deleted,
so it is surprising that that would have affected the results,
although I can see how turning off optimizaton would probably
force the multiply to be fully performed.
355 buff = 500;
356
357 if( im_SegA.im.MA_gain_multipier ){
- 0xefb4 <MA_get_brightness+18>: cmp.b #0, &0x1011 ;subst
r3 with As==00
- 0xefb8 <MA_get_brightness+22>: jz $+64 ;abs
dst addr 0xeff8
358 // Berechnung <FC>ber Gain
359 buff *= im_SegA.im.MA_gain_multipier;
- 0xefba <MA_get_brightness+24>: mov #16000, r12 ;#0x3e80
- 0xefbe <MA_get_brightness+28>: clr r13 ;
ps.
was Merten_Argus.i really the whole preprocessed output?
