[avr-gcc-list] Using hardware multiplier on ATmega128L

[Alessio Mattÿffffe8]

Hi all,
 
I'm working with "Wireless Sensor Networks" using the platform Crossbow MICA2 
(www.xbow.com) + TinyOS (www.tinyos.net)
 
In particular, inside the MICA2 sensor node there is an ATMEL Atmega128L 
microcontroller ....
 
I want to use the hardware multiplier located inside Atmega128L microcontroller 
to improve the performances of my FFT algorithm under MICA2.
In particular I have this kind of instruction on my NeSC program (NeSC is a 
"dialect" of C language ...):
 
Wr = ((int32_t)c * (int32_t)real[k+n1] - (int32_t)s * (int32_t)imag[k+n1]) >> 
15;
 
where Wr,c,real[k+n1],s and imag[k+n1] are declared as int16_t variables.
 
The problem is that the computation time of above multiply is very high ... and 
therefore (if possible !!!) I want to improve the performances ...
 
On Internet I discover this document 
(http://www.atmel.com/dyn/resources/prod_documents/DOC1631.PDF) which describes 
the use of the hardware multiplier on AVR Microcontroller: the problem is that 
the 16-bit x 16-bit =32 bit operation is written using Assembly code, so I have 
to use Inline Asm inside my program and I have never used it before.
 
Has anybody implemented a solution like this yet ???
 
Does anybody known if documentation (I found only 
http://hubbard.engr.scu.edu/embedded/avr/doc/avr-libc/avr-libc-user-manual/inline_asm.html)
 about Inline Assembly is available ???
 
Please help me !!!
 
Thanks in advance,
 
Alessio


                                
---------------------------------
Nuovo Yahoo! Messenger E' molto più divertente: Audibles, Avatar, Webcam, 
Giochi, Rubrica… Scaricalo ora!