Hi

or32 cannot support multiply calculation?
Are there any option to create assembler to emulate this?

or32-uclinux-ld -Ttext 0x0 start.o utils.o cpu_model.o mb_compress.o
-o cpu_model
cpu_model.o: In function `compress':
or1k/practice10/soft01/cpu_model.c:127: undefined reference to `___mulsi3'

Line 127 (cpp_model.c)
read(PPMBASE+(fh*info->exp_w+bh*info->exp_w+fv+bv)*4, d);

Regards,
swaka

On Thu, 2009-01-22 at 12:49 +0100, swaka at hdlab.co.jp wrote: Hi Swaka, OR32 supports multiplication just fine - you wouldn't be able to run Linux otherwise! The simple example you are building on has no library support - it's a bare metal test. That's why I have to add utils.c, to provide simple character I/O by using the extended l.nop operations. However if you link in the Linux/uClibc infrastructure, which are part of the tool chain this will all work. There are other, simpler, operating systems which run on OR32 (notably RTEMS), but I've never tried using them. Alternatively you can dig out just the implementation of ___mulsi3 etc (it should be somewhere in the uClibc/Linux stuff and build a mini-library of your own. There may be someone else on this list, who has done more with this, who can help. Try looking on Rich D'Addio's website: www.meansoffreedom.com. If you make progress in this area, perhaps you could write up an explanation and we'll put it on the website for the benefit of others. HTH, Jeremy -- Tel: +44 (1202) 416955 Cell: +44 (7970) 676050 SkypeID: jeremybennett Email: jeremy.bennett at embecosm.com Web: www.embecosm.com

On Thu, 2009-01-22 at 12:49 +0100, swaka at hdlab.co.jp wrote: Hi Swaka, My earlier comments were about using soft multiplication and division. However all the functions are immediately available with hardware multiplication and division. Just add the -mhard-div switch to your GCC compilations: or32-uclinux-gcc -mhard-div .... That should be your easiest solution. HTH, Jeremy -- Tel: +44 (1202) 416955 Cell: +44 (7970) 676050 SkypeID: jeremybennett Email: jeremy.bennett at embecosm.com Web: www.embecosm.com

Hi,

"-mhard-div" could not resolve this issue.

But I found two solution.

1)
libgcc.a was missing my environment.
There are ___mulsi3 in libgcc.a.

2)
or32-uclinux-gcc seems not create multiply/divider instruction by default.
"-mhard-mul" option creates mul instruction.

Regards,
swaka

On Mon, 2009-01-26 at 03:21 +0100, swaka at hdlab.co.jp wrote:

Hi Swaka, Looks like you have the problem solved. I'm not sure how much of libgcc.a will work outside a Linux environment - let us know if you run into problems. My mistake on the hardware option. I thought -mhard-div implied -mhard-mul. I now know correctly :-). When you come to use the actual Verilog, bear in mind that the default configuration creates the l.mul/l.mulu instructions, but NOT the l.div/l.divu instructions. You have to configure these in one of the Verilog headers if you want them. Best wishes, Jeremy -- Tel: +44 (1202) 416955 Cell: +44 (7970) 676050 SkypeID: jeremybennett Email: jeremy.bennett at embecosm.com Web: www.embecosm.com

Hi all,

I use the pre-compiled toolchain and have the same issue.
Before adding "-mhard-mul", "-mhard-div" in or32-elf-gcc option, there is no any "l.mul/l.mulu/l.div/l.divu" instruction used in compiled program.

Is there any other similar option?
Any document about this option?
Thanks.

Nick