FPGA Central

hi,

i'm a dsp programming newcomer, i was wondering what's the cleanest way
to read the V flag in the ST0 register using C...

i quickly searched through the optimizing compiler docs and ccs headers
and haven't found a directly suitable intrinsic yet...

what's your suggestion?

should i go to asm level?
maybe should i copy the status register in a safe place and then mask
it to expose the bit?

....thank you in advance

gst

gst wrote:
> hi,
>
> i'm a dsp programming newcomer, i was wondering what's the cleanest way
> to read the V flag in the ST0 register using C...
>
> i quickly searched through the optimizing compiler docs and ccs headers
> and haven't found a directly suitable intrinsic yet...
>
> what's your suggestion?
>
> should i go to asm level?
> maybe should i copy the status register in a safe place and then mask
> it to expose the bit?
>
> ...thank you in advance

HLLs deal with flags poorly if at all. For one thing, they're supposed
to insulate the programmer from grungy details. For another, such code
wouldn't be portable. (Flag sets differ among processors. E.g., alpha
doesn't have a carry flag.) Assembly is the way. A C-callable routine
pretties up the code.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Hi,
The C and asm both produce exectly same code if you do the V flag
polling, the AND instruction. No difference from performance, but code
in C is more readable and more portable.
Good luck
Alex

Too quick post. Why do you need for V polling if you can implement
algorithms to avoid overflow at all? If you check out Cirrus Logic'
modems (good old days!) I developed some DSP routines for them and OV
flag was not there in the HW. All ariphmetic was done without OV even
for V90 modulation .
Good luck,
Alex

(warning: horrible english writing ahead)

First of all, thank you to both of you for replying!

Sorry for having been disappeared for so long; anyway: you have been
read :-)

But I'm a still newbie, so: I can't find anything in the standard
headers that points to the overflow flag (or the whole registry), so,
to summarize, if I understand correctly, the suggestion is:

"do the operation in C, write a function intended to be used in C but
written in ASM to check the flag only"?

thank you again!

(operations without overflows: unfortunately i'm not that skilled, but
someday maybe i'll be able to do that kind of things, or at least
understand them ;-)) )

gst

Hi,
Does your question mean 'How OV status flag is detected?' or "How OV
flag is toggled?'.
The 1st questions normally ask a person who develops high level code,
like C/asm programmers. The 2nd question usually comes from hardware a
developer, who wants to implement arithmetic in Verilog/VHDL. Look like
you are from first group, means you have to poll OV all time you want
to get values exceeding your data format limits. You may poll one in
asm or in C, just check your hardware manual for software developers to
understand what kind of overhead C command has for OV polling on your
platform. May be, that is acceptable for you.
The 2nd question actually is an algorithm selection answer. Some
algorithms toggle OV flag before arithmetic is actually executed, the
other ones do it right after result of calculation is saved or in
parallel with write. You need to reference your hardware design
specification to understand which one is more appropriate for you.
Good luck,
Alex