Greetings All,
Please forgive me for crossposting the following (it's also posted to the
newsgroup codewarrior.palm), but I wanted to get maximum exposure to this.
Also I was curious if PODS 1.1 can handle these types of huge switch
statements without choking...
-------------------
I've come across a somewhat disappointing problem with CW9.3 with regards to
PNO development that I was hoping someone might know a workaround for. Let's
say I was given a 10-bit integer X that I wanted to do a switch on for all
1024 cases. Setting this up as a jump table in C is easily done:
switch (X & 0x3ff)
{
case 0: do_something_0... ; break;
case 1: do_something_1... ; break;
...
case 1023: do_something_1023 ; break;
}
The PNO compiler of CW 9.3 correctly converts this to an efficient
pc-relative jump table without much effort.
The problem that I'm seeing is if this jump table is extended to encompass a
13-bit integer (8192 unique cases) or larger. The PNO compiler in CW9.3
chokes on this totally, regardless of the complexity of the
do_something_xxxx entries. They could be as simple as adding 1 to a local
variable, and the compiler will still choke to the point of crashing
completely and bringing the CW IDE down with it.
I tried to do a work around with inline assembly as follows:
void assmebly_jump_table(UInt16 X)
{
asm
{
: assembly code to set up a pc_relative jump to a table of branches
to C code
b case_0
b case_1
...
b case_8191
}
case_0: do_something_0... ; goto jump_end;
case_1: do_something_1... ; goto jump_end;
...
case 8192: do_something_8191... ; goto jump_end;
jump_end:
do_finish_up_code;
return;
}
The compiler can handle the large inline assembly portion just fine, but
then I encounter a different problem with the PNO compiler. Since the
compiler essentially ignores everything that happens in the inline assembly
portion of the function with regards to how it compiles the C code, it never
sees all the different branches to each code case. Since it isn't aware that
there is a possibility of branching to anything other than the case_0, it
dead strips all remaining cases, and I can't figure out a way to keep it
from doing this...
So anyway, after all that, does anyone have any idea how t get around this?
Regards,
-Robert Hildinger
rh_public AT mobilevoodoo DOT com
--
For information on using the Palm Developer Forums, or to unsubscribe, please see
http://www.palmos.com/dev/support/forums/
