SEO

Hi again -

While trying to get some old code to work under HP-UX 11.11 with a
PA-RISC 2.0 CPU-based machine in 64 bit mode I have some assembler
code that doesn't compile.

Note that I am using gcc 4.2.3 to do this and so I am using the gnu as
assembler.

Still, this is the "offending" code part:

.import $$dyncall,MILLICODE /* sitting at the top of the file */

/* somewhere in a procedure: */

bl $$dyncall,%r31

The linker then barfs and says that $$dyncall is an unsatisfied
symbol. Looking at docs from HP (man, I used to work for HP for more
than 12 years back then) I see that there's a $$dyncall entry in the
32 bit /usr/lib/libmilli.a but no such symbol in
/usr/lib/PA20_64/libmilli.a ...

The 64 bit porting info I found also doesn't help.

Any ideas?

Thanks!!!

Best,
Frank

--

Frank Goenninger

frgo(at)mac(dot)com

"Don't ask me! I haven't been reading comp.lang.lisp long enough to
really know ..."

"Frank Goenninger" <dont-email-me@nomail.org> writes:

> Hi again -
>
> While trying to get some old code to work under HP-UX 11.11 with a
> PA-RISC 2.0 CPU-based machine in 64 bit mode I have some assembler
> code that doesn't compile.
>
> Note that I am using gcc 4.2.3 to do this and so I am using the gnu as
> assembler.
>
> Still, this is the "offending" code part:
>
> .import $$dyncall,MILLICODE /* sitting at the top of the file */
>
> /* somewhere in a procedure: */
>
> bl $$dyncall,%r31
>
> The linker then barfs and says that $$dyncall is an unsatisfied
> symbol. Looking at docs from HP (man, I used to work for HP for more
> than 12 years back then) I see that there's a $$dyncall entry in the
> 32 bit /usr/lib/libmilli.a but no such symbol in
> /usr/lib/PA20_64/libmilli.a ...

Solved:

gcc -march2.0

and static linking ...

(I don't know the details, though - it was a monkey style problem
solving technique ...)

Best,
Frank

--

Frank Goenninger

frgo(at)mac(dot)com

"Don't ask me! I haven't been reading comp.lang.lisp long enough to
really know ..."

Frank Bennington's wrote:
> While trying to get some old code to work under HP-UX 11.11 with a
> PA-RISC 2.0 CPU-based machine in 64 bit mode I have some assembler
> code that doesn't compile.
> bl $$dyncall,%r31
> The linker then barfs and says that $$dyncall is an unsatisfied
> symbol. but no such symbol in /usr/lib/PA20_64/libmilli.a ...

There is no need for $$dyncall on PA2.0 since there is an instruction
that does exactly that BVE,L.

>it was a monkey style problem solving technique

Yes, looking at a simple example of an indirect call would show that.

Dennis Handly <dhandly@convex.hp.com> writes:

> Frank Goenninger wrote:
>> While trying to get some old code to work under HP-UX 11.11 with a
>> PA-RISC 2.0 CPU-based machine in 64 bit mode I have some assembler
>> code that doesn't compile.
>> bl $$dyncall,%r31
>> The linker then barfs and says that $$dyncall is an unsatisfied
>> symbol. but no such symbol in /usr/lib/PA20_64/libmilli.a ...
>
> There is no need for $$dyncall on PA2.0 since there is an instruction
> that does exactly that BVE,L.
>
>>it was a monkey style problem solving technique
>
> Yes, looking at a simple example of an indirect call would show that.

You mean like

COPY gp, save_gp
LDD 16(fp), tmp
LDD 24(fp), gp
BVE,L (tmp), r2

COPY save_gp, gp

???

Thx!
Frank

--

Frank Goenninger

frgo(at)mac(dot)com

"Don't ask me! I haven't been reading comp.lang.lisp long enough to
really know ..."

Frank "frgo" a.k.a DG1SBG wrote:
> You mean like
> COPY gp, save_gp
> LDD 16(fp), tmp
> LDD 24(fp), gp
> BVE,L (tmp), r2
>
> COPY save_gp, gp

Yes and you'll need to set R29 (same as other calls):
LDO -48(%r30),%r29
LDD 16(%r31),%r19
.CALL ;in=26,29;out=28;
BVE,L (%r19),%r2
LDD 24(%r31),%r27
LDD -128(%r30),%r27

Dennis Handly <dhandly@convex.hp.com> writes:

> Yes and you'll need to set R29 (same as other calls):
> LDO -48(%r30),%r29
> LDD 16(%r31),%r19
> .CALL ;in=26,29;out=28;
> BVE,L (%r19),%r2
> LDD 24(%r31),%r27
> LDD -128(%r30),%r27

Thx!!

Best,
Frank

(formerly been working for HP in Germany for 12 years)
--

Frank Goenninger

frgo(at)mac(dot)com

"Don't ask me! I haven't been reading comp.lang.lisp long enough to
really know ..."