Re: PPC compiler splits up ints & shorts?

[Andrew Johnson <anj@aps.anl.gov>]

Laznovsky, Michael wrote:
> Hi again- now that I know it's the "-mstrict-align" that's causing the
> byte operations, I still have 2 questions...
>
> 1) I see that this switch is provided by EPICS/base/config/CONFIG.Vx.mv2400:
>
> 	ARCH_DEP_CFLAGS = -mcpu=604 --no-builtin -mstrict-align -mlongcall
>
>    What's the proper way to override this in a local work area config directory?

You should be able to just include a new definition of ARCH_DEP_CFLAGS in 
your Makefile, inside an architecture test comparing $(T_A) against your 
specific target arch.

> 2) What switch (or lack of) does this compiler need to turn this feature off?
>    A quick naive test of re-running just the gmake-generated ccpcc command without
>    the mstrict switch makes the assembly phase barf, though I'm easily convinced
>    that this may be due to the lack of proper environment caused by running it
>    in isolation:

Correct, the R3.14 build system sets up some environment variables that 
tell the compiler where to find its components.

You should read this tech-talks message first though, which describes why 
we added that -mstrict-align flag in the first place:
  http://www.aps.anl.gov/epics/tech-talk/2000/msg00529.php

You can discover whether your compiler includes the fix for that original 
problem by running
    <tornado>/host/<arch>/bin/ccppc -v
If you find the number 28373 in the output from that command you *may* be 
ok to omit the -mstrict-align in that our original bug *probably* wont recur.

- Andrew
--
Dear God, I didn't think orange went with purple until I saw
the sunset you made last night.  That was really cool. - Caro