REG_WORDS_BIG_ENDIAN

[Jeff Law]

On 07/12/2018 06:19 PM, Michael Eager wrote:
> Hi All --
> 
> I'm updating GCC on a dual-endianity processor.  In little-endian mode,
> memory is little-endian, but the core remains big-endian.  Double-word
> values are in even/odd register pairs, with the MSW in the
> lower-numbered (even) register.  Values in memory are stored with LSW in
> the low-address word.
How unpleasant.

> 
> From info:
> REG_WORDS_BIG_ENDIAN [Macro]
> Â  On some machines, the order of words in a multiword object differs
> Â  between registers in memory. In such a situation, define this macro
> Â  to describe the order of words in a register. The macro
> Â  WORDS_BIG_ENDIAN controls the order of words in memory.
> 
> I set REG_WORDS_BIG_ENDIAN to 1 and WORDS_BIG_ENDIAN to the memory
> endianity.
> 
> None of the targets in the GCC repo define this.  Does anyone know of an
> out-of-repo target which sets REG_WORDS_BIG_ENDIAN?
The closest in-tree target is probably c6x, allows selecting big/little
for WORDS_BIG_ENDIAN, but REG_WORDS_BIG_ENDIAN is always 0.


> 
> Most test cases pass, but I'm finding a few cases which fail to
> calculate a subreg correctly.  Subreg_lowpart_offset() is called while
> emitting RTL, for example, to truncate a DImode integer to SImode.  But
> subreg_lowpart_offset() does not know whether it is computing the offset
> of a memory value or register value.  It returns the word offset
> according to memory endianity, ignoring REG_WORDS_BIG_ENDIAN.  (This is
> also the case with subreg_size_lowpart_offset()).
I think you'd probably want to avoid (subreg (mem)) if at all possible.
That's probably good advice in general anyway.

> 
> I'm working on a patch to pass the inner RTX to subreg_lowpart_offset(),
> but this function is called in many places and it isn't always clear
> what object the subreg applies to, or even if there is an object.  For
> example, maybe_mode_change() calls subreg_size_lowpart_offset() before
> generating a REG rtx.  This seems to be the wrong approach.
> 
> There are other subreg functions which seem to be specifically for
> registers, like subreg_get_info() and subreg_regno_offset().  It isn't
> clear whether these can replace any of the calls to
> subreg_lowpart_offset().
About the only advice I'd have would be to look at the c6x and see if
it's handling any of this stuff correctly.  BUt I suspect you're largely
going to be on your own.

Jeff