On Wed, Aug 08, 2018 at 04:03:32AM +0100, Ramsay Jones wrote:
>
>
> On 07/08/18 14:03, Luc Van Oostenryck wrote:
> > No functional changes, just moving the core around
>
> s/core/code/
>
> > in preparation for shift-shift simplification.
> >
> > Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx>
> > ---
> > simplify.c | 15 ++++++++++-----
> > 1 file changed, 10 insertions(+), 5 deletions(-)
> >
> > diff --git a/simplify.c b/simplify.c
> > index 9e4a496fd..911ab11a9 100644
> > --- a/simplify.c
> > +++ b/simplify.c
> > @@ -627,14 +627,19 @@ static int simplify_shift(struct instruction *insn, pseudo_t pseudo, long long v
> > break;
> > case OP_LSR:
> > size = operand_size(insn, pseudo);
> > - /* fall through */
> > + if (value >= size)
> > + goto zero;
> > + switch(DEF_OPCODE(def, pseudo)) {
> > + case OP_LSR:
> > + goto case_shift_shift;
> > + }
> > + break;
> > case OP_SHL:
> > if (value >= size)
> > goto zero;
> > - if (pseudo->type != PSEUDO_REG)
> > - break;
> > - def = pseudo->def;
> > - if (def->opcode == insn->opcode) {
> > + switch(DEF_OPCODE(def, pseudo)) {
> > + case OP_SHL:
> > + case_shift_shift: // also for LSR - lSR
> > if (def == insn) // cyclic DAG!
> > break;
> > src2 = def->src2;
> >
>
> Hmm, OK, it took me quite some time to convince myself that
> there were, in fact, no functional changes. I could not do
> that with the patch text alone - I needed to look up the
> definition of the DEF_OPCODE macro. (Even then, I had to
> stare at the 'pseudo->type != PSEUDO_REG' condition along
> with the following control flow, very very hard! ;-) ).
I know. I could maybe have splitted this patch in two but ...
And this DEF_OPCODE() is a bit disgusting. At several times
I was like: "It's too ugly and needs to go away" and then a
few days later "Grr, I really need something like this".
But I think that once the real shift-shift simplification
is added, the code is quite readable.
-- Luc
--
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
