On Thu, Sep 19, 2019 at 5:43 PM Uros Bizjak <ubiz...@gmail.com> wrote:
>
> On Thu, Sep 19, 2019 at 5:30 PM Richard Biener <rguent...@suse.de> wrote:
> >
> >
> > Boostrapped and tested on x86_64-unknown-linux-gnu.
> >
> > OK?
>
> OK.
Hm, something is not working correctly here. For the testcase, I get:
main:
vmovq %xmm0, %xmm0
vpxor %xmm1, %xmm1, %xmm1
vpsubq %xmm0, %xmm1, %xmm1
vpmaxsq %xmm1, %xmm0, %xmm0
vmovq %xmm0, %rax
movabsq %rax, .LC0+11532131096
xorl %eax, %eax
ret
The first insn uses uninitialized reg.
The _.stv pass misses initialization of r94 reg:
(note 2 3 7 2 NOTE_INSN_FUNCTION_BEG)
(note 7 2 24 2 NOTE_INSN_DELETED)
(insn 24 7 8 2 (set (subreg:V2DI (reg:DI 93) 0)
(vec_concat:V2DI (reg:DI 94)
(const_int 0 [0]))) "pr67271.c":11:17 -1
(nil))
Uros.
> Thanks,
> Uros.
>
> > Thanks,
> > Richard.
> >
> > 2019-09-19 Richard Biener <rguent...@suse.de>
> >
> > PR target/91814
> > * config/i386/i386-features.c (gen_gpr_to_xmm_move_src):
> > Force operand to a register if it isn't nonimmediate_operand.
> >
> > Index: gcc/config/i386/i386-features.c
> > ===================================================================
> > --- gcc/config/i386/i386-features.c (revision 275959)
> > +++ gcc/config/i386/i386-features.c (working copy)
> > @@ -668,10 +668,13 @@ scalar_chain::emit_conversion_insns (rtx
> > static rtx
> > gen_gpr_to_xmm_move_src (enum machine_mode vmode, rtx gpr)
> > {
> > + if (!nonimmediate_operand (gpr, GET_MODE_INNER (vmode)))
> > + gpr = force_reg (GET_MODE_INNER (vmode), gpr);
> > switch (GET_MODE_NUNITS (vmode))
> > {
> > case 1:
> > - return gen_rtx_SUBREG (vmode, gpr, 0);
> > + /* We are not using this case currently. */
> > + gcc_unreachable ();
> > case 2:
> > return gen_rtx_VEC_CONCAT (vmode, gpr,
> > CONST0_RTX (GET_MODE_INNER (vmode)));
