On 20/01/2020 18:25, Mihail Ionescu wrote: > Hi, > > > This patch fixes the uninitialised 'last_regno' variable introduced in: > https://gcc.gnu.org/ml/gcc-patches/2019-12/msg01299.html > and makes the clear_operation_p code more readable. > > *** gcc/ChangeLog *** > > 2020-01-20 Mihail-Calin Ionescu <mihail.ione...@arm.com> > > * gcc/config/arm/arm.c (clear_operation_p): > Initialise last_regno, skip first iteration > based on the first_set value and use ints instead > of the unnecessary HOST_WIDE_INTs. > > > > Is it ok for trunk? > > > Regards, > Mihail > > > ############### Attachment also inlined for ease of reply > ############### > > > diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c > index > 4bfceb973c47f64b2705c1ab47cc61c9f78508a9..52e1db22ac18bc281ca3deb8cf950eb3438c7e52 > 100644 > --- a/gcc/config/arm/arm.c > +++ b/gcc/config/arm/arm.c > @@ -13751,13 +13751,14 @@ ldm_stm_operation_p (rtx op, bool load, > machine_mode mode, > bool > clear_operation_p (rtx op, bool vfp) > { > - unsigned regno, last_regno; > + unsigned regno; > + unsigned last_regno = INVALID_REGNUM; > rtx elt, reg, zero; > - HOST_WIDE_INT count = XVECLEN (op, 0); > - HOST_WIDE_INT i, first_set = vfp ? 1 : 0; > + int count = XVECLEN (op, 0); > + int first_set = vfp ? 1 : 0; > machine_mode expected_mode = vfp ? E_SFmode : E_SImode; > > - for (i = first_set; i < count; i++) > + for (int i = first_set; i < count; i++) > { > elt = XVECEXP (op, 0, i); > > @@ -13789,14 +13790,14 @@ clear_operation_p (rtx op, bool vfp) > > if (vfp) > { > - if (i != 1 && regno != last_regno + 1) > + if (i != first_set && regno != last_regno + 1) > return false; > } > else > { > if (regno == SP_REGNUM || regno == PC_REGNUM) > return false; > - if (i != 0 && regno <= last_regno) > + if (i != first_set && regno <= last_regno) > return false; > } > >
OK. R.
