lra: fix spill_hard_reg_in_range clobber check

FROM..TO range might contain NOTE_INSN_DELETED insns, for which the
corresponding entries in lra_insn_recog_data[] are NULLs.  Example from
the problematic code from PR87596:

    (note 148 154 68 7 NOTE_INSN_DELETED)

lra_insn_recog_data[] is used directly only when the insn in question
is taken from insn_bitmap, which is not the case here.  In other
situations lra_get_insn_recog_data () guarded by INSN_P () or other
stricter predicate are used.  So we need to do this here as well.

A tiny detail worth noting: I put the INSN_P () check before the
insn_bitmap check, because I believe that insn_bitmap can contain only
real insns anyway.

gcc/ChangeLog:

2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>

	PR rtl-optimization/87596
	* lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
	lra_get_insn_recog_data () instead of lra_insn_recog_data[]
	for instructions in FROM..TO range.

gcc/testsuite/ChangeLog:

2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>

	PR rtl-optimization/87596
	* gcc.target/i386/pr87596.c: New test.

From-SVN: r265306
This commit is contained in:
Ilya Leoshkevich 2018-10-19 08:33:52 +00:00 committed by Ilya Leoshkevich
parent ba9a8625b0
commit 3664a0f184
4 changed files with 34 additions and 3 deletions

View File

@ -1,3 +1,10 @@
2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
PR rtl-optimization/87596
* lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
lra_get_insn_recog_data () instead of lra_insn_recog_data[]
for instructions in FROM..TO range.
2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
* cfgexpand.c (expand_one_var): Use specific wording in error message

View File

@ -5696,12 +5696,15 @@ spill_hard_reg_in_range (int regno, enum reg_class rclass, rtx_insn *from, rtx_i
continue;
for (insn = from; insn != NEXT_INSN (to); insn = NEXT_INSN (insn))
{
lra_insn_recog_data_t id = lra_insn_recog_data[uid = INSN_UID (insn)];
struct lra_static_insn_data *static_id = id->insn_static_data;
struct lra_static_insn_data *static_id;
struct lra_insn_reg *reg;
if (bitmap_bit_p (&lra_reg_info[hard_regno].insn_bitmap, uid))
if (!INSN_P (insn))
continue;
if (bitmap_bit_p (&lra_reg_info[hard_regno].insn_bitmap,
INSN_UID (insn)))
break;
static_id = lra_get_insn_recog_data (insn)->insn_static_data;
for (reg = static_id->hard_regs; reg != NULL; reg = reg->next)
if (reg->regno == hard_regno)
break;

View File

@ -1,3 +1,8 @@
2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
PR rtl-optimization/87596
* gcc.target/i386/pr87596.c: New test.
2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
* gnat.dg/frame_overflow2.adb: New test.

View File

@ -0,0 +1,16 @@
/* LRA corner case which triggered a segfault. */
/* Reduced testcase by Arseny Solokha. */
/* { dg-do compile { target int128 } } */
/* { dg-options "-O1 -fschedule-insns -ftrapv -funroll-all-loops -fno-tree-dominator-opts -fno-tree-loop-im" } */
void
wh (__int128 *ku)
{
unsigned int *dp;
while (*ku < 1)
{
*dp <<= 32; /* { dg-warning "left shift count >= width of type" } */
++*ku;
}
}