[PATCH] bitint: Fix bitfield loads in handle_cast [PR114433]

[PATCH] bitint: Fix bitfield loads in handle_cast [PR114433]

[Jakub Jelinek]

Hi!

We ICE on the following testcase, because handle_cast was incorrectly
testing !m_first to see whether it should use m_data[m_bitfld_load + 1]
or fresh SSA_NAME for a PHI result.
Now, m_first is in the routine sometimes temporarily cleared in between
doing prepare_data_in_out and the !m_first check and only before returning
restored from the save_first copy.
Without this patch, we try to use the same SSA_NAME (_12 here) in 2
different PHI results which is obviously invalid IL and ICEs very quickly.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2024-03-23  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/114433
	* gimple-lower-bitint.cc (bitint_large_huge::handle_cast): For
	m_bitfld_load check save_first rather than m_first.

	* gcc.dg/torture/bitint-68.c: New test.

--- gcc/gimple-lower-bitint.cc.jj	2024-03-22 14:54:41.767972684 +0100
+++ gcc/gimple-lower-bitint.cc	2024-03-22 17:13:43.702249666 +0100
@@ -1506,7 +1506,7 @@ bitint_large_huge::handle_cast (tree lhs
 	  if (m_bitfld_load)
 	    {
 	      tree t4;
-	      if (!m_first)
+	      if (!save_first)
 		t4 = m_data[m_bitfld_load + 1];
 	      else
 		t4 = make_ssa_name (m_limb_type);
--- gcc/testsuite/gcc.dg/torture/bitint-68.c.jj	2024-03-22 17:26:29.841713520 +0100
+++ gcc/testsuite/gcc.dg/torture/bitint-68.c	2024-03-22 17:26:51.941409603 +0100
@@ -0,0 +1,28 @@
+/* PR tree-optimization/114433 */
+/* { dg-do run { target bitint } } */
+/* { dg-options "-std=c23" } */
+/* { dg-skip-if "" { ! run_expensive_tests }  { "*" } { "-O0" "-O2" } } */
+/* { dg-skip-if "" { ! run_expensive_tests } { "-flto" } { "" } } */
+
+#if __BITINT_MAXWIDTH__ >= 511
+struct S { int : 31; _BitInt(511) b : 300; } s;
+
+__attribute__((noipa)) _BitInt(511)
+foo (void) 
+{
+  return s.b << 1;
+}
+#endif
+
+int
+main ()
+{
+#if __BITINT_MAXWIDTH__ >= 511
+  s.b = 642460398785925402356009598661384732715767737595497615767135001949421105426024498988100867wb;
+  if (foo () != ((_BitInt(511)) 642460398785925402356009598661384732715767737595497615767135001949421105426024498988100867wb) << 1)
+    __builtin_abort ();
+  s.b = 2655156766298562299560755420298083843774074962786295887660222690220887wb;
+  if (foo () != ((_BitInt(511)) 2655156766298562299560755420298083843774074962786295887660222690220887wb) << 1)
+    __builtin_abort ();
+#endif
+}

	Jakub

Re: [PATCH] bitint: Fix bitfield loads in handle_cast [PR114433]

[Richard Biener]

> Am 23.03.2024 um 08:59 schrieb Jakub Jelinek <jakub@redhat.com>:
> 
> Hi!
> 
> We ICE on the following testcase, because handle_cast was incorrectly
> testing !m_first to see whether it should use m_data[m_bitfld_load + 1]
> or fresh SSA_NAME for a PHI result.
> Now, m_first is in the routine sometimes temporarily cleared in between
> doing prepare_data_in_out and the !m_first check and only before returning
> restored from the save_first copy.
> Without this patch, we try to use the same SSA_NAME (_12 here) in 2
> different PHI results which is obviously invalid IL and ICEs very quickly.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

Ok

Richard 

> 2024-03-23  Jakub Jelinek  <jakub@redhat.com>
> 
>    PR tree-optimization/114433
>    * gimple-lower-bitint.cc (bitint_large_huge::handle_cast): For
>    m_bitfld_load check save_first rather than m_first.
> 
>    * gcc.dg/torture/bitint-68.c: New test.
> 
> --- gcc/gimple-lower-bitint.cc.jj    2024-03-22 14:54:41.767972684 +0100
> +++ gcc/gimple-lower-bitint.cc    2024-03-22 17:13:43.702249666 +0100
> @@ -1506,7 +1506,7 @@ bitint_large_huge::handle_cast (tree lhs
>      if (m_bitfld_load)
>        {
>          tree t4;
> -          if (!m_first)
> +          if (!save_first)
>        t4 = m_data[m_bitfld_load + 1];
>          else
>        t4 = make_ssa_name (m_limb_type);
> --- gcc/testsuite/gcc.dg/torture/bitint-68.c.jj    2024-03-22 17:26:29.841713520 +0100
> +++ gcc/testsuite/gcc.dg/torture/bitint-68.c    2024-03-22 17:26:51.941409603 +0100
> @@ -0,0 +1,28 @@
> +/* PR tree-optimization/114433 */
> +/* { dg-do run { target bitint } } */
> +/* { dg-options "-std=c23" } */
> +/* { dg-skip-if "" { ! run_expensive_tests }  { "*" } { "-O0" "-O2" } } */
> +/* { dg-skip-if "" { ! run_expensive_tests } { "-flto" } { "" } } */
> +
> +#if __BITINT_MAXWIDTH__ >= 511
> +struct S { int : 31; _BitInt(511) b : 300; } s;
> +
> +__attribute__((noipa)) _BitInt(511)
> +foo (void)
> +{
> +  return s.b << 1;
> +}
> +#endif
> +
> +int
> +main ()
> +{
> +#if __BITINT_MAXWIDTH__ >= 511
> +  s.b = 642460398785925402356009598661384732715767737595497615767135001949421105426024498988100867wb;
> +  if (foo () != ((_BitInt(511)) 642460398785925402356009598661384732715767737595497615767135001949421105426024498988100867wb) << 1)
> +    __builtin_abort ();
> +  s.b = 2655156766298562299560755420298083843774074962786295887660222690220887wb;
> +  if (foo () != ((_BitInt(511)) 2655156766298562299560755420298083843774074962786295887660222690220887wb) << 1)
> +    __builtin_abort ();
> +#endif
> +}
> 
>    Jakub
>