Hi Kyrill
On 04/01/18 16:36, Kyrill Tkachov wrote:
Hi Sudi,
On 04/01/18 15:35, Sudakshina Das wrote:
Hi
The bug reported a particular test di-longlong64-sync-1.c failing when
run on arm-linux-gnueabi with options -mthumb -march=armv5t -O[g,1,2,3]
and -mthumb -march=armv6 -O[g,1,2,3].
According to what I could see, the crash was caused because of the
explicit VOIDmode argument that was sent to emit_store_flag_force ().
Since the comparing argument was a long long, it was being forced into a
VOID type register before the comparison (in prepare_cmp_insn()) is done.
As pointed out by Kyrill, there is a comment on emit_store_flag() which
says "MODE is the mode to use for OP0 and OP1 should they be CONST_INTs.
If it is VOIDmode, they cannot both be CONST_INT". This condition is
not true in this case and thus I think it is suitable to change the
argument.
Testing done: Checked for regressions on bootstrapped
arm-none-linux-gnueabi and arm-none-linux-gnueabihf and added new test
cases.
Sudi
ChangeLog entries:
*** gcc/ChangeLog ***
2017-01-04 Sudakshina Das <sudi....@arm.com>
PR target/82096
* optabs.c (expand_atomic_compare_and_swap): Change argument
to emit_store_flag_force.
*** gcc/testsuite/ChangeLog ***
2017-01-04 Sudakshina Das <sudi....@arm.com>
PR target/82096
* gcc.c-torture/compile/pr82096-1.c: New test.
* gcc.c-torture/compile/pr82096-2.c: Likwise.
diff --git a/gcc/testsuite/gcc.c-torture/compile/pr82096-1.c
b/gcc/testsuite/gcc.c-torture/compile/pr82096-1.c
new file mode 100644
index 0000000..07eb5f6
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/compile/pr82096-1.c
@@ -0,0 +1,10 @@
+/* { dg-do compile { target { arm*-*-* } } } */
+/* { dg-options "-march=armv5t -mthumb -mfloat-abi=soft" } */
The tests in gcc.c-torture/compile/ are supposed to be
target-independent, so
it's best to not gate it on target arm*-*-*. Best to add the arm-specific
options that you want using a dg-additional-options directive gated on
target arm*-*-*.
+
+static long long AL[24];
+
+int
+check_ok (void)
+{
+ return (__sync_bool_compare_and_swap (AL+1, 0x200000003ll,
0x1234567890ll));
+}
diff --git a/gcc/testsuite/gcc.c-torture/compile/pr82096-2.c
b/gcc/testsuite/gcc.c-torture/compile/pr82096-2.c
new file mode 100644
index 0000000..2570b16
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/compile/pr82096-2.c
@@ -0,0 +1,10 @@
+/* { dg-do compile { target { arm*-*-* } } } */
+/* { dg-options "-march=armv6 -mthumb -mfloat-abi=soft" } */
+
+static long long AL[24];
+
+int
+check_ok (void)
+{
+ return (__sync_bool_compare_and_swap (AL+1, 0x200000003ll,
0x1234567890ll));
+}
I don't think we need an armv6 test here as the root cause is the same
as on armv5t AFAICT
so this won't give us any extra code coverage.
I have made the changes in the test files as requested.
Thanks
Sudi
Thanks,
Kyrill
diff --git a/gcc/optabs.c b/gcc/optabs.c
index 225e955..45b018e 100644
--- a/gcc/optabs.c
+++ b/gcc/optabs.c
@@ -6295,7 +6295,7 @@ expand_atomic_compare_and_swap (rtx *ptarget_bool, rtx *ptarget_oval,
if (cc_reg)
{
target_bool = emit_store_flag_force (target_bool, EQ, cc_reg,
- const0_rtx, VOIDmode, 0, 1);
+ const0_rtx, mode, 0, 1);
goto success;
}
goto success_bool_from_val;
@@ -6323,7 +6323,7 @@ expand_atomic_compare_and_swap (rtx *ptarget_bool, rtx *ptarget_oval,
success_bool_from_val:
target_bool = emit_store_flag_force (target_bool, EQ, target_oval,
- expected, VOIDmode, 1, 1);
+ expected, mode, 1, 1);
success:
/* Make sure that the oval output winds up where the caller asked. */
if (ptarget_oval)
diff --git a/gcc/testsuite/gcc.c-torture/compile/pr82096.c b/gcc/testsuite/gcc.c-torture/compile/pr82096.c
new file mode 100644
index 0000000..9fed28c
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/compile/pr82096.c
@@ -0,0 +1,9 @@
+/* { dg-additional-options "-march=armv5t -mthumb -mfloat-abi=soft" { target arm*-*-* } } */
+
+static long long AL[24];
+
+int
+check_ok (void)
+{
+ return (__sync_bool_compare_and_swap (AL+1, 0x200000003ll, 0x1234567890ll));
+}
