Hi,
While testing a GCC patch (which I am not ready to submit yet), I
wrote a testcase which had produced wrong code (with an earlier
version of the patch). There was no other testcase for it in the
testsuite so I adding one.
This is a reduced testcase from the uboot PCIe code.
Thanks,
Andrew Pinski
* gcc.c-torture/execute/bswap-3.c: New testcase.
Index: ChangeLog
===================================================================
--- ChangeLog (revision 279099)
+++ ChangeLog (working copy)
@@ -1,3 +1,7 @@
+2019-12-08 Andrew Pinski <apin...@marvell.com>
+
+ * gcc.c-torture/execute/bswap-3.c: New test.
+
2019-12-08 Sandra Loosemore <san...@codesourcery.com>
Revert:
Index: gcc.c-torture/execute/bswap-3.c
===================================================================
--- gcc.c-torture/execute/bswap-3.c (nonexistent)
+++ gcc.c-torture/execute/bswap-3.c (working copy)
@@ -0,0 +1,25 @@
+/* { dg-require-effective-target int32plus } */
+
+int f(unsigned int a) __attribute__((noipa));
+int f(unsigned int a)
+{
+ return ((__builtin_bswap32(a))>>24) & 0x3;
+}
+
+
+int g(unsigned int a) __attribute__((noipa));
+int g(unsigned int a)
+{
+ return a&0x3;
+}
+
+int main(void)
+{
+ for (int b = 0; b <= 0xF; b++)
+ {
+ if (f(b) != g(b))
+ __builtin_abort ();
+ }
+ return 0;
+}
+
