Re: Test case for PR 321

[Neil Booth <neil at daikokuya dot demon dot co dot uk>]

Jeffrey A Law wrote:-

> Doesn't this test depend on the endianness of the target?

I don't think so - it's just doing long long masks and rotates and
indexing on the result and checking if it matches a number.  There's
no assumption about byte order (that I can see anyway).

> You also need to disable the test if long long isn't a 64bit type
> (consider that we have some ports that don't have supprot for 64bit
> integer types).

OK.  What's the standard way of doing this?  Is the test below enough?

Neil.

	* gcc.c-torture/execute/longlong.c: New test.

Index: gcc.c-torture/execute/longlong.c
===================================================================
RCS file: longlong.c
diff -N longlong.c
--- /dev/null	Tue May  5 13:32:27 1998
+++ longlong.c	Sun Jan 14 23:57:07 2001
@@ -0,0 +1,39 @@
+/* Source: PR 321 modified for test suite by Neil Booth 14 Jan 2001.  */
+
+typedef unsigned long long uint64;
+unsigned int pars;
+
+uint64 b[32];
+uint64 *r = b;
+
+void alpha_ep_extbl_i_eq_0()
+{
+  unsigned int rb, ra, rc;
+
+  rb  = (((unsigned int)(pars) >> 27)) & 0x1fUL;
+  ra  = (((unsigned int)(pars) >> 5)) & 0x1fUL;
+  rc  = (((unsigned int)(pars) >> 0)) & 0x1fUL;
+  {
+    uint64 temp = ((r[ra] >> ((r[rb] & 0x7) << 3)) & 0x00000000000000FFLL); 
+    if (rc != 31) 
+      r[rc] = temp;  
+  }
+}
+
+int 
+main(void)
+{
+  if (sizeof (uint64) == 64)
+    {
+      b[17] = 0x0000000000303882; /* rb */
+      b[2] = 0x534f4f4c494d000a; /* ra & rc */
+
+      pars = 0x88000042;	/* 17, 2, 2 coded */
+      alpha_ep_extbl_i_eq_0();
+
+      if (b[2] != 0x4d)
+	abort ();
+    }
+
+  exit (0);
+}