LoongArch: Fix atomic_exchange expanding [PR107713]

We used to expand atomic_exchange_n(ptr, new, mem_order) for subword types
into something like:

    {
      __typeof__(*ptr) t = atomic_load_n(ptr, mem_order);
      atomic_compare_exchange_n(ptr, &t, new, true, mem_order, mem_order);
      return t;
    }

It's incorrect because another thread may store a different value into *ptr
after atomic_load_n.  Then atomic_compare_exchange_n will not store into
*ptr, but atomic_exchange_n should always perform the store.

gcc/ChangeLog:

	PR target/107713
	* config/loongarch/sync.md
	(atomic_cas_value_exchange_7_<mode>): New define_insn.
	(atomic_exchange): Use atomic_cas_value_exchange_7_si instead of
	atomic_cas_value_cmp_and_7_si.

gcc/testsuite/ChangeLog:

	PR target/107713
	* gcc.target/loongarch/pr107713-1.c: New test.
	* gcc.target/loongarch/pr107713-2.c: New test.

gcc/config/loongarch/sync.md

@@ -448,6 +448,29 @@
 }
   [(set (attr "length") (const_int 32))])
 
+(define_insn "atomic_cas_value_exchange_7_<mode>"
+  [(set (match_operand:GPR 0 "register_operand" "=&r")
+	(match_operand:GPR 1 "memory_operand" "+ZC"))
+   (set (match_dup 1)
+	(unspec_volatile:GPR [(match_operand:GPR 2 "reg_or_0_operand" "rJ")
+			      (match_operand:GPR 3 "reg_or_0_operand" "rJ")
+			      (match_operand:GPR 4 "reg_or_0_operand" "rJ")
+			      (match_operand:GPR 5 "reg_or_0_operand"  "rJ")
+			      (match_operand:SI 6 "const_int_operand")] ;; model
+	 UNSPEC_SYNC_EXCHANGE))
+   (clobber (match_scratch:GPR 7 "=&r"))]
+  ""
+{
+  return "%G6\\n\\t"
+	 "1:\\n\\t"
+	 "ll.<amo>\\t%0,%1\\n\\t"
+	 "and\\t%7,%0,%z3\\n\\t"
+	 "or%i5\\t%7,%7,%5\\n\\t"
+	 "sc.<amo>\\t%7,%1\\n\\t"
+	 "beqz\\t%7,1b\\n\\t";
+}
+  [(set (attr "length") (const_int 20))])
+
 (define_expand "atomic_exchange<mode>"
   [(set (match_operand:SHORT 0 "register_operand")
 	(unspec_volatile:SHORT
@@ -459,9 +482,9 @@
   ""
 {
   union loongarch_gen_fn_ptrs generator;
-  generator.fn_7 = gen_atomic_cas_value_cmp_and_7_si;
+  generator.fn_7 = gen_atomic_cas_value_exchange_7_si;
   loongarch_expand_atomic_qihi (generator, operands[0], operands[1],
-				operands[1], operands[2], operands[3]);
+				const0_rtx, operands[2], operands[3]);
   DONE;
 })
 

gcc/testsuite/gcc.target/loongarch/pr107713-1.c

@@ -0,0 +1,50 @@
+/* { dg-do run } */
+/* { dg-require-effective-target pthread } */
+/* { dg-options "-pthread" } */
+
+#include <pthread.h>
+
+char x, x1, x2;
+
+void *
+work1 (void *)
+{
+  for (int i = 0; i < 100; i++)
+    x1 = __atomic_exchange_n (&x, x1, __ATOMIC_SEQ_CST);
+  return NULL;
+}
+
+void *
+work2 (void *)
+{
+  for (int i = 0; i < 100; i++)
+    x2 = __atomic_exchange_n (&x, x2, __ATOMIC_SEQ_CST);
+  return NULL;
+}
+
+void
+test (void)
+{
+  x = 0;
+  x1 = 1;
+  x2 = 2;
+  pthread_t w1, w2;
+  if (pthread_create (&w1, NULL, work1, NULL) != 0)
+    __builtin_abort ();
+  if (pthread_create (&w2, NULL, work2, NULL) != 0)
+    __builtin_abort ();
+  if (pthread_join (w1, NULL) != 0)
+    __builtin_abort ();
+  if (pthread_join (w2, NULL) != 0)
+    __builtin_abort ();
+  if ((x ^ x1 ^ x2) != 3)
+    __builtin_abort ();
+}
+
+int
+main ()
+{
+  int i;
+  for (i = 0; i < 10000; i++)
+    test ();
+}

gcc/testsuite/gcc.target/loongarch/pr107713-2.c

@@ -0,0 +1,9 @@
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+/* { dg-final { scan-assembler-times "beq|bne" 1 } } */
+
+char
+t (char *p, char x)
+{
+  return __atomic_exchange_n (p, x, __ATOMIC_RELAXED);
+}