public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [trans-mem] Test that nested txns started from pure/unsafe code work correctly
@ 2011-08-03 11:52 Torvald Riegel
  2011-08-04 15:54 ` Richard Henderson
  0 siblings, 1 reply; 2+ messages in thread
From: Torvald Riegel @ 2011-08-03 11:52 UTC (permalink / raw)
  To: GCC Patches; +Cc: Aldy Hernandez, Richard Henderson

[-- Attachment #1: Type: text/plain, Size: 376 bytes --]

Tests that new transactions can be started from both transaction_pure
and transaction_unsafe code. This also checks proper handling of
reentrant nesting in the serial_lock implementation (reentrant in the
sense that we go from transactional to nontransactional to transactional
code).
This test currently does not compile due to a GCC bug (no bug report
yet).

OK for branch?

[-- Attachment #2: patch6 --]
[-- Type: text/plain, Size: 1721 bytes --]

commit c1eafd7cfbdb71dadb3ac5f797fb2a596026f1be
Author: Torvald Riegel <triegel@redhat.com>
Date:   Mon Aug 1 15:43:05 2011 +0200

    Test that nested txns started from pure/unsafe code work correctly.
    
    	* testsuite/libitm.c/reentrant.c: New file.

diff --git a/libitm/testsuite/libitm.c/reentrant.c b/libitm/testsuite/libitm.c/reentrant.c
new file mode 100644
index 0000000..aeb9a0e
--- /dev/null
+++ b/libitm/testsuite/libitm.c/reentrant.c
@@ -0,0 +1,64 @@
+/* Tests that new transactions can be started from both transaction_pure and
+   transaction_unsafe code. This also requires proper handling of reentrant
+   nesting in the serial_lock implementation. */
+
+#include <stdlib.h>
+#include <pthread.h>
+#include <libitm.h>
+
+int x = 0;
+
+int __attribute__((transaction_pure)) pure(int i)
+{
+  __transaction {
+    x++;
+  }
+  if (_ITM_inTransaction() == outsideTransaction)
+    abort();
+  return i+1;
+}
+
+int __attribute__((transaction_unsafe)) unsafe(int i)
+{
+  if (_ITM_inTransaction() != inIrrevocableTransaction)
+    abort();
+  __transaction {
+    x++;
+  }
+  if (_ITM_inTransaction() != inIrrevocableTransaction)
+    abort();
+  return i+1;
+}
+
+static void *thread (void *dummy __attribute__((unused)))
+{
+  __transaction {
+    pure(1);
+  }
+  __transaction[[relaxed]] {
+    unsafe(1);
+  }
+  return 0;
+}
+
+int main()
+{
+  pthread_t pt;
+  int r = 0;
+
+  __transaction {
+    r += pure(1) + x;
+  }
+  __transaction[[relaxed]] {
+    r += unsafe(1) + x;
+  }
+  if (r != 7)
+    abort();
+
+  // Spawn a new thread to check that the serial lock is not held.
+  pthread_create(&pt, NULL, thread, NULL);
+  pthread_join(pt, NULL);
+  if (x != 4)
+    abort();
+  return 0;
+}

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [trans-mem] Test that nested txns started from pure/unsafe code work correctly
  2011-08-03 11:52 [trans-mem] Test that nested txns started from pure/unsafe code work correctly Torvald Riegel
@ 2011-08-04 15:54 ` Richard Henderson
  0 siblings, 0 replies; 2+ messages in thread
From: Richard Henderson @ 2011-08-04 15:54 UTC (permalink / raw)
  To: Torvald Riegel; +Cc: GCC Patches, Aldy Hernandez

On 08/03/2011 04:51 AM, Torvald Riegel wrote:
>     Test that nested txns started from pure/unsafe code work correctly.
>     
>     	* testsuite/libitm.c/reentrant.c: New file.

Ok.


r~

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2011-08-04 15:54 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-08-03 11:52 [trans-mem] Test that nested txns started from pure/unsafe code work correctly Torvald Riegel
2011-08-04 15:54 ` Richard Henderson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).