[Gc]: libatomic-ops patch for Renesas SH

Nobuhiro Iwamatsu iwamatsu at nigauri.org
Tue Oct 6 17:42:48 PDT 2009


Hi,

Sorry that reply is late.

2009/10/3 Boehm, Hans <hans.boehm at hp.com>:
>
>> - AO_test_and_set_full should return values within
>> AO_TS_VAL_t type (i.e., AO_TS_CLEAR(0) and AO_TS_SET(0xff)
>> instead of 0 and 0x80).
> Unless I'm missing something, this looks to me like it's not broken.  The AO_TS_SET and AO_TS_CLEAR values are allowed to be hardware dependent.  But I admit, it's not clear to me why the nonstandard value helps here.  Does anything break if the redeinition of AO_TS_SET is just removed?

Thank your for your comment.

Redeinition of AO_TS_SET is mistake.
Please remove these line.

--- a/src/atomic_ops/sysdeps/gcc/sh.h	2009-10-07 09:37:20.000000000 +0900
+++ b/src/atomic_ops/sysdeps/gcc/sh.h	2009-10-07 09:37:30.000000000 +0900
@@ -18,10 +18,6 @@
 /* sh has tas.b(byte) only */
 #include "../test_and_set_t_is_char.h"

-/* FIXME: Is AO_TS_SET redefining needed? */
-/* #undef AO_TS_SET */
-/* #define AO_TS_SET (1<<7) */
-
 AO_INLINE AO_TS_VAL_t
 AO_test_and_set_full(volatile AO_TS_t *addr)
 {


Best regards,
  Nobuhiro

-- 
Nobuhiro Iwamatsu



More information about the Gc mailing list