Re: [Maria-developers] Rev 2740: Group commit for maria storage engine. in file:///Users/bell/maria/bzr/work-maria-5.2-groupcommit/

23 Feb 2010

      Hi!
...
...
...
...
...
"Sergei" == Sergei Golubchik  writes:
Sergei> Hi, Michael!
Sergei> On Feb 11, Michael Widenius wrote:
...
...
Yes, but if you are only looking at one value, without regard of the
other, then you don't need to use my_atomic_load32()
Sergei> if 32-bit reads are always atomic - yes.
Sergei> I suppose we can assume that they are.
...
...
Another problem is that if the variables are related in any way,
atomic_load32 is not good enough as you can get 'a' and 'b' from
different instances in time.
for example:
a=b= 0;
T1: my_atomic_store32(&a, 1);
T1: my_atomic_store32(&b, 2);
T2: my_atomic_store32(&a, 3);
T2: my_atomic_store32(&b, 4);
T3: a1= my_atomic_load32(&a);
T3: b1= my_atomic_load32(&b);
this means that you can get the following values for a1 and b1:
a1= 0  b1= 0
a1= 0  b1= 2
a1= 1  b1= 2
a1= 1  b1= 4
a1= 3  b1= 2
a1= 3  b1= 4
Sergei> and also

Sergei>   a1=0   b1=4
Sergei>   a1=1   b1=0

ok, I missed the above one.

Sergei>   a1=3   b1=0

Ok, that is right. However, my intention was actually to have T2 to be
T1 (ie same process), in which case this option would be impossible.

Sergei> all the 9 variants are possible.

Sanja, when considering the above, are you sure your usage of
my_atomic_load/my_atmoic_store is correct?

For example:

+      my_atomic_rwlock_rdlock(&soft_sync_rwl);
+      min= my_atomic_load32(&soft_sync_min);
+      max= my_atomic_load32(&soft_sync_max);
+      sync_request= my_atomic_load32(&soft_need_sync);
+      my_atomic_store32(&soft_sync_min, max);
+      my_atomic_store32(&soft_need_sync, 0);
+      my_atomic_rwlock_rdunlock(&soft_sync_rwl);

In the above code, there is no gurantee that you read or write
related soft_sync_min and soft_sync_max values.
(See above table).

Also, as we concluded with Sergei, any code like:

+  my_atomic_rwlock_rdlock(&soft_sync_rwl);
+  min= my_atomic_load32(&soft_sync_min);
+  my_atomic_rwlock_rdunlock(&soft_sync_rwl);

There is no reason to have an my_atomic_rwlock instruction.

Regards,
Monty