Will other threads see a variable entry saying "volatile" in a reasonable amount of time?

Question

When you ask about a more specific issue , I find this to be a key issue when people are not completely sure.

The following assumptions can be made:

• The processor uses a cache coherence protocol such as MESI (F) (examples: x86 / x86_64 and ARMv7mp) It is assumed that the variable
• has a size that is atomically written / read by the processor (aligned and proper word size)
• Variable declared volatile

Questions:

• If I write to a variable in one thread, will other threads see the change?
• What is the order of magnitude of the timeframe in which other threads will see the change?
• Do you know about architectures where cache coherence is not enough to provide transparency between processors / cross-core?

The question is NOT:

• Is it possible to use such a variable?
• reordering issues
• about c ++ 11 atomics

This may be considered a duplicate. In C / C ++, are there volatile variables that guarantee consistent semantics of betwen threads? and other similar questions, but I think that none of these clear requirements regarding the target architecture, which leads to a lot of confusion regarding various assumptions.

Answer 1

, --?

, . , - , Intel, Intel QPI, .

Wiki Intel QPI :

http://en.wikipedia.org/wiki/Intel_QuickPath_Interconnect

Answer 2

, ?

. , , , .

, ?

. .

, --?

, ++, . , , , "" ++.

, "". , (, , ).

Answer 3

, , . , ( ) . , . , , , , , , , , .

, , , .

Answer 4

x86/64:

, ?

. , /-.

, ?

, . , = 200 NUMA node. node, 2 node. . , +/- 50 .

60-70ns 3GHz node.

, --?

, - . , , , , , x86, . , .

Answer 5

, , volatile "" .

, ( ) .

() PowerPC , - volatile. , , ( ).

, , , (, , ..) , , .

volatile ++ , --. volatile, , , , ( ). - .

, . , , , , ( ..), .

, - volatile, . , , , . ( ) , , , .

++ ( , , , ), , , , , volatile. - - - , . (, , , , ).