Onward with our miniseries on lock-free many-producer\/single-consumer patterns. Today, we’re going to look at the case where you have a work queue where there can be multiple work items, and you need to perform them all, but each item is identical. <\/p>\n
For example, you may have a Buy It<\/i> button. Each time the user clicks the button, you want to run a transaction. But each button press is equivalent; all that’s important is the number of times the user pushed the button. <\/p>\n
Okay, that’s not a very good example, but it’ll have to do. <\/p>\n
One way of doing this is with a semaphore, where the number of tokens in the semaphore is the number of work items left to be done. But let’s stick with our current pattern where the producers need to manually wake the consumer, say with a message, and we want to minimize the number of times we need to perform the wake ritual. <\/p>\n
\nLONG WorkCount;\n\nvoid RequestWork()\n{\n if (InterlockedIncrement(&WorkCount) == 1) {\n \/\/ You provide the WakeConsumer() function.\n WakeConsumer();\n }\n}\n\n\/\/ You call this function when the consumer receives the\n\/\/ signal raised by WakeConsumer().\nvoid ConsumeWork()\n{\n while (InterlockedDecrementToZero(&WorkCount)) {\n DoSomeWork();\n }\n}\n\nbool InterlockedDecrementToZero(LONG volatile* value)\n{\n LONG original, result;\n do {\n original = *value;\n if (original == 0) return false;\n result = original - 1;\n } while (InterlockedCompareExchange(value, result,\n original) != original);\n return true;\n}\n<\/pre>\n